id
int64 0
458k
| file_name
stringlengths 4
119
| file_path
stringlengths 14
227
| content
stringlengths 24
9.96M
| size
int64 24
9.96M
| language
stringclasses 1
value | extension
stringclasses 14
values | total_lines
int64 1
219k
| avg_line_length
float64 2.52
4.63M
| max_line_length
int64 5
9.91M
| alphanum_fraction
float64 0
1
| repo_name
stringlengths 7
101
| repo_stars
int64 100
139k
| repo_forks
int64 0
26.4k
| repo_open_issues
int64 0
2.27k
| repo_license
stringclasses 12
values | repo_extraction_date
stringclasses 433
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2,288,300 | attaches.js | Nigel2392_wagtail_editorjs/wagtail_editorjs/static/wagtail_editorjs/js/tools/attaches.js | class CSRFAttachesTool extends window.AttachesTool {
constructor({ data, config, api, readOnly }) {
config['additionalRequestHeaders'] = {
'X-CSRFToken': document.querySelector('input[name="csrfmiddlewaretoken"]').value
};
super({ data, config, api, readOnly });
}
}
window.CSRFAttachesTool = CSRFAttachesTool; | 355 | Python | .tac | 9 | 33.555556 | 92 | 0.682081 | Nigel2392/wagtail_editorjs | 8 | 0 | 3 | GPL-2.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,301 | attaches.js | Nigel2392_wagtail_editorjs/wagtail_editorjs/static/wagtail_editorjs/vendor/editorjs/tools/attaches.js | /**
* Skipped minification because the original files appears to be already minified.
* Original file: /npm/@editorjs/[email protected]/dist/bundle.js
*
* Do NOT use SRI with dynamically generated files! More information: https://www.jsdelivr.com/using-sri-with-dynamic-files
*/
!function(e,t){"object"==typeof exports&&"object"==typeof module?module.exports=t():"function"==typeof define&&define.amd?define([],t):"object"==typeof exports?exports.AttachesTool=t():e.AttachesTool=t()}(window,(function(){return function(e){var t={};function n(r){if(t[r])return t[r].exports;var o=t[r]={i:r,l:!1,exports:{}};return e[r].call(o.exports,o,o.exports,n),o.l=!0,o.exports}return n.m=e,n.c=t,n.d=function(e,t,r){n.o(e,t)||Object.defineProperty(e,t,{enumerable:!0,get:r})},n.r=function(e){"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},n.t=function(e,t){if(1&t&&(e=n(e)),8&t)return e;if(4&t&&"object"==typeof e&&e&&e.__esModule)return e;var r=Object.create(null);if(n.r(r),Object.defineProperty(r,"default",{enumerable:!0,value:e}),2&t&&"string"!=typeof e)for(var o in e)n.d(r,o,function(t){return e[t]}.bind(null,o));return r},n.n=function(e){var t=e&&e.__esModule?function(){return e.default}:function(){return e};return n.d(t,"a",t),t},n.o=function(e,t){return Object.prototype.hasOwnProperty.call(e,t)},n.p="/",n(n.s=6)}([function(e,t,n){window,e.exports=function(e){var t={};function n(r){if(t[r])return t[r].exports;var o=t[r]={i:r,l:!1,exports:{}};return e[r].call(o.exports,o,o.exports,n),o.l=!0,o.exports}return n.m=e,n.c=t,n.d=function(e,t,r){n.o(e,t)||Object.defineProperty(e,t,{enumerable:!0,get:r})},n.r=function(e){"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},n.t=function(e,t){if(1&t&&(e=n(e)),8&t)return e;if(4&t&&"object"==typeof e&&e&&e.__esModule)return e;var r=Object.create(null);if(n.r(r),Object.defineProperty(r,"default",{enumerable:!0,value:e}),2&t&&"string"!=typeof e)for(var o in e)n.d(r,o,function(t){return e[t]}.bind(null,o));return r},n.n=function(e){var t=e&&e.__esModule?function(){return e.default}:function(){return e};return n.d(t,"a",t),t},n.o=function(e,t){return Object.prototype.hasOwnProperty.call(e,t)},n.p="",n(n.s=3)}([function(e,t){var n;n=function(){return this}();try{n=n||new Function("return this")()}catch(e){"object"==typeof window&&(n=window)}e.exports=n},function(e,t,n){"use strict";(function(e){var r=n(2),o=setTimeout;function i(){}function a(e){if(!(this instanceof a))throw new TypeError("Promises must be constructed via new");if("function"!=typeof e)throw new TypeError("not a function");this._state=0,this._handled=!1,this._value=void 0,this._deferreds=[],d(e,this)}function s(e,t){for(;3===e._state;)e=e._value;0!==e._state?(e._handled=!0,a._immediateFn((function(){var n=1===e._state?t.onFulfilled:t.onRejected;if(null!==n){var r;try{r=n(e._value)}catch(e){return void l(t.promise,e)}c(t.promise,r)}else(1===e._state?c:l)(t.promise,e._value)}))):e._deferreds.push(t)}function c(e,t){try{if(t===e)throw new TypeError("A promise cannot be resolved with itself.");if(t&&("object"==typeof t||"function"==typeof t)){var n=t.then;if(t instanceof a)return e._state=3,e._value=t,void u(e);if("function"==typeof n)return void d((r=n,o=t,function(){r.apply(o,arguments)}),e)}e._state=1,e._value=t,u(e)}catch(t){l(e,t)}var r,o}function l(e,t){e._state=2,e._value=t,u(e)}function u(e){2===e._state&&0===e._deferreds.length&&a._immediateFn((function(){e._handled||a._unhandledRejectionFn(e._value)}));for(var t=0,n=e._deferreds.length;t<n;t++)s(e,e._deferreds[t]);e._deferreds=null}function f(e,t,n){this.onFulfilled="function"==typeof e?e:null,this.onRejected="function"==typeof t?t:null,this.promise=n}function d(e,t){var n=!1;try{e((function(e){n||(n=!0,c(t,e))}),(function(e){n||(n=!0,l(t,e))}))}catch(e){if(n)return;n=!0,l(t,e)}}a.prototype.catch=function(e){return this.then(null,e)},a.prototype.then=function(e,t){var n=new this.constructor(i);return s(this,new f(e,t,n)),n},a.prototype.finally=r.a,a.all=function(e){return new a((function(t,n){if(!e||void 0===e.length)throw new TypeError("Promise.all accepts an array");var r=Array.prototype.slice.call(e);if(0===r.length)return t([]);var o=r.length;function i(e,a){try{if(a&&("object"==typeof a||"function"==typeof a)){var s=a.then;if("function"==typeof s)return void s.call(a,(function(t){i(e,t)}),n)}r[e]=a,0==--o&&t(r)}catch(e){n(e)}}for(var a=0;a<r.length;a++)i(a,r[a])}))},a.resolve=function(e){return e&&"object"==typeof e&&e.constructor===a?e:new a((function(t){t(e)}))},a.reject=function(e){return new a((function(t,n){n(e)}))},a.race=function(e){return new a((function(t,n){for(var r=0,o=e.length;r<o;r++)e[r].then(t,n)}))},a._immediateFn="function"==typeof e&&function(t){e(t)}||function(e){o(e,0)},a._unhandledRejectionFn=function(e){"undefined"!=typeof console&&console&&console.warn("Possible Unhandled Promise Rejection:",e)},t.a=a}).call(this,n(5).setImmediate)},function(e,t,n){"use strict";t.a=function(e){var t=this.constructor;return this.then((function(n){return t.resolve(e()).then((function(){return n}))}),(function(n){return t.resolve(e()).then((function(){return t.reject(n)}))}))}},function(e,t,n){"use strict";function r(e){return(r="function"==typeof Symbol&&"symbol"==typeof Symbol.iterator?function(e){return typeof e}:function(e){return e&&"function"==typeof Symbol&&e.constructor===Symbol&&e!==Symbol.prototype?"symbol":typeof e})(e)}n(4);var o,i,a,s,c,l,u,f=n(8),d=(i=function(e){return new Promise((function(t,n){e=s(e),(e=c(e)).beforeSend&&e.beforeSend();var r=window.XMLHttpRequest?new window.XMLHttpRequest:new window.ActiveXObject("Microsoft.XMLHTTP");r.open(e.method,e.url),r.setRequestHeader("X-Requested-With","XMLHttpRequest"),Object.keys(e.headers).forEach((function(t){var n=e.headers[t];r.setRequestHeader(t,n)}));var o=e.ratio;r.upload.addEventListener("progress",(function(t){var n=Math.round(t.loaded/t.total*100),r=Math.ceil(n*o/100);e.progress(Math.min(r,100))}),!1),r.addEventListener("progress",(function(t){var n=Math.round(t.loaded/t.total*100),r=Math.ceil(n*(100-o)/100)+o;e.progress(Math.min(r,100))}),!1),r.onreadystatechange=function(){if(4===r.readyState){var e=r.response;try{e=JSON.parse(e)}catch(e){}var o=f.parseHeaders(r.getAllResponseHeaders()),i={body:e,code:r.status,headers:o};u(r.status)?t(i):n(i)}},r.send(e.data)}))},a=function(e){return e.method="POST",i(e)},s=function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{};if(e.url&&"string"!=typeof e.url)throw new Error("Url must be a string");if(e.url=e.url||"",e.method&&"string"!=typeof e.method)throw new Error("`method` must be a string or null");if(e.method=e.method?e.method.toUpperCase():"GET",e.headers&&"object"!==r(e.headers))throw new Error("`headers` must be an object or null");if(e.headers=e.headers||{},e.type&&("string"!=typeof e.type||!Object.values(o).includes(e.type)))throw new Error("`type` must be taken from module's «contentType» library");if(e.progress&&"function"!=typeof e.progress)throw new Error("`progress` must be a function or null");if(e.progress=e.progress||function(e){},e.beforeSend=e.beforeSend||function(e){},e.ratio&&"number"!=typeof e.ratio)throw new Error("`ratio` must be a number");if(e.ratio<0||e.ratio>100)throw new Error("`ratio` must be in a 0-100 interval");if(e.ratio=e.ratio||90,e.accept&&"string"!=typeof e.accept)throw new Error("`accept` must be a string with a list of allowed mime-types");if(e.accept=e.accept||"*/*",e.multiple&&"boolean"!=typeof e.multiple)throw new Error("`multiple` must be a true or false");if(e.multiple=e.multiple||!1,e.fieldName&&"string"!=typeof e.fieldName)throw new Error("`fieldName` must be a string");return e.fieldName=e.fieldName||"files",e},c=function(e){switch(e.method){case"GET":var t=l(e.data,o.URLENCODED);delete e.data,e.url=/\?/.test(e.url)?e.url+"&"+t:e.url+"?"+t;break;case"POST":case"PUT":case"DELETE":case"UPDATE":var n=function(){return(arguments.length>0&&void 0!==arguments[0]?arguments[0]:{}).type||o.JSON}(e);(f.isFormData(e.data)||f.isFormElement(e.data))&&(n=o.FORM),e.data=l(e.data,n),n!==d.contentType.FORM&&(e.headers["content-type"]=n)}return e},l=function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{};switch(arguments.length>1?arguments[1]:void 0){case o.URLENCODED:return f.urlEncode(e);case o.JSON:return f.jsonEncode(e);case o.FORM:return f.formEncode(e);default:return e}},u=function(e){return e>=200&&e<300},{contentType:o={URLENCODED:"application/x-www-form-urlencoded; charset=utf-8",FORM:"multipart/form-data",JSON:"application/json; charset=utf-8"},request:i,get:function(e){return e.method="GET",i(e)},post:a,transport:function(e){return e=s(e),f.selectFiles(e).then((function(t){for(var n=new FormData,r=0;r<t.length;r++)n.append(e.fieldName,t[r],t[r].name);f.isObject(e.data)&&Object.keys(e.data).forEach((function(t){var r=e.data[t];n.append(t,r)}));var o=e.beforeSend;return e.beforeSend=function(){return o(t)},e.data=n,a(e)}))},selectFiles:function(e){return delete(e=s(e)).beforeSend,f.selectFiles(e)}});e.exports=d},function(e,t,n){"use strict";n.r(t);var r=n(1);window.Promise=window.Promise||r.a},function(e,t,n){(function(e){var r=void 0!==e&&e||"undefined"!=typeof self&&self||window,o=Function.prototype.apply;function i(e,t){this._id=e,this._clearFn=t}t.setTimeout=function(){return new i(o.call(setTimeout,r,arguments),clearTimeout)},t.setInterval=function(){return new i(o.call(setInterval,r,arguments),clearInterval)},t.clearTimeout=t.clearInterval=function(e){e&&e.close()},i.prototype.unref=i.prototype.ref=function(){},i.prototype.close=function(){this._clearFn.call(r,this._id)},t.enroll=function(e,t){clearTimeout(e._idleTimeoutId),e._idleTimeout=t},t.unenroll=function(e){clearTimeout(e._idleTimeoutId),e._idleTimeout=-1},t._unrefActive=t.active=function(e){clearTimeout(e._idleTimeoutId);var t=e._idleTimeout;t>=0&&(e._idleTimeoutId=setTimeout((function(){e._onTimeout&&e._onTimeout()}),t))},n(6),t.setImmediate="undefined"!=typeof self&&self.setImmediate||void 0!==e&&e.setImmediate||this&&this.setImmediate,t.clearImmediate="undefined"!=typeof self&&self.clearImmediate||void 0!==e&&e.clearImmediate||this&&this.clearImmediate}).call(this,n(0))},function(e,t,n){(function(e,t){!function(e,n){"use strict";if(!e.setImmediate){var r,o,i,a,s,c=1,l={},u=!1,f=e.document,d=Object.getPrototypeOf&&Object.getPrototypeOf(e);d=d&&d.setTimeout?d:e,"[object process]"==={}.toString.call(e.process)?r=function(e){t.nextTick((function(){h(e)}))}:function(){if(e.postMessage&&!e.importScripts){var t=!0,n=e.onmessage;return e.onmessage=function(){t=!1},e.postMessage("","*"),e.onmessage=n,t}}()?(a="setImmediate$"+Math.random()+"$",s=function(t){t.source===e&&"string"==typeof t.data&&0===t.data.indexOf(a)&&h(+t.data.slice(a.length))},e.addEventListener?e.addEventListener("message",s,!1):e.attachEvent("onmessage",s),r=function(t){e.postMessage(a+t,"*")}):e.MessageChannel?((i=new MessageChannel).port1.onmessage=function(e){h(e.data)},r=function(e){i.port2.postMessage(e)}):f&&"onreadystatechange"in f.createElement("script")?(o=f.documentElement,r=function(e){var t=f.createElement("script");t.onreadystatechange=function(){h(e),t.onreadystatechange=null,o.removeChild(t),t=null},o.appendChild(t)}):r=function(e){setTimeout(h,0,e)},d.setImmediate=function(e){"function"!=typeof e&&(e=new Function(""+e));for(var t=new Array(arguments.length-1),n=0;n<t.length;n++)t[n]=arguments[n+1];var o={callback:e,args:t};return l[c]=o,r(c),c++},d.clearImmediate=p}function p(e){delete l[e]}function h(e){if(u)setTimeout(h,0,e);else{var t=l[e];if(t){u=!0;try{!function(e){var t=e.callback,n=e.args;switch(n.length){case 0:t();break;case 1:t(n[0]);break;case 2:t(n[0],n[1]);break;case 3:t(n[0],n[1],n[2]);break;default:t.apply(void 0,n)}}(t)}finally{p(e),u=!1}}}}}("undefined"==typeof self?void 0===e?this:e:self)}).call(this,n(0),n(7))},function(e,t){var n,r,o=e.exports={};function i(){throw new Error("setTimeout has not been defined")}function a(){throw new Error("clearTimeout has not been defined")}function s(e){if(n===setTimeout)return setTimeout(e,0);if((n===i||!n)&&setTimeout)return n=setTimeout,setTimeout(e,0);try{return n(e,0)}catch(t){try{return n.call(null,e,0)}catch(t){return n.call(this,e,0)}}}!function(){try{n="function"==typeof setTimeout?setTimeout:i}catch(e){n=i}try{r="function"==typeof clearTimeout?clearTimeout:a}catch(e){r=a}}();var c,l=[],u=!1,f=-1;function d(){u&&c&&(u=!1,c.length?l=c.concat(l):f=-1,l.length&&p())}function p(){if(!u){var e=s(d);u=!0;for(var t=l.length;t;){for(c=l,l=[];++f<t;)c&&c[f].run();f=-1,t=l.length}c=null,u=!1,function(e){if(r===clearTimeout)return clearTimeout(e);if((r===a||!r)&&clearTimeout)return r=clearTimeout,clearTimeout(e);try{r(e)}catch(t){try{return r.call(null,e)}catch(t){return r.call(this,e)}}}(e)}}function h(e,t){this.fun=e,this.array=t}function m(){}o.nextTick=function(e){var t=new Array(arguments.length-1);if(arguments.length>1)for(var n=1;n<arguments.length;n++)t[n-1]=arguments[n];l.push(new h(e,t)),1!==l.length||u||s(p)},h.prototype.run=function(){this.fun.apply(null,this.array)},o.title="browser",o.browser=!0,o.env={},o.argv=[],o.version="",o.versions={},o.on=m,o.addListener=m,o.once=m,o.off=m,o.removeListener=m,o.removeAllListeners=m,o.emit=m,o.prependListener=m,o.prependOnceListener=m,o.listeners=function(e){return[]},o.binding=function(e){throw new Error("process.binding is not supported")},o.cwd=function(){return"/"},o.chdir=function(e){throw new Error("process.chdir is not supported")},o.umask=function(){return 0}},function(e,t,n){function r(e,t){for(var n=0;n<t.length;n++){var r=t[n];r.enumerable=r.enumerable||!1,r.configurable=!0,"value"in r&&(r.writable=!0),Object.defineProperty(e,r.key,r)}}var o=n(9);e.exports=function(){function e(){!function(e,t){if(!(e instanceof t))throw new TypeError("Cannot call a class as a function")}(this,e)}var t,n;return t=e,(n=[{key:"urlEncode",value:function(e){return o(e)}},{key:"jsonEncode",value:function(e){return JSON.stringify(e)}},{key:"formEncode",value:function(e){if(this.isFormData(e))return e;if(this.isFormElement(e))return new FormData(e);if(this.isObject(e)){var t=new FormData;return Object.keys(e).forEach((function(n){var r=e[n];t.append(n,r)})),t}throw new Error("`data` must be an instance of Object, FormData or <FORM> HTMLElement")}},{key:"isObject",value:function(e){return"[object Object]"===Object.prototype.toString.call(e)}},{key:"isFormData",value:function(e){return e instanceof FormData}},{key:"isFormElement",value:function(e){return e instanceof HTMLFormElement}},{key:"selectFiles",value:function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:{};return new Promise((function(t,n){var r=document.createElement("INPUT");r.type="file",e.multiple&&r.setAttribute("multiple","multiple"),e.accept&&r.setAttribute("accept",e.accept),r.style.display="none",document.body.appendChild(r),r.addEventListener("change",(function(e){var n=e.target.files;t(n),document.body.removeChild(r)}),!1),r.click()}))}},{key:"parseHeaders",value:function(e){var t=e.trim().split(/[\r\n]+/),n={};return t.forEach((function(e){var t=e.split(": "),r=t.shift(),o=t.join(": ");r&&(n[r]=o)})),n}}])&&r(t,n),e}()},function(e,t){var n=function(e){return encodeURIComponent(e).replace(/[!'()*]/g,escape).replace(/%20/g,"+")},r=function(e,t,o,i){return t=t||null,o=o||"&",i=i||null,e?function(e){for(var t=new Array,n=0;n<e.length;n++)e[n]&&t.push(e[n]);return t}(Object.keys(e).map((function(a){var s,c,l=a;if(i&&(l=i+"["+l+"]"),"object"==typeof e[a]&&null!==e[a])s=r(e[a],null,o,l);else{t&&(c=l,l=!isNaN(parseFloat(c))&&isFinite(c)?t+Number(l):l);var u=e[a];u=(u=0===(u=!1===(u=!0===u?"1":u)?"0":u)?"0":u)||"",s=n(l)+"="+n(u)}return s}))).join(o).replace(/[!'()*]/g,""):""};e.exports=r}])},function(e,t,n){var r=n(2);"string"==typeof r&&(r=[[e.i,r,""]]);var o={hmr:!0,transform:void 0,insertInto:void 0};n(4)(r,o);r.locals&&(e.exports=r.locals)},function(e,t,n){(e.exports=n(3)(!1)).push([e.i,".cdx-attaches {\n --color-line: #EFF0F1;\n --color-bg: #fff;\n --color-bg-secondary: #F8F8F8;\n --color-bg-secondary--hover: #f2f2f2;\n --color-text-secondary: #707684;\n}\n\n .cdx-attaches--with-file {\n display: flex;\n align-items: center;\n padding: 10px 12px;\n border: 1px solid var(--color-line);\n border-radius: 7px;\n background: var(--color-bg);\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-info {\n display: grid;\n grid-gap: 4px;\n max-width: calc(100% - 80px);\n margin: auto 0;\n flex-grow: 2;\n }\n\n .cdx-attaches--with-file .cdx-attaches__download-button {\n display: flex;\n align-items: center;\n background: var(--color-bg-secondary);\n padding: 6px;\n border-radius: 6px;\n margin: auto 0 auto auto;\n }\n\n .cdx-attaches--with-file .cdx-attaches__download-button:hover {\n background: var(--color-bg-secondary--hover);\n }\n\n .cdx-attaches--with-file .cdx-attaches__download-button svg {\n width: 20px;\n height: 20px;\n fill: none;\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-icon {\n position: relative;\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-icon-background {\n background-color: #333;\n\n width: 27px;\n height: 30px;\n margin-right: 12px;\n border-radius: 8px;\n display: flex;\n align-items: center;\n justify-content: center;\n }\n\n @supports(-webkit-mask-box-image: url('')){\n\n .cdx-attaches--with-file .cdx-attaches__file-icon-background {\n border-radius: 0;\n -webkit-mask-box-image: url(\"data:image/svg+xml,%3Csvg width='24' height='24' viewBox='0 0 24 24' fill='none' xmlns='http://www.w3.org/2000/svg'%3E%3Cpath d='M0 10.3872C0 1.83334 1.83334 0 10.3872 0H13.6128C22.1667 0 24 1.83334 24 10.3872V13.6128C24 22.1667 22.1667 24 13.6128 24H10.3872C1.83334 24 0 22.1667 0 13.6128V10.3872Z' fill='black'/%3E%3C/svg%3E%0A\") 48% 41% 37.9% 53.3%\n };\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-icon-label {\n position: absolute;\n left: 3px;\n top: 11px;\n background: inherit;\n text-transform: uppercase;\n line-height: 1em;\n color: #fff;\n padding: 1px 2px;\n border-radius: 3px;\n font-size: 10px;\n font-weight: bold;\n /* box-shadow: 0px 4px 6px rgba(0, 0, 0, 0.22); */\n font-family: ui-monospace,SFMono-Regular,SF Mono,Menlo,Consolas,Liberation Mono,monospace;\n letter-spacing: 0.02em;\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-icon svg {\n width: 20px;\n height: 20px;\n }\n\n .cdx-attaches--with-file .cdx-attaches__file-icon path {\n stroke: #fff;\n }\n\n .cdx-attaches--with-file .cdx-attaches__size {\n color: var(--color-text-secondary);\n font-size: 12px;\n line-height: 1em;\n }\n\n .cdx-attaches--with-file .cdx-attaches__size::after {\n content: attr(data-size);\n margin-left: 0.2em;\n }\n\n .cdx-attaches--with-file .cdx-attaches__title {\n white-space: nowrap;\n text-overflow: ellipsis;\n overflow: hidden;\n outline: none;\n max-width: 90%;\n font-size: 14px;\n font-weight: 500;\n line-height: 1em;\n }\n\n .cdx-attaches--with-file .cdx-attaches__title:empty::before {\n content: attr(data-placeholder);\n color: #7b7e89;\n }\n\n .cdx-attaches--loading .cdx-attaches__title,\n .cdx-attaches--loading .cdx-attaches__file-icon,\n .cdx-attaches--loading .cdx-attaches__size,\n .cdx-attaches--loading .cdx-attaches__download-button,\n .cdx-attaches--loading .cdx-attaches__button {\n opacity: 0;\n font-size: 0;\n }\n\n .cdx-attaches__button {\n display: flex;\n align-items: center;\n justify-content: center;\n color: #000;\n border-radius: 7px;\n font-weight: 500;\n }\n\n .cdx-attaches__button svg {\n margin-top: 0;\n }\n",""])},function(e,t){e.exports=function(e){var t=[];return t.toString=function(){return this.map((function(t){var n=function(e,t){var n=e[1]||"",r=e[3];if(!r)return n;if(t&&"function"==typeof btoa){var o=(a=r,"/*# sourceMappingURL=data:application/json;charset=utf-8;base64,"+btoa(unescape(encodeURIComponent(JSON.stringify(a))))+" */"),i=r.sources.map((function(e){return"/*# sourceURL="+r.sourceRoot+e+" */"}));return[n].concat(i).concat([o]).join("\n")}var a;return[n].join("\n")}(t,e);return t[2]?"@media "+t[2]+"{"+n+"}":n})).join("")},t.i=function(e,n){"string"==typeof e&&(e=[[null,e,""]]);for(var r={},o=0;o<this.length;o++){var i=this[o][0];"number"==typeof i&&(r[i]=!0)}for(o=0;o<e.length;o++){var a=e[o];"number"==typeof a[0]&&r[a[0]]||(n&&!a[2]?a[2]=n:n&&(a[2]="("+a[2]+") and ("+n+")"),t.push(a))}},t}},function(e,t,n){var r,o,i={},a=(r=function(){return window&&document&&document.all&&!window.atob},function(){return void 0===o&&(o=r.apply(this,arguments)),o}),s=function(e){return document.querySelector(e)},c=function(e){var t={};return function(e){if("function"==typeof e)return e();if(void 0===t[e]){var n=s.call(this,e);if(window.HTMLIFrameElement&&n instanceof window.HTMLIFrameElement)try{n=n.contentDocument.head}catch(e){n=null}t[e]=n}return t[e]}}(),l=null,u=0,f=[],d=n(5);function p(e,t){for(var n=0;n<e.length;n++){var r=e[n],o=i[r.id];if(o){o.refs++;for(var a=0;a<o.parts.length;a++)o.parts[a](r.parts[a]);for(;a<r.parts.length;a++)o.parts.push(b(r.parts[a],t))}else{var s=[];for(a=0;a<r.parts.length;a++)s.push(b(r.parts[a],t));i[r.id]={id:r.id,refs:1,parts:s}}}}function h(e,t){for(var n=[],r={},o=0;o<e.length;o++){var i=e[o],a=t.base?i[0]+t.base:i[0],s={css:i[1],media:i[2],sourceMap:i[3]};r[a]?r[a].parts.push(s):n.push(r[a]={id:a,parts:[s]})}return n}function m(e,t){var n=c(e.insertInto);if(!n)throw new Error("Couldn't find a style target. This probably means that the value for the 'insertInto' parameter is invalid.");var r=f[f.length-1];if("top"===e.insertAt)r?r.nextSibling?n.insertBefore(t,r.nextSibling):n.appendChild(t):n.insertBefore(t,n.firstChild),f.push(t);else if("bottom"===e.insertAt)n.appendChild(t);else{if("object"!=typeof e.insertAt||!e.insertAt.before)throw new Error("[Style Loader]\n\n Invalid value for parameter 'insertAt' ('options.insertAt') found.\n Must be 'top', 'bottom', or Object.\n (https://github.com/webpack-contrib/style-loader#insertat)\n");var o=c(e.insertInto+" "+e.insertAt.before);n.insertBefore(t,o)}}function v(e){if(null===e.parentNode)return!1;e.parentNode.removeChild(e);var t=f.indexOf(e);t>=0&&f.splice(t,1)}function y(e){var t=document.createElement("style");return void 0===e.attrs.type&&(e.attrs.type="text/css"),g(t,e.attrs),m(e,t),t}function g(e,t){Object.keys(t).forEach((function(n){e.setAttribute(n,t[n])}))}function b(e,t){var n,r,o,i;if(t.transform&&e.css){if(!(i=t.transform(e.css)))return function(){};e.css=i}if(t.singleton){var a=u++;n=l||(l=y(t)),r=_.bind(null,n,a,!1),o=_.bind(null,n,a,!0)}else e.sourceMap&&"function"==typeof URL&&"function"==typeof URL.createObjectURL&&"function"==typeof URL.revokeObjectURL&&"function"==typeof Blob&&"function"==typeof btoa?(n=function(e){var t=document.createElement("link");return void 0===e.attrs.type&&(e.attrs.type="text/css"),e.attrs.rel="stylesheet",g(t,e.attrs),m(e,t),t}(t),r=E.bind(null,n,t),o=function(){v(n),n.href&&URL.revokeObjectURL(n.href)}):(n=y(t),r=S.bind(null,n),o=function(){v(n)});return r(e),function(t){if(t){if(t.css===e.css&&t.media===e.media&&t.sourceMap===e.sourceMap)return;r(e=t)}else o()}}e.exports=function(e,t){if("undefined"!=typeof DEBUG&&DEBUG&&"object"!=typeof document)throw new Error("The style-loader cannot be used in a non-browser environment");(t=t||{}).attrs="object"==typeof t.attrs?t.attrs:{},t.singleton||"boolean"==typeof t.singleton||(t.singleton=a()),t.insertInto||(t.insertInto="head"),t.insertAt||(t.insertAt="bottom");var n=h(e,t);return p(n,t),function(e){for(var r=[],o=0;o<n.length;o++){var a=n[o];(s=i[a.id]).refs--,r.push(s)}e&&p(h(e,t),t);for(o=0;o<r.length;o++){var s;if(0===(s=r[o]).refs){for(var c=0;c<s.parts.length;c++)s.parts[c]();delete i[s.id]}}}};var w,x=(w=[],function(e,t){return w[e]=t,w.filter(Boolean).join("\n")});function _(e,t,n,r){var o=n?"":r.css;if(e.styleSheet)e.styleSheet.cssText=x(t,o);else{var i=document.createTextNode(o),a=e.childNodes;a[t]&&e.removeChild(a[t]),a.length?e.insertBefore(i,a[t]):e.appendChild(i)}}function S(e,t){var n=t.css,r=t.media;if(r&&e.setAttribute("media",r),e.styleSheet)e.styleSheet.cssText=n;else{for(;e.firstChild;)e.removeChild(e.firstChild);e.appendChild(document.createTextNode(n))}}function E(e,t,n){var r=n.css,o=n.sourceMap,i=void 0===t.convertToAbsoluteUrls&&o;(t.convertToAbsoluteUrls||i)&&(r=d(r)),o&&(r+="\n/*# sourceMappingURL=data:application/json;base64,"+btoa(unescape(encodeURIComponent(JSON.stringify(o))))+" */");var a=new Blob([r],{type:"text/css"}),s=e.href;e.href=URL.createObjectURL(a),s&&URL.revokeObjectURL(s)}},function(e,t){e.exports=function(e){var t="undefined"!=typeof window&&window.location;if(!t)throw new Error("fixUrls requires window.location");if(!e||"string"!=typeof e)return e;var n=t.protocol+"//"+t.host,r=n+t.pathname.replace(/\/[^\/]*$/,"/");return e.replace(/url\s*\(((?:[^)(]|\((?:[^)(]+|\([^)(]*\))*\))*)\)/gi,(function(e,t){var o,i=t.trim().replace(/^"(.*)"$/,(function(e,t){return t})).replace(/^'(.*)'$/,(function(e,t){return t}));return/^(#|data:|http:\/\/|https:\/\/|file:\/\/\/|\s*$)/i.test(i)?e:(o=0===i.indexOf("//")?i:0===i.indexOf("/")?n+i:r+i.replace(/^\.\//,""),"url("+JSON.stringify(o)+")")}))}},function(e,t,n){"use strict";n.r(t),n.d(t,"default",(function(){return p}));n(1);var r=n(0),o=n.n(r);function i(e,t){for(var n=0;n<t.length;n++){var r=t[n];r.enumerable=r.enumerable||!1,r.configurable=!0,"value"in r&&(r.writable=!0),Object.defineProperty(e,r.key,r)}}var a=function(){function e(t){var n=t.config,r=t.onUpload,o=t.onError;!function(e,t){if(!(e instanceof t))throw new TypeError("Cannot call a class as a function")}(this,e),this.config=n,this.onUpload=r,this.onError=o}var t,n,r;return t=e,(n=[{key:"uploadSelectedFile",value:function(e){var t=this,n=e.onPreview;(this.config.uploader&&"function"==typeof this.config.uploader.uploadByFile?o.a.selectFiles({accept:this.config.types}).then((function(e){n();var r,o=t.config.uploader.uploadByFile(e[0]);return(r=o)&&"function"==typeof r.then||console.warn("Custom uploader method uploadByFile should return a Promise"),o})):o.a.transport({url:this.config.endpoint,accept:this.config.types,beforeSend:function(){return n()},fieldName:this.config.field,headers:this.config.additionalRequestHeaders||{}}).then((function(e){return e.body}))).then((function(e){t.onUpload(e)})).catch((function(e){var n=e.body,r=n&&n.message?n.message:t.config.errorMessage;t.onError(r)}))}}])&&i(t.prototype,n),r&&i(t,r),e}();function s(e){return function(e){if(Array.isArray(e))return c(e)}(e)||function(e){if("undefined"!=typeof Symbol&&null!=e[Symbol.iterator]||null!=e["@@iterator"])return Array.from(e)}(e)||function(e,t){if(!e)return;if("string"==typeof e)return c(e,t);var n=Object.prototype.toString.call(e).slice(8,-1);"Object"===n&&e.constructor&&(n=e.constructor.name);if("Map"===n||"Set"===n)return Array.from(e);if("Arguments"===n||/^(?:Ui|I)nt(?:8|16|32)(?:Clamped)?Array$/.test(n))return c(e,t)}(e)||function(){throw new TypeError("Invalid attempt to spread non-iterable instance.\nIn order to be iterable, non-array objects must have a [Symbol.iterator]() method.")}()}function c(e,t){(null==t||t>e.length)&&(t=e.length);for(var n=0,r=new Array(t);n<t;n++)r[n]=e[n];return r}function l(e){var t,n=arguments.length>1&&void 0!==arguments[1]?arguments[1]:null,r=arguments.length>2&&void 0!==arguments[2]?arguments[2]:{},o=document.createElement(e);Array.isArray(n)?(t=o.classList).add.apply(t,s(n)):n&&o.classList.add(n);for(var i in r)o[i]=r[i];return o}function u(e){return 0===Object.keys(e).length}const f='<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" fill="none" viewBox="0 0 24 24"><path stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M13.3236 8.43554L9.49533 12.1908C9.13119 12.5505 8.93118 13.043 8.9393 13.5598C8.94741 14.0767 9.163 14.5757 9.53862 14.947C9.91424 15.3182 10.4191 15.5314 10.9422 15.5397C11.4653 15.5479 11.9637 15.3504 12.3279 14.9908L16.1562 11.2355C16.8845 10.5161 17.2845 9.53123 17.2682 8.4975C17.252 7.46376 16.8208 6.46583 16.0696 5.72324C15.3184 4.98066 14.3086 4.55425 13.2624 4.53782C12.2162 4.52138 11.2193 4.91627 10.4911 5.63562L6.66277 9.39093C5.57035 10.4699 4.97032 11.9473 4.99467 13.4979C5.01903 15.0485 5.66578 16.5454 6.79264 17.6592C7.9195 18.7731 9.43417 19.4127 11.0034 19.4374C12.5727 19.462 14.068 18.8697 15.1604 17.7907L18.9887 14.0354"/></svg>';function d(e,t){for(var n=0;n<t.length;n++){var r=t[n];r.enumerable=r.enumerable||!1,r.configurable=!0,"value"in r&&(r.writable=!0),Object.defineProperty(e,r.key,r)}}var p=function(){function e(t){var n=this,r=t.data,o=t.config,i=t.api,s=t.readOnly;!function(e,t){if(!(e instanceof t))throw new TypeError("Cannot call a class as a function")}(this,e),this.api=i,this.readOnly=s,this.nodes={wrapper:null,button:null,title:null},this._data={file:{},title:""},this.config={endpoint:o.endpoint||"",field:o.field||"file",types:o.types||"*",buttonText:o.buttonText||"Select file to upload",errorMessage:o.errorMessage||"File upload failed",uploader:o.uploader||void 0,additionalRequestHeaders:o.additionalRequestHeaders||{}},void 0===r||u(r)||(this.data=r),this.uploader=new a({config:this.config,onUpload:function(e){return n.onUpload(e)},onError:function(e){return n.uploadingFailed(e)}}),this.enableFileUpload=this.enableFileUpload.bind(this)}var t,n,r;return t=e,r=[{key:"toolbox",get:function(){return{icon:f,title:"Attachment"}}},{key:"isReadOnlySupported",get:function(){return!0}}],(n=[{key:"CSS",get:function(){return{baseClass:this.api.styles.block,apiButton:this.api.styles.button,loader:this.api.styles.loader,wrapper:"cdx-attaches",wrapperWithFile:"cdx-attaches--with-file",wrapperLoading:"cdx-attaches--loading",button:"cdx-attaches__button",title:"cdx-attaches__title",size:"cdx-attaches__size",downloadButton:"cdx-attaches__download-button",fileInfo:"cdx-attaches__file-info",fileIcon:"cdx-attaches__file-icon",fileIconBackground:"cdx-attaches__file-icon-background",fileIconLabel:"cdx-attaches__file-icon-label"}}},{key:"EXTENSIONS",get:function(){return{doc:"#1483E9",docx:"#1483E9",odt:"#1483E9",pdf:"#DB2F2F",rtf:"#744FDC",tex:"#5a5a5b",txt:"#5a5a5b",pptx:"#E35200",ppt:"#E35200",mp3:"#eab456",mp4:"#f676a6",xls:"#11AE3D",html:"#2988f0",htm:"#2988f0",png:"#AA2284",jpg:"#D13359",jpeg:"#D13359",gif:"#f6af76",zip:"#4f566f",rar:"#4f566f",exe:"#e26f6f",svg:"#bf5252",key:"#00B2FF",sketch:"#FFC700",ai:"#FB601D",psd:"#388ae5",dmg:"#e26f6f",json:"#2988f0",csv:"#11AE3D"}}},{key:"validate",value:function(e){return!u(e.file)}},{key:"save",value:function(e){if(this.pluginHasData()){var t=e.querySelector(".".concat(this.CSS.title));t&&Object.assign(this.data,{title:t.innerHTML})}return this.data}},{key:"render",value:function(){var e=l("div",this.CSS.baseClass);return this.nodes.wrapper=l("div",this.CSS.wrapper),this.pluginHasData()?this.showFileData():this.prepareUploadButton(),e.appendChild(this.nodes.wrapper),e}},{key:"prepareUploadButton",value:function(){this.nodes.button=l("div",[this.CSS.apiButton,this.CSS.button]),this.nodes.button.innerHTML="".concat(f," ").concat(this.config.buttonText),this.readOnly||this.nodes.button.addEventListener("click",this.enableFileUpload),this.nodes.wrapper.appendChild(this.nodes.button)}},{key:"appendCallback",value:function(){this.nodes.button.click()}},{key:"pluginHasData",value:function(){return""!==this.data.title||Object.values(this.data.file).some((function(e){return void 0!==e}))}},{key:"enableFileUpload",value:function(){var e=this;this.uploader.uploadSelectedFile({onPreview:function(){e.nodes.wrapper.classList.add(e.CSS.wrapperLoading,e.CSS.loader)}})}},{key:"onUpload",value:function(e){var t,n,r,o=e;try{o.success&&void 0!==o.file&&!u(o.file)?(this.data={file:o.file,title:o.file.title||""},this.nodes.button.remove(),this.showFileData(),t=this.nodes.title,n=document.createRange(),r=window.getSelection(),n.selectNodeContents(t),n.collapse(!1),r.removeAllRanges(),r.addRange(n),this.removeLoader()):this.uploadingFailed(this.config.errorMessage)}catch(e){console.error("Attaches tool error:",e),this.uploadingFailed(this.config.errorMessage)}this.api.blocks.getBlockByIndex(this.api.blocks.getCurrentBlockIndex()).dispatchChange()}},{key:"appendFileIcon",value:function(e){var t,n=e.extension||(void 0===(t=e.name)?"":t.split(".").pop()),r=this.EXTENSIONS[n],o=l("div",this.CSS.fileIcon),i=l("div",this.CSS.fileIconBackground);if(r&&(i.style.backgroundColor=r),o.appendChild(i),n){var a=n;n.length>4&&(a=n.substring(0,4)+"…");var s=l("div",this.CSS.fileIconLabel,{textContent:a,title:n});r&&(s.style.backgroundColor=r),o.appendChild(s)}else i.innerHTML=f;this.nodes.wrapper.appendChild(o)}},{key:"removeLoader",value:function(){var e=this;setTimeout((function(){return e.nodes.wrapper.classList.remove(e.CSS.wrapperLoading,e.CSS.loader)}),500)}},{key:"showFileData",value:function(){this.nodes.wrapper.classList.add(this.CSS.wrapperWithFile);var e=this.data,t=e.file,n=e.title;this.appendFileIcon(t);var r=l("div",this.CSS.fileInfo);if(this.nodes.title=l("div",this.CSS.title,{contentEditable:!1===this.readOnly}),this.nodes.title.dataset.placeholder=this.api.i18n.t("File title"),this.nodes.title.textContent=n||"",r.appendChild(this.nodes.title),t.size){var o,i,a=l("div",this.CSS.size);Math.log10(+t.size)>=6?(o="MiB",i=t.size/Math.pow(2,20)):(o="KiB",i=t.size/Math.pow(2,10)),a.textContent=i.toFixed(1),a.setAttribute("data-size",o),r.appendChild(a)}if(this.nodes.wrapper.appendChild(r),void 0!==t.url){var s=l("a",this.CSS.downloadButton,{innerHTML:'<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" fill="none" viewBox="0 0 24 24"><path stroke="currentColor" stroke-linecap="round" stroke-width="2" d="M7 10L11.8586 14.8586C11.9367 14.9367 12.0633 14.9367 12.1414 14.8586L17 10"/></svg>',href:t.url,target:"_blank",rel:"nofollow noindex noreferrer"});this.nodes.wrapper.appendChild(s)}}},{key:"uploadingFailed",value:function(e){this.api.notifier.show({message:e,style:"error"}),this.removeLoader()}},{key:"data",get:function(){return this._data},set:function(e){var t=e.file,n=e.title;this._data={file:t,title:n}}}])&&d(t.prototype,n),r&&d(t,r),e}()}]).default})); | 34,797 | Python | .tac | 7 | 4,968.857143 | 34,512 | 0.689568 | Nigel2392/wagtail_editorjs | 8 | 0 | 3 | GPL-2.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,302 | entity-builder.py | pdxlocations_meshtastic-ha-entity-builder/entity-builder.py | from meshtastic.serial_interface import SerialInterface
from meshtastic.tcp_interface import TCPInterface
from meshtastic.ble_interface import BLEInterface
import argparse
import sys
### Add arguments to parse
parser = argparse.ArgumentParser(
add_help=False,
epilog="If no connection arguments are specified, we attempt a serial connection and then a TCP connection to localhost.")
helpGroup = parser.add_argument_group("Help")
helpGroup.add_argument("-h", "--help", action="help", help="Show this help message and exit.")
connOuter = parser.add_argument_group('Connection', 'Optional arguments to specify a device to connect to and how.')
conn = connOuter.add_mutually_exclusive_group()
conn.add_argument(
"--port",
help="The port to connect to via serial, e.g. `/dev/ttyUSB0`.",
default=None,
)
conn.add_argument(
"--host",
help="The hostname or IP address to connect to using TCP.",
default=None,
)
conn.add_argument(
"--ble",
help="The BLE device MAC address or name to connect to.",
default=None,
)
mqtt = parser.add_argument_group("MQTT", "Arguments to specify the gateway node and root MQTT topics")
mqtt.add_argument(
"--gateway",
help="The ID of the MQTT gateway node, e.g. !12345678. If not provided, will use the ID of the locally connected node.",
default=None,
)
# it would be nice to have this request settings if the gateway isn't the remote node
mqtt.add_argument(
"--root-topic",
help="The root topic to use in MQTT for the generated files. If not provided, will attempt to get the root path from the local node and use all channels. Wildcard: `+`. Example: to include all channels with the root topic `msh/`, use `msh/2/json/+`. To include just LongFast, use `msh/2/json/LongFast`",
default=None,
)
includes = parser.add_argument_group("Includes", "Arguments to specify what sensors to generate for each node.")
includes.add_argument(
"--no-messages",
help="Don't include a sensor for messages from the node.",
action='store_true',
)
includes.add_argument(
"--fahrenheit",
help="Use Fahrenheit instead of Celsius.",
action='store_true',
)
includes.add_argument(
"--no-temperature",
help="Don't include a temperature sensor.",
action='store_true',
)
includes.add_argument(
"--no-humidity",
help="Don't include a humidity sensor.",
action='store_true',
)
includes.add_argument(
"--no-pressure",
help="Don't include a barometric pressure sensor.",
action='store_true',
)
includes.add_argument(
"--gas-resistance",
help="Include a gas resistance sensor.",
action='store_true',
)
includes.add_argument(
"--power-ch1",
help="Include current & voltage channel 1 sensor.",
action='store_true',
)
includes.add_argument(
"--power-ch2",
help="Include current & voltage channel 2 sensor.",
action='store_true',
)
includes.add_argument(
"--power-ch3",
help="Include current & voltage channel 3 sensor.",
action='store_true',
)
parser.add_argument(
"--nodes",
help="Only generate sensors for these nodes. If not provided, all nodes in the NodeDB will be included. Example: `\"!XXXXXXXX\", \"!YYYYYYYY\"`.",
nargs='*',
action='store',
)
args = parser.parse_args()
### Create an interface
if args.ble:
iface = BLEInterface(args.ble)
elif args.host:
iface = TCPInterface(args.host)
else:
try:
iface = SerialInterface(args.port)
except PermissionError as ex:
print("You probably need to add yourself to the `dialout` group to use a serial connection.")
if iface.devPath is None:
iface = TCPInterface("localhost")
if args.gateway:
gateway_id = args.gateway
else:
gateway_id = f"!{iface.localNode.nodeNum:08x}"
if args.root_topic:
root_topic = args.root_topic
else:
mqttRoot = iface.localNode.moduleConfig.mqtt.root
if mqttRoot != "":
root_topic = mqttRoot + '/2/json/+'
print(f"Using a gateway ID of {gateway_id} and a root topic of {root_topic}")
node_list = []
use_node_list = False # only use nodes from the node list. If False, create for all nodes in db.
if args.nodes and len(args.nodes) > 0:
use_node_list = True
node_list = args.nodes
print(f"Using node list: {node_list}")
fahrenheit = args.fahrenheit
include_messages = not args.no_messages
include_temperature = not args.no_temperature
include_humidity = not args.no_humidity
include_pressure = not args.no_pressure
include_gas_resistance = args.gas_resistance
include_power_ch1 = args.power_ch1
include_power_ch2 = args.power_ch2
include_power_ch3 = args.power_ch3
# initialize the file with the 'sensor' header
with open("mqtt.yaml", "w", encoding="utf-8") as file:
file.write('sensor:\n')
# initialize the file as empty so we have something to append to
with open("automations.yaml", "w", encoding="utf-8") as file:
file.write('')
for node_num, node in iface.nodes.items():
node_short_name = f"{node['user']['shortName']}"
node_long_name = f"{node['user']['longName']}"
node_id = f"{node['user']['id']}"
node_num = f"{node['num']}"
hardware_model = f"{node['user']['hwModel']}"
automation_config = f"""
- id: 'update_location_{node_num}'
alias: update {node_id} location
trigger:
- platform: mqtt
topic: "{root_topic}/{gateway_id}"
payload: 'on'
value_template: >-
{{%- if value_json.from == {node_num} and
value_json.payload.latitude_i is defined and
value_json.payload.longitude_i is defined -%}}
on
{{%- endif -%}}
condition: []
action:
- service: device_tracker.see
metadata: {{}}
data:
dev_id: "{int(node_num):08x}"
gps:
- '{{{{ (trigger.payload | from_json).payload.latitude_i | int * 1e-7 }}}}'
- '{{{{ (trigger.payload | from_json).payload.longitude_i | int * 1e-7 }}}}'
mode: single
"""
config = f'''
- name: "{node_short_name} Last Heard"
unique_id: "{int(node_num):08x}_last_heard"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
device_class: timestamp
value_template: >-
{{% if value_json.from == {node_num} and
value_json.timestamp is defined %}}
{{{{ as_datetime(value_json.timestamp) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
device:
name: "{node_long_name}"
model: "{hardware_model}"
identifiers:
- "meshtastic_{node_num}"
- name: "{node_short_name} SNR"
unique_id: "{int(node_num):08x}_snr"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
device_class: signal_strength
value_template: >-
{{% if value_json.from == {node_num} and value_json.snr is defined %}}
{{{{ value_json.snr}}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
icon: "mdi:signal"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} RSSI"
unique_id: "{int(node_num):08x}_rssi"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
device_class: signal_strength
value_template: >-
{{% if value_json.from == {node_num} and value_json.rssi is defined %}}
{{{{ value_json.rssi | int}}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
icon: "mdi:signal-variant"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Hops Away"
unique_id: "{int(node_num):08x}_hops_away"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
device_class: distance
value_template: >-
{{% if value_json.from == {node_num} and value_json.hops_away is defined %}}
{{{{ value_json.hops_away | int }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
icon: "mdi:rabbit"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Battery Voltage"
unique_id: "{int(node_num):08x}_battery_voltage"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and
value_json.payload.voltage is defined and
value_json.payload.temperature is not defined %}}
{{{{ (value_json.payload.voltage | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "V"
icon: "mdi:lightning-bolt"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Battery Percent"
unique_id: "{int(node_num):08x}_battery_percent"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.battery_level is defined %}}
{{{{ (value_json.payload.battery_level | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
device_class: battery
unit_of_measurement: "%"
icon: "mdi:battery-high"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Uptime"
unique_id: "{int(node_num):08x}_uptime"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
device_class: duration
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.uptime_seconds is defined %}}
{{{{ value_json.payload.uptime_seconds | int }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "s"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} ChUtil"
unique_id: "{int(node_num):08x}_chutil"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.channel_utilization is defined %}}
{{{{ (value_json.payload.channel_utilization | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "%"
icon: "mdi:signal-distance-variant"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} AirUtilTX"
unique_id: "{int(node_num):08x}_airutiltx"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.air_util_tx is defined %}}
{{{{ (value_json.payload.air_util_tx | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "%"
icon: "mdi:percent-box-outline"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_messages:
config += f'''
- name: "{node_short_name} Messages"
unique_id: "{int(node_num):08x}_messages"
state_topic: "{root_topic}/{gateway_id}"
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.text is defined %}}
{{{{ value_json.payload.text }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
device:
identifiers: "meshtastic_{node_num}"
icon: "mdi:chat"
'''
if include_temperature:
if fahrenheit:
temp_mod = "* 1.8) +32"
temp_unit = "F"
else:
temp_mod = ")"
temp_unit = "C"
config += f'''
- name: "{node_short_name} Temperature"
unique_id: "{int(node_num):08x}_temperature"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.temperature is defined %}}
{{{{ (((value_json.payload.temperature | float) {temp_mod}) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "{temp_unit}"
icon: "mdi:sun-thermometer"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_humidity:
config += f'''
- name: "{node_short_name} Humidity"
unique_id: "{int(node_num):08x}_humidity"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.relative_humidity is defined %}}
{{{{ (value_json.payload.relative_humidity | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "%"
icon: "mdi:water-percent"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_pressure:
config += f'''
- name: "{node_short_name} Barometric Pressure"
unique_id: "{int(node_num):08x}_pressure"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.barometric_pressure is defined %}}
{{{{ (value_json.payload.barometric_pressure | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "hPa"
icon: "mdi:chevron-double-down"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_gas_resistance:
config += f'''
- name: "{node_short_name} Gas Resistance"
unique_id: "{int(node_num):08x}_gas_resistance"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.gas_resistance is defined %}}
{{{{ (value_json.payload.gas_resistance | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "MOhms"
icon: "mdi:dots-hexagon"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_power_ch1:
config += f'''
# {node_long_name}
- name: "{node_short_name} Voltage Sensor Ch1"
unique_id: "{int(node_num):08x}_voltage_sensor_ch1"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.voltage_ch1 is defined %}}
{{{{ (value_json.payload.voltage_ch1 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "V"
icon: "mdi:lightning-bolt"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Current Sensor Ch1"
unique_id: "{int(node_num):08x}_current_sensor_ch1"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.current_ch1 is defined %}}
{{{{ (value_json.payload.current_ch1 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "mA"
icon: "mdi:waves"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_power_ch2:
config += f'''
- name: "{node_short_name} Voltage Sensor Ch2"
unique_id: "{int(node_num):08x}_voltage_sensor_ch2"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.voltage_ch2 is defined %}}
{{{{ (value_json.payload.voltage_ch2 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "V"
icon: "mdi:lightning-bolt"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Current Sensor Ch2"
unique_id: "{int(node_num):08x}_current_sensor_ch2"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.current_ch2 is defined %}}
{{{{ (value_json.payload.current_ch2 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "mA"
icon: "mdi:waves"
device:
identifiers: "meshtastic_{node_num}"
'''
if include_power_ch3:
config += f'''
- name: "{node_short_name} Voltage Sensor Ch3"
unique_id: "{int(node_num):08x}_voltage_sensor_ch3"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.voltage_ch3 is defined %}}
{{{{ (value_json.payload.voltage_ch3 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "V"
icon: "mdi:lightning-bolt"
device:
identifiers: "meshtastic_{node_num}"
- name: "{node_short_name} Current Sensor Ch3"
unique_id: "{int(node_num):08x}_current_sensor_ch3"
state_topic: "{root_topic}/{gateway_id}"
state_class: measurement
value_template: >-
{{% if value_json.from == {node_num} and value_json.payload.current_ch3 is defined %}}
{{{{ (value_json.payload.current_ch3 | float) | round(2) }}}}
{{% else %}}
{{{{ this.state }}}}
{{% endif %}}
unit_of_measurement: "mA"
icon: "mdi:waves"
device:
identifiers: "meshtastic_{node_num}"
'''
if node_id in node_list or (not use_node_list):
with open("mqtt.yaml", "a", encoding="utf-8") as file:
file.write(config + '\n')
with open("automations.yaml", "a", encoding="utf-8") as file:
file.write(automation_config)
iface.close()
| 17,607 | Python | .py | 490 | 30.34898 | 307 | 0.618082 | pdxlocations/meshtastic-ha-entity-builder | 8 | 3 | 1 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,303 | deleteAllRedirects.py | nightscout_trio-docs/utils/deleteAllRedirects.py | import requests
import json
import time
import sys
import re
PROJECT = 'androidaps'
URL = 'https://readthedocs.org/api/v3/projects/' + PROJECT + '/redirects/'
TOKEN = len(sys.argv) == 2 and sys.argv[1]
HEADERS = {'Authorization': f'token {TOKEN}'}
def delItem (url) :
delResponse = requests.delete(url, headers=HEADERS)
if delResponse.status_code == 204 :
print('removed ' + url)
elif delResponse.status_code == 429:
detail = delResponse.json()['detail']
wait = int(re.search(r'\d+', detail).group())
print('Throttled, wait for ' + str(wait + 1) + ' seconds')
time.sleep(wait + 1)
delItem(url)
else :
results = delResponse.json()
print(results)
def deleteList() :
response = requests.get(URL, headers=HEADERS)
listResult = response.json()
if response.status_code == 200:
redirects = listResult['results']
for redirect in redirects :
url = redirect['_links']['_self']
delItem(url)
return listResult['count']
elif response.status_code == 429:
detail = response.json()['detail']
wait = int(re.search(r'\d+', detail).group())
print('Throttled, wait for ' + str(wait + 1) + ' seconds')
time.sleep(wait + 1)
deleteList()
else :
print('ERROR code:', response.status_code, listResult)
return 0
def main() :
while True:
count = deleteList()
print('Removed, still counting: ' + str(count))
if count == 0 :
break
print('done')
if not TOKEN :
print('Please provide a API token as parameter')
print('useage: $ python deleteAllRedirects.py <apikey>')
else :
main()
| 1,776 | Python | .py | 52 | 26.903846 | 75 | 0.596963 | nightscout/trio-docs | 8 | 11 | 6 | AGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,304 | importRedirects.py | nightscout_trio-docs/utils/importRedirects.py | import requests
import json
import time
import sys
import re
PROJECT = 'androidaps'
URL = 'https://readthedocs.org/api/v3/projects/' + PROJECT + '/redirects/'
TOKEN = len(sys.argv) == 2 and sys.argv[1]
HEADERS = {'Authorization': f'token {TOKEN}'}
FILE = 'redirect.json'
def create(redirect, index) :
response = requests.post(
URL,
json=redirect,
headers=HEADERS,
)
if response.status_code == 201 :
print ('create redirect (' + str(index) + ') ' + redirect['from_url'])
elif response.status_code == 429:
detail = response.json()['detail']
wait = int(re.search(r'\d+', detail).group())
print('Throttled, wait for ' + str(wait + 1) + ' seconds ')
time.sleep(wait + 1)
create(redirect, index)
else :
print(response.status_code , response.json())
def main():
try:
with open(FILE) as json_file:
redirects = json.load(json_file)
print('Creating ' + str(len(redirects)) + ' redirects ')
for index, redirect in enumerate(redirects):
create(redirect, index)
print ('done')
except IOError:
print('File ' + FILE + ' is not accessible, please make sure you run the "generateRedirect" script')
if not TOKEN :
print('Please provide a API token as parameter')
print('useage: $ python importRedirects.py <apikey>')
else :
main()
| 1,473 | Python | .py | 41 | 28.585366 | 109 | 0.600563 | nightscout/trio-docs | 8 | 11 | 6 | AGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,305 | generateRedirects.py | nightscout_trio-docs/utils/generateRedirects.py | import os
import json
FILE = 'redirect.json'
dir_path = os.path.dirname(os.path.realpath(__file__))
language_path = os.path.join(dir_path, '..\\', 'docs', 'CROWDIN') + '\\'
print('Generate redirect for al .md and .rst files in path ' + language_path)
redirects = []
class Object:
def toJSON(self):
return json.dumps(self, default=lambda o: o.__dict__,
sort_keys=True, indent=4)
for path, subdirs, files in os.walk(language_path):
for filename in files:
if filename.endswith(('.md', '.rst')):
file = os.path.splitext(os.path.join(path, filename))[0] + '.html'
relative_file = file[len(language_path):]
original = '\\' + os.path.join('en', 'latest', 'CROWDIN', relative_file)
to = '\\' + os.path.join(relative_file[:2], 'latest', relative_file[3:])
r = Object()
r.from_url = original.replace('\\', '/')
r.to_url = to.replace('\\', '/')
r.type = 'exact'
redirects.append(r)
def obj_dict(obj):
return obj.__dict__
json_object = json.dumps(redirects, indent=4, default=obj_dict)
with open(FILE, 'w') as outfile:
outfile.write(json_object)
print('Done, ' + str(len(redirects)) + ' results are stored in ' + FILE)
| 1,309 | Python | .py | 29 | 37.172414 | 85 | 0.586371 | nightscout/trio-docs | 8 | 11 | 6 | AGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,306 | shared.conf.py | nightscout_trio-docs/docs/shared.conf.py | # -*- coding: utf-8 -*-
#
# Trio documentation build configuration file
#
# This file is execfile()d with the current directory set to its
# containing dir.
#
# Note that not all possible configuration values are present in this
# autogenerated file.
#
# All configuration values have a default; values that are commented out
# serve to show the default.
import sys
import os
# If extensions (or modules to document with autodoc) are in another directory,
# add these directories to sys.path here. If the directory is relative to the
# documentation root, use os.path.abspath to make it absolute, like shown here.
#sys.path.insert(0, os.path.abspath('.'))
#from recommonmark.parser import CommonMarkParser
#from recommonmark.transform import AutoStructify
import alabaster
# -- General configuration ------------------------------------------------
# RTD
on_rtd = os.environ.get('READTHEDOCS', None) == 'True'
# If your documentation needs a minimal Sphinx version, state it here.
#needs_sphinx = '1.0'
# Add any Sphinx extension module names here, as strings. They can be
# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
# ones.
extensions = [
'sphinx.ext.ifconfig',
'sphinx.ext.autodoc',
'sphinx.ext.todo',
'myst_parser',
'sphinx_copybutton',
'sphinx_new_tab_link',
# 'alabaster',
]
myst_enable_extensions = [
"colon_fence",
]
myst_heading_anchors = 3
# Add any paths that contain templates here, relative to this directory.
templates_path = ['../../_templates'] # path relative to languages conf.py
#source_parsers = {
# '.md': CommonMarkParser,
#}
# The suffix(es) of source filenames.
# You can specify multiple suffix as a list of string:
#source_suffix = ['.md', '.rst', ]
# source_suffix = '.rst'
# The encoding of source files.
#source_encoding = 'utf-8-sig'
# The master toctree document.
master_doc = 'index'
# General information about the project.
project = u'Trio'
copyright = u'Trio Community'
author = u'Trio Community'
# The version info for the project you're documenting, acts as replacement for
# |version| and |release|, also used in various other places throughout the
# built documents.
#
# The short X.Y version.
version = u'0.0.1'
# The full version, including alpha/beta/rc tags.
release = u'0.0.1'
# The language for content autogenerated by Sphinx. Refer to documentation
# for a list of supported languages.
#
# This is also used if you do content translation via gettext catalogs.
# Usually you set "language" from the command line for these cases.
#language = None
# There are two options for replacing |today|: either, you set today to some
# non-false value, then it is used:
#today = ''
# Else, today_fmt is used as the format for a strftime call.
#today_fmt = '%B %d, %Y'
# List of patterns, relative to source directory, that match files and
# directories to ignore when looking for source files.
exclude_patterns = []
# The reST default role (used for this markup: `text`) to use for all
# documents.
#default_role = None
# If true, '()' will be appended to :func: etc. cross-reference text.
#add_function_parentheses = True
# If true, the current module name will be prepended to all description
# unit titles (such as .. function::).
#add_module_names = True
# If true, sectionauthor and moduleauthor directives will be shown in the
# output. They are ignored by default.
#show_authors = False
# The name of the Pygments (syntax highlighting) style to use.
pygments_style = 'sphinx'
# A list of ignored prefixes for module index sorting.
#modindex_common_prefix = []
# If true, keep warnings as "system message" paragraphs in the built documents.
#keep_warnings = False
# If true, `todo` and `todoList` produce output, else they produce nothing.
todo_include_todos = True
# -- Options for HTML output ----------------------------------------------
# The theme to use for HTML and HTML Help pages. See the documentation for
# a list of builtin themes.
html_theme = 'sphinx_rtd_theme'
html_theme_options = {
"collapse_navigation" : False,
"navigation_with_keys": True
}
# Theme options are theme-specific and customize the look and feel of a theme
# further. For a list of options available for each theme, see the
# documentation.
# alabaster
# theme_github_user = ''
# theme_github_repo = ''
#"""
#html_theme = 'default'
#html_theme_options = {
# 'display_github': True,
# 'github_user': 'nightscout',
# 'github_repo': 'trio-docs',
# 'navigation_depth': 6,
#}
import sphinx_rtd_theme
html_theme_path = [sphinx_rtd_theme.get_html_theme_path( )]
#"""
# Add any paths that contain custom themes here, relative to this directory.
html_theme_path = []
# html_theme_path = [alabaster.get_path( )]
# The name for this set of Sphinx documents. If None, it defaults to
# "<project> v<release> documentation".
#html_title = None
# A shorter title for the navigation bar. Default is the same as html_title.
#html_short_title = None
# The name of an image file (relative to this directory) to place at the top
# of the sidebar.
#html_logo = None
html_logo = '../trio-logo.png' # path relative to languages conf.py
# The name of an image file (relative to this directory) to use as a favicon of
# the docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32
# pixels large.
#html_favicon = None
html_favicon = '../favicon.ico' # path relative to languages conf.py
# Add any paths that contain custom static files (such as style sheets) here,
# relative to this directory. They are copied after the builtin static files,
# so a file named "default.css" will overwrite the builtin "default.css".
#html_static_path = ['_static']
html_static_path = ['../../_static'] # path relative to languages conf.py
# Add any extra paths that contain custom files (such as robots.txt or
# .htaccess) here, relative to this directory. These files are copied
# directly to the root of the documentation.
#html_extra_path = []
# If not '', a 'Last updated on:' timestamp is inserted at every page bottom,
# using the given strftime format.
#html_last_updated_fmt = '%b %d, %Y'
html_last_updated_fmt = '%b %d, %Y'
# If true, SmartyPants will be used to convert quotes and dashes to
# typographically correct entities.
#html_use_smartypants = True
# Custom sidebar templates, maps document names to template names.
#html_sidebars = {}
# for alabaster
# """
#html_sidebars = { }
#"""
#html_sidebars = {
# '**': [
# 'about.html',
# 'navigation.html',
# 'relations.html',
# 'searchbox.html',
# 'donate.html',
# ]
#}
# Additional templates that should be rendered to pages, maps page names to
# template names.
#html_additional_pages = {}
# If false, no module index is generated.
#html_domain_indices = True
# If false, no index is generated.
#html_use_index = True
# If true, the index is split into individual pages for each letter.
#html_split_index = False
# If true, links to the reST sources are added to the pages.
#html_show_sourcelink = True
# If true, "Created using Sphinx" is shown in the HTML footer. Default is True.
#html_show_sphinx = True
# If true, "(C) Copyright ..." is shown in the HTML footer. Default is True.
#html_show_copyright = True
# If true, an OpenSearch description file will be output, and all pages will
# contain a <link> tag referring to it. The value of this option must be the
# base URL from which the finished HTML is served.
#html_use_opensearch = ''
# This is the file name suffix for HTML files (e.g. ".xhtml").
#html_file_suffix = None
# Language to be used for generating the HTML full-text search index.
# Sphinx supports the following languages:
# 'da', 'de', 'en', 'es', 'fi', 'fr', 'hu', 'it', 'ja'
# 'nl', 'no', 'pt', 'ro', 'ru', 'sv', 'tr'
#html_search_language = 'en'
# A dictionary with options for the search language support, empty by default.
# Now only 'ja' uses this config value
#html_search_options = {'type': 'default'}
# The name of a javascript file (relative to the configuration directory) that
# implements a search results scorer. If empty, the default will be used.
#html_search_scorer = 'scorer.js'
# Output file base name for HTML help builder.
htmlhelp_basename = 'trio-docs'
# -- Options for manual page output ---------------------------------------
# One entry per manual page. List of tuples
# (source start file, name, description, authors, manual section).
man_pages = [
(master_doc, 'Trio', u'Trio Documentation',
[author], 1)
]
# If true, show URL addresses after external links.
#man_show_urls = False
# -- Options for Texinfo output -------------------------------------------
# Grouping the document tree into Texinfo files. List of tuples
# (source start file, target name, title, author,
# dir menu entry, description, category)
#texinfo_documents = [
# (master_doc, 'AndroidAPS', u'AndroidAPS Documentation',
# author, 'AndroidAPS', 'One line description of project.',
# 'Miscellaneous'),
#]
# Documents to append as an appendix to all manuals.
#texinfo_appendices = []
# If false, no module index is generated.
#texinfo_domain_indices = True
# How to display URL addresses: 'footnote', 'no', or 'inline'.
#texinfo_show_urls = 'footnote'
# If true, do not generate a @detailmenu in the "Top" node's menu.
#texinfo_no_detailmenu = False
# -- Options for Epub output ----------------------------------------------
# Bibliographic Dublin Core info.
epub_title = project
epub_author = author
epub_publisher = author
epub_copyright = copyright
# The basename for the epub file. It defaults to the project name.
#epub_basename = project
# The HTML theme for the epub output. Since the default themes are not
# optimized for small screen space, using the same theme for HTML and epub
# output is usually not wise. This defaults to 'epub', a theme designed to save
# visual space.
#epub_theme = 'epub'
# The language of the text. It defaults to the language option
# or 'en' if the language is not set.
#epub_language = ''
# The scheme of the identifier. Typical schemes are ISBN or URL.
#epub_scheme = ''
# The unique identifier of the text. This can be a ISBN number
# or the project homepage.
#epub_identifier = ''
# A unique identification for the text.
#epub_uid = ''
# A tuple containing the cover image and cover page html template filenames.
#epub_cover = ()
# A sequence of (type, uri, title) tuples for the guide element of content.opf.
#epub_guide = ()
# HTML files that should be inserted before the pages created by sphinx.
# The format is a list of tuples containing the path and title.
#epub_pre_files = []
# HTML files that should be inserted after the pages created by sphinx.
# The format is a list of tuples containing the path and title.
#epub_post_files = []
# A list of files that should not be packed into the epub file.
epub_exclude_files = ['search.html']
# The depth of the table of contents in toc.ncx.
#epub_tocdepth = 3
# Allow duplicate toc entries.
#epub_tocdup = True
# Choose between 'default' and 'includehidden'.
#epub_tocscope = 'default'
# Fix unsupported image types using the Pillow.
#epub_fix_images = False
# Scale large images.
#epub_max_image_width = 0
# How to display URL addresses: 'footnote', 'no', or 'inline'.
#epub_show_urls = 'inline'
# If false, no index is generated.
#epub_use_index = True
#github_doc_root = 'https://github.com/openaps/AndroidAPSdocs/tree/master/'
#hosted_root = 'http://localhost:8000/'
#on_rtd = os.environ.get('READTHEDOCS', None) == 'True'
#if on_rtd:
# rtd_version = os.environ.get('READTHEDOCS_VERSION')
# rtd_lang = os.environ.get('READTHEDOCS_LANGUAGE')
# rtd_sub = os.environ.get('SUB_DOMAIN') or 'AndroidAPS'
#
# hosted_root = 'https://%s.readthedocs.org/%s/%s/' % (rtd_sub, rtd_lang, rtd_version)
#
#def setup(app):
# app.add_config_value('recommonmark_config', {
# # 'url_resolver': lambda url: github_doc_root + url,
# 'url_resolver': lambda url: hosted_root + url + '.html',
# 'auto_toc_tree_section': 'Summary',
# 'enable_auto_doc_ref': True,
# 'enable_eval_rst': True,
# }, True)
# #app.add_transform(AutoStructify)
# Allows preventing copy button from being added to code blocks like so:
# ```{code-block}
# :class: no-copybutton
# the code you want to display goes here
# ```
copybutton_selector = "div:not(.no-copybutton) > div.highlight > pre"
| 12,422 | Python | .py | 307 | 39.009772 | 87 | 0.71804 | nightscout/trio-docs | 8 | 11 | 6 | AGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,307 | conf.py | nightscout_trio-docs/docs/EN/conf.py | import os
exec (open("../shared.conf.py").read())
# The name of an image file (relative to this directory) to place at the top
# of the sidebar.
#html_logo = None
# html_logo = '../drawing.png'
# The name of an image file (relative to this directory) to use as a favicon of
# the docs. This file should be a Windows icon file (.ico) being 16x16 or 32x32
# pixels large.
#html_favicon = None
# html_favicon = '../favicon.ico'
# Add any paths that contain custom static files (such as style sheets) here,
# relative to this directory. They are copied after the builtin static files,
# so a file named "default.css" will overwrite the builtin "default.css".
html_static_path = ['../_static']
# A list of paths that contain extra templates (or templates that overwrite builtin/theme-specific templates). Relative paths
# are taken as relative to the configuration directory.
templates_path = ['../_templates']
| 912 | Python | .py | 18 | 49.388889 | 125 | 0.748031 | nightscout/trio-docs | 8 | 11 | 6 | AGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,308 | nfc_helper.py | CardToolz_libnfc_mitm_cffi/nfc_helper.py | #!/usr/bin/python3
import ctypes
import binascii
from hexdump import *
from typing import List, Callable, Generic, TypeVar
from enum import Enum
from dataclasses import dataclass
import dataclasses
import json
import logging
logger = logging.getLogger(__name__)
hex2str = lambda x: bytearray.fromhex(x.replace(" ", ""))
str2hex = lambda x: x.hex()
int32tole = lambda x: x.to_bytes(4, byteorder='little')
c_uint8 = ctypes.c_uint8
class ISO14443_PCB_bits(ctypes.LittleEndianStructure):
_fields_ = [
("b1", c_uint8, 1),
("b2", c_uint8, 1),
("b3", c_uint8, 1),
("b4", c_uint8, 1),
("b5", c_uint8, 1),
("b6", c_uint8, 1),
("b7_b8", c_uint8, 2),
]
class ISO14443_PCB_IBlock(ctypes.LittleEndianStructure):
_fields_ = [
("block_num", c_uint8, 1),
("b2_const_1", c_uint8, 1),
("hasNAD", c_uint8, 1),
("hasCID", c_uint8, 1),
("chaining", c_uint8, 1),
("b6_const_0", c_uint8, 1),
("blockType", c_uint8, 2),
]
class ISO14443_PCB_RBlock(ctypes.LittleEndianStructure):
_fields_ = [
("block_num", c_uint8, 1),
("b2_const_1", c_uint8, 1),
("b3_const_0", c_uint8, 1),
("hasCID", c_uint8, 1),
("ACK_NAK", c_uint8, 1),
("b6_const_1", c_uint8, 1),
("b7_b8_blockType", c_uint8, 2),
]
class ISO14443_PCB_SBlock(ctypes.LittleEndianStructure):
_fields_ = [
("b1_const_0", c_uint8, 1),
("b2_const_1", c_uint8, 1),
("b3_const_0", c_uint8, 1),
("b4_hasCID", c_uint8, 1),
("DESELECT_WTX", c_uint8, 2),
("b7_b8_blockType", c_uint8, 2),
]
class ISO14443_PCB(ctypes.Union):
_fields_ = [("bits", ISO14443_PCB_bits),
("iblock", ISO14443_PCB_IBlock),
("rblock", ISO14443_PCB_RBlock),
("sblock", ISO14443_PCB_SBlock),
("asbyte", c_uint8)]
def cstruct_pprint(s):
return "{}: {\t{\t\t{}}}".format(s.__class__.__name__,
", ".join(["{}: {}".format(field[0],
getattr(s,
field[0]))
for field in s._fields_]))
def print_target(target):
ret = '\n'
ret += ("\tUID \t: {}\n".format(binascii.hexlify(bytearray(target.nti.nai.abtUid)[:target.nti.nai.szUidLen])))
ret += ("\tATQA\t: {}\n".format(binascii.hexlify(bytearray(target.nti.nai.abtAtqa))))
ret += ("\tSAK \t: {}\n".format(binascii.hexlify(bytearray(target.nti.nai.btSak)[:1])))
ret += ("\tATS \t: {}\n".format(binascii.hexlify(bytearray(target.nti.nai.abtAts)[:target.nti.nai.szAtsLen])))
return ret
def print_frame(frame):
# frame_data = frame['data'].encode('utf-8')
# frame_len = len(frame_data)
# result = frame['result']
if frame.data == None:
print('frame_data is None')
frame_data = bytearray()
frame_data = frame.data
frame_len = len(frame_data)
result = frame.result
time = frame.time
direction = frame.direction
index = frame.index
print('{} - {} ({}):\tlen: {} \tret: {}'.format(index, direction, time, frame_len, result))
if frame_len != 0:
if not frame.easy_framing:
PCB = ISO14443_PCB(asbyte=frame_data[0])
if PCB.bits.b7_b8 == 0b00:
print('\tI-Block')
print('\t\tPCB: block_num: {}\n\t\tb2_const_1: {}\n\t\thasNAD: {}\n\t\thasCID: {}\n\t\tchaining: {}\n\t\tb6_const_0: {}\n\t\tblockType: {}'.format(
PCB.iblock.block_num, PCB.iblock.b2_const_1, PCB.iblock.hasNAD, PCB.iblock.hasCID, PCB.iblock.chaining, PCB.iblock.b6_const_0, PCB.iblock.blockType
))
elif PCB.bits.b7_b8 == 0b01:
print('\tRFU')
print('\t\tPCB: b1: {}\n\t\tb2: {}\n\t\tb3: {}\n\t\tb4: {}\n\t\tb5: {}\n\t\tb6: {}\n\t\tb7_b8: {}'.format(
PCB.bits.b1, PCB.bits.b2, PCB.bits.b3, PCB.bits.b4, PCB.bits.b5, PCB.bits.b6, PCB.bits.b7_b8
))
elif PCB.bits.b7_b8 == 0b10:
print('\tR-Block')
print('\t\tPCB: block_num: {}\n\t\tb2_const_1: {}\n\t\tb3_const_0: {}\n\t\thasCID: {}\n\t\tACK_NAK: {}\n\t\tb6_const_1: {}\n\t\tb7_b8_blockType: {}'.format(
PCB.rblock.block_num, PCB.rblock.b2_const_1, PCB.rblock.b3_const_0, PCB.rblock.hasCID, PCB.rblock.ACK_NAK, PCB.rblock.b6_const_1, PCB.rblock.b7_b8_blockType
))
elif PCB.bits.b7_b8 == 0b11:
print('\tS-Block')
print('\t\tPCB: b1_const_0: {}\n\t\tb2_const_1: {}\n\t\tb3_const_0: {}\n\t\tb4_hasCID: {}\n\t\tDESELECT_WTX: {}\n\t\tb7_b8_blockType: {}'.format(
PCB.sblock.b1_const_0, PCB.sblock.b2_const_1, PCB.sblock.b3_const_0, PCB.sblock.b4_hasCID, PCB.sblock.DESELECT_WTX, PCB.sblock.b7_b8_blockType
))
print(hexdump(frame_data, result='return'))
# print ("\tPCB \t:", binascii.hexlify(bytearray(frame[:1])))
# print ("\tCID \t:", binascii.hexlify(bytearray(frame[1:2])))
# print ("\tNAD \t:", binascii.hexlify(bytearray(frame[2:3])))
# print ("\tINF \t:", binascii.hexlify(bytearray(frame[3:-2])))
# print ("\tCRC \t:", binascii.hexlify(bytearray(frame[-2:])))
# print ("\tCRC OK \t:", bool(frame[-1] == 0x00 and frame[-2] == 0x00))
def obj_dict(obj):
return obj.__dict__
class BytearrayEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj, bytearray) or isinstance(obj, bytes):
# bytearray decode to hex string
return obj.hex()
return json.JSONEncoder.default(self, obj)
# class BytearrayDecoder(json.JSONDecoder):
class FrameDirection(str, Enum):
FromReader = "FromReader"
ToReader = "ToReader"
FromCard = "FromCard"
ToCard = "ToCard"
@dataclass
class Frame:
index: int
time : int
data: bytearray
result: int
direction: FrameDirection
easy_framing: bool = True
def print_data(self):
print_frame(self)
# print_frame(self, 0)
pass
def __iter__(self):
yield from{
'index': self.index,
'time': self.time,
'data': self.data,
'result': self.result,
'direction': self.direction,
'easy_framing': self.easy_framing
}.items()
def __dict__(self) -> dict:
return dataclasses.asdict(self)
def to_json(self):
d = dict(self)
print(d)
return json.dumps(d, cls=BytearrayEncoder)
def __str__(self) -> str:
return self.to_json()
def __repr__(self) -> str:
return self.to_json()
class FrameEncoder(json.JSONEncoder):
def default(self, obj):
if isinstance(obj, Frame):
return obj.to_json()
return json.JSONEncoder.default(self, obj)
# def frame_from_json(self, json_str):
# j = json.loads(json_str)
# self.index = j['index']
# self.time = j['time']
# self.data = j['data']
# self.result = j['result']
# self.direction = j['direction']
# return self
def frame_from_json(json_str):
j = json.loads(json_str)
index = j['index']
time = j['time']
data = bytearray.fromhex(j['data'])
result = j['result']
direction = j['direction']
easy_framing = j['easy_framing']
return Frame(index, time, data, result, direction, easy_framing)
class FrameList:
def __init__(self, easy_framing=True):
self.frame_list: List[Frame] = []
# self.easy_framing = False
self.easy_framing = easy_framing
pass
def clear(self):
self.frame_list.clear()
def add_frame(self, frame):
self.frame_list.append(frame)
def add_frame_by_data(self, index, time, data, result, direction, easy_framing=None):
if easy_framing == None:
easy_framing = self.easy_framing
frame = Frame(index, time, data, result, direction, easy_framing)
self.add_frame(frame)
def get_frame(self, index):
return self.frame_list[index]
def get_frame_list(self):
return self.frame_list
def get_frame_list_len(self):
return len(self.frame_list)
@dataclass
class TargetData:
abtUid: bytearray
abtAtqa: bytearray
btSak: bytearray
def __iter__(self):
yield from{
'abtUid': self.abtUid,
'abtAtqa': self.abtAtqa,
'btSak': self.btSak
}.items()
def __dict__(self) -> dict:
return dataclasses.asdict(self)
def to_json(self):
d = dict(self)
print(d)
return json.dumps(d, cls=BytearrayEncoder)
def __str__(self) -> str:
return self.to_json()
def __repr__(self) -> str:
return self.to_json()
class FrameLogger(FrameList):
log_fname: str = None
def __init__(self, easy_framing=True, log_fname=None):
FrameList.__init__(self, easy_framing)
self.easy_framing = easy_framing
self.log_fname = log_fname
def print(self):
for frame in self.frame_list:
# print(type(frame))
frame.print_data()
def to_json(self):
return json.dumps([frame.__dict__() for frame in self.frame_list], cls=BytearrayEncoder)
def to_json_pretty(self):
return json.dumps([frame.__dict__() for frame in self.frame_list], cls=BytearrayEncoder, indent=4)
def save_to(self, log_fname):
with open(log_fname, 'w') as f:
j = self.to_json_pretty()
f.write(j)
def save(self):
if self.log_fname == None:
return
self.save_to(self.log_fname)
def load_from(self, log_fname):
self.clear()
with open(log_fname, 'r') as f:
j = f.read()
a = json.loads(j)
for frame in a:
self.add_frame(frame_from_json(json.dumps(frame)))
def load(self):
if self.log_fname == None:
return
self.load_from(self.log_fname)
class EmulatedInitiator(FrameLogger):
def configure(self, option, value): # for backward compatibility from relay as data source
pass
def initiator_transceive_bytes(self, data, timeout=0): # for backward compatibility from relay as data source
# print("initiator_transceive_bytes: ", data)
for req in self.frame_list:
if (req.direction == FrameDirection.FromReader) and (req.data[:5] == data[:5]):
idx = req.index
# print("Found req frame: ", idx, req)
for resp in self.frame_list:
if resp.direction == FrameDirection.FromCard and resp.index == idx+1:
# print("Found resp frame: ", resp)
return resp.data, resp.result
print("Can't find frame for request: ", data)
return b'', 0
def chunks(lst, n):
"""Yield successive n-sized chunks from lst."""
for i in range(0, len(lst), n):
yield lst[i:i + n]
| 11,427 | Python | .py | 278 | 31.44964 | 180 | 0.558167 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,309 | log_parser.py | CardToolz_libnfc_mitm_cffi/log_parser.py | #!/usr/bin/python3
# from nfc_ctypes import *
# from nfc_wrapper import *
from nfc_helper import *
# from NFCReplay import *
from argparse import ArgumentParser
def main():
parser = ArgumentParser()
parser.add_argument("-f", "--filename", dest="log_fname", default=0, type=str, help="Input JSON log filename")
fl = FrameLogger(easy_framing=True, log_fname=parser.parse_args().log_fname)
print ("Log file name: %s" % fl.log_fname)
fl.load()
print ("Loaded %d frames" % fl.get_frame_list_len())
fl.print()
if __name__ == "__main__":
main() | 572 | Python | .py | 16 | 32.625 | 114 | 0.67509 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,310 | nfc_wrapper.py | CardToolz_libnfc_mitm_cffi/nfc_wrapper.py | #!/usr/bin/python3
# to trace shared lib calls use "ltrace --library="*libnfc*" python3 ./nfc_wrapper.py"
from libnfc_ffi.libnfc_ffi import ffi, libnfc as nfc
import nfc_helper
from hexdump import *
import time
from pprint import pprint
from inspect import getmembers
import logging
logger = logging.getLogger(__name__)
# def eprint(*args, **kwargs):
# print('\033[91m', *args, '\033[0m', file=sys.stderr, **kwargs)
# def log(*args, **kwargs):
# print(*args, **kwargs)
def cdata_dict(cd):
if isinstance(cd, ffi.CData):
try:
return ffi.string(cd)
except TypeError:
try:
return [cdata_dict(x) for x in cd]
except TypeError:
return {k: cdata_dict(v) for k, v in getmembers(cd)}
else:
return cd
def pprint_cdata(cd, print_hex=False):
logger.debug('cdata info: {}, sizeof()={}'.format(cd, ffi.sizeof(cd)))
logger.debug(cdata_dict(cd))
if print_hex:
c = ffi.buffer(cd)
logger.debug(hexdump(c, result='return'))
def hexbytes(data):
return " ".join(["{:02x}".format(x) for x in data])
sErrorMessages = {
# /* Chip-level errors (internal errors, RF errors, etc.) */
nfc.NFC_SUCCESS: "Success",
nfc.NFC_EIO: "Input / Output Error",
nfc.NFC_EINVARG: "Invalid argument(s)",
nfc.NFC_EDEVNOTSUPP: "Not Supported by Device",
nfc.NFC_ENOTSUCHDEV: "No Such Device",
nfc.NFC_EOVFLOW: "Buffer Overflow",
nfc.NFC_ETIMEOUT: "Timeout",
nfc.NFC_EOPABORTED: "Operation Aborted",
nfc.NFC_ENOTIMPL: "Not (yet) Implemented",
nfc.NFC_ETGRELEASED: "Target Released",
nfc.NFC_EMFCAUTHFAIL: "Mifare Authentication Failed",
nfc.NFC_ERFTRANS: "RF Transmission Error",
nfc.NFC_ECHIP: "Device's Internal Chip Error",
}
cffi_chars_to_str = lambda c: ffi.string(c).decode("utf-8")
NFC_DEVICE_LIST_SIZE = 10
NFC_DEVICE_LIST = ffi.new("nfc_connstring[{0}]".format(NFC_DEVICE_LIST_SIZE))
MAX_FRAME_LEN = 264
ctx = ffi.new("nfc_context**")
nfc.nfc_init(ctx)
c = ctx[0]
def get_version_str():
return cffi_chars_to_str(nfc.nfc_version())
def nfc_exit():
nfc.nfc_exit(c)
def list_devices(verbose=False):
result = []
num_devices = nfc.nfc_list_devices(c, NFC_DEVICE_LIST, NFC_DEVICE_LIST_SIZE)
for i in range(num_devices):
result.append(NFC_DEVICE_LIST[i])
if verbose:
print ('libNFC devices ({}):'.format(num_devices))
for i in range(num_devices):
dev = nfc.nfc_open(c, result[i])
devname = cffi_chars_to_str(nfc.nfc_device_get_name(dev))
print('\tNo: {}\t\t{}'.format(i, devname))
# nfc.nfc_close(dev)
return result
class NfcDevice(object):
def __init__(self, devdesc=None, verbosity=0, modtype=nfc.NMT_ISO14443A, baudrate=nfc.NBR_106, timeout=5000):
logger.debug("NfcDevice init")
self._device = nfc.nfc_open(c, devdesc)
self._device_name = cffi_chars_to_str(nfc.nfc_device_get_name(self._device))
self._txbytes = ffi.new("uint8_t[{}]".format(MAX_FRAME_LEN))
self._rxbytes = ffi.new("uint8_t[{}]".format(MAX_FRAME_LEN))
self.verbosity = verbosity
self.nm = ffi.new("nfc_modulation*")
self.nm.nmt = modtype
self.nm.nbr = baudrate
self.timeout = timeout
self.last_err = nfc.NFC_SUCCESS
# time.sleep(0.5) # 50ms removes error "libnfc.driver.pn532_spi Unable to wait for SPI data. (RX)"
def get_last_err(self):
logger.debug("get_lasst_err: {}, {}".format(self.last_err, sErrorMessages[self.last_err]))
return self.last_err
def set_modulation(self, modtype, baudrate):
logger.debug("set_modulation")
self.nm.nmt = modtype
self.nm.nbr = baudrate
def set_property_bool(self, option, value: bool):
logger.debug("set_property_bool")
"""Configures the NFC device options"""
ret = nfc.nfc_device_set_property_bool(self._device, option, value)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("set_property_bool() setting option {0} to {1}".format(option, value))
return ret
def set_property_int(self, option, value: int):
logger.debug("set_property_int")
"""Configures the NFC device options"""
ret = nfc.nfc_device_set_property_int(self._device, option, value)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("set_property_int() setting option {0} to {1}".format(option, value))
return ret
class NfcTarget(NfcDevice):
def __init__(self, devdesc, targettype=None, timeout=5000, verbosity=0):
super().__init__(devdesc, verbosity)
ret = self.init(targettype, timeout)
logger.debug("Target dev name: {}".format(self._device_name))
self.last_err = ret
def init(self, targettype=None, timeout=0):
logger.debug("NfcTarget init")
if targettype==None:
targettype = self.prepare_emulated_target()
ret = nfc.nfc_target_init(self._device, targettype, self._rxbytes, MAX_FRAME_LEN, timeout)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("init() error: {}, {}".format(ret, sErrorMessages[ret]))
self._nt = targettype
return ret
def get_target(self):
return self._nt
def prepare_emulated_target(self):
logger.debug("prepare_emulated_target")
abtAtqa = [0x03, 0x04]
abtUid = [0x08, 0xba, 0xdf, 0x0d] # abtUid[0] = 0x08 Needed for PN532 emulation
abtAts = [0x75, 0x33, 0x92, 0x03]
# https://de.wikipedia.org/wiki/Answer_to_Select
# ATS = (05) 75 33 92 03
# (TL) T0 TA TB TC
# | | | +-- CID supported, NAD supported
# | | +----- FWI=9 SFGI=2 => FWT=154ms, SFGT=1.21ms
# | +-------- DR=2,4 DS=2,4 => supports 106, 212 & 424bps in both directions
# +----------- TA,TB,TC, FSCI=5 => FSC=64
# It seems hazardous to tell we support NAD if the tag doesn't support NAD but I don't know how to disable it
# PC/SC pseudo-ATR = 3B 80 80 01 01 if there is no historical bytes
nti = ffi.new("nfc_target_info*")
nti.nai.abtAtqa = abtAtqa
nti.nai.abtUid = abtUid
nti.nai.szUidLen = len(abtUid)
nti.nai.btSak = 0x20
nti.nai.abtAts = abtAts
nti.nai.szAtsLen = len(abtAts)
nt = ffi.new("nfc_target*")
nt.nm = self.nm[0]
nt.nti = nti[0]
logger.info("Emulated target:")
logger.info(nfc_helper.print_target(nt))
return nt
def receive_bytes(self, timeout=None):
logger.debug("receive_bytes")
if timeout is None:
timeout = self.timeout
ret = nfc.nfc_target_receive_bytes(self._device, self._rxbytes, MAX_FRAME_LEN, timeout)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("receive_bytes() error{}: ".format(ret))
data = bytearray(ffi.buffer(self._rxbytes, ret))
logger.info('T<I[%2X]: %s' % (len(data), hexbytes(data)))
return data, ret
def send_bytes(self, txbytes, timeout=None):
logger.debug("send_bytes")
if timeout is None:
timeout = self.timeout
logger.info('T>I[%2X]: %s' % (len(txbytes), hexbytes(txbytes)))
tx_len = len(txbytes)
self._txbytes[0:tx_len] = txbytes
ret = nfc.nfc_target_send_bytes(self._device, self._txbytes, tx_len, timeout)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("send_bytes() error: {}".format(ret))
return ret
def receive_bits(self, *args, **kwargs):
raise NotImplementedError("receive_bits() not implemented")
def send_bits(self, *args, **kwargs):
raise NotImplementedError("send_bits() not implemented")
class NfcInitiator(NfcDevice):
def __init__(self, devdesc=None, verbosity=0):
super().__init__(devdesc, verbosity)
ret = self.init()
logger.info("Initiator dev name:{}".format(self._device_name))
self.last_err = ret
def init(self):
logger.debug("NfcInitiator init()")
ret = nfc.nfc_initiator_init(self._device)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("init() error{}: ".format(ret))
# self.set_property_bool(nfc.NP_HANDLE_CRC, False)
# self.set_property_bool(nfc.NP_ACCEPT_INVALID_FRAMES, True)
# self.set_property_bool(nfc.NP_AUTO_ISO14443_4, False)
# self.set_property_bool(nfc.NP_EASY_FRAMING, False)
self.set_property_int(nfc.NP_TIMEOUT_COMMAND, 5000)
self.set_property_int(nfc.NP_TIMEOUT_COM, 1000)
self.set_property_int(nfc.NP_TIMEOUT_ATR, 1000)
# self.set_property_bool(nfc.NP_INFINITE_SELECT, False)
def list_passive_targets(self):
logger.debug("list_passive_targets()")
result = []
max_targets_length = 16
nt = ffi.new("nfc_target[{}]".format(max_targets_length))
# time.sleep(0.5) # 50ms removes error "libnfc.driver.pn532_spi Unable to wait for SPI data. (RX)"
ret = nfc.nfc_initiator_list_passive_targets(self._device, self.nm[0],
nt, max_targets_length)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("list_passive_targets() error: {}".format(sErrorMessages[ret]))
for target_n in range(ret):
result.append(nt[target_n])
logger.info("list_passive_targets() num_targets: {}".format(ret))
logger.info(*(nfc_helper.print_target(target) for target in result))
# time.sleep(0.5) # 50ms removes error "libnfc.driver.pn532_spi Unable to wait for SPI data. (RX)"
return ret, result
def select_passive_target(self, initdata=None):
logger.debug("select_passive_target")
nt = ffi.new("nfc_target*")
if initdata is None:
ret = nfc.nfc_initiator_select_passive_target(self._device, self.nm[0], ffi.NULL, 0, nt)
else:
ret = nfc.nfc_initiator_select_passive_target(self._device, self.nm[0],
initdata, len(initdata), nt)
self.last_err = ret
if ret < nfc.NFC_SUCCESS:
logger.warning("select_passive_target() error: {}, {}".format(ret, sErrorMessages[ret]))
return ret, nt
def deselect_target(self, *args, **kwargs):
raise NotImplementedError("deselect_target() not implemented")
def select_dep_target(self, *args, **kwargs):
raise NotImplementedError("select_dep_target() not implemented")
def poll_targets(self, *args, **kwargs):
raise NotImplementedError("poll_targets() not implemented")
def transceive_bytes(self, txbytes, timeout=None):
logger.debug("transceive_bytes()")
if timeout is None:
timeout = self.timeout
logger.info('I>T[%2X]: %s' % (len(txbytes), hexbytes(txbytes)))
tx_len = len(txbytes)
self._txbytes[0:tx_len] = txbytes
rx_len = MAX_FRAME_LEN
ret = nfc.nfc_initiator_transceive_bytes(self._device, self._txbytes, tx_len,
self._rxbytes, rx_len, timeout)
self.last_err = ret
data = bytearray(ffi.buffer(self._rxbytes, ret))
logger.info('I<T[%2X]: %s' % (len(data), hexbytes(data)))
return data, ret
def transceive_bits(self, *args, **kwargs):
raise NotImplementedError("transceive_bits() not implemented")
if __name__ == "__main__":
logging.basicConfig(level=logging.DEBUG)
logger.debug("Debug message")
logger.info("Info message")
logger.warning("Warning message")
logger.error("Error message")
logger.critical("Critical message")
print("logger.level: ", logger.level)
print(__name__)
print("logger.name: ", logger.name)
version = get_version_str()
print("libNfc version:", version)
dev_list = list_devices(verbose=True)
i = NfcInitiator(dev_list[1], verbosity=1)
# print(i._device_name)
# try:
# i.set_property_bool(nfc.NP_HANDLE_PARITY+100, True)
# except Exception as e:
# print("Exception test:", e)
res, nt = i.list_passive_targets()
# for k in range(res):
# nfc_helper.print_target(nt[k])
# print("list_passive_targets() res: ", res)
# pprint_cdata(nt)
if res == 0:
print("No tags found")
exit(0)
initdata = nt[0].nti.nai.abtUid[0:nt[0].nti.nai.szUidLen]
# print("initdata: ", initdata, len(initdata))
try:
res = i.select_passive_target(initdata=initdata)
except Exception as e:
print("Exception test:", e)
res = i.select_passive_target()
# res = i.select_passive_target()
print("select_passive_target() res: ", res)
nt = res
# print("nt: ", nt)
# print("nt.nti.nai.abtUid: ", nt.nti.nai.abtUid)
# print("nt.nti.nai.szUidLen: ", nt.nti.nai.szUidLen)
# print("nt.nti.nai.abtAtqa: ", nt.nti.nai.abtAtqa)
# print("nt.nti.nai.btSak: ", nt.nti.nai.btSak)
# print("nt.nti.nai.abtAts: ", nt.nti.nai.abtAts)
# print("nt.nti.nai.szAtsLen: ", nt.nti.nai.szAtsLen)
print("select_passive_target: nt: ", nt, type(nt))
# pprint_cdata(nt)
# for i in range(nt.nti.nai.szAtsLen):
# print("nt.nti.nai.szAtsLen[{}]: {:02x}".format(i, nt.nti.nai.abtAts[i]))
# for i in range(nt.nti.nai.szUidLen):
# print("nt.nti.nai.szUidLen[{}]: {:02x}".format(i, nt.nti.nai.abtUid[i]))
t = NfcTarget(dev_list[0], verbosity=1)
data, data_len = t.receive_bytes()
# print("receive_bytes() ret: ", data, len)
data, data_len = i.transceive_bytes(data)
# print("transceive_bytes() ret: ", ret)
data_len = t.send_bytes(data)
| 13,994 | Python | .py | 309 | 37.333333 | 117 | 0.617171 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,311 | cffi_main.py | CardToolz_libnfc_mitm_cffi/cffi_main.py | #!/usr/bin/python3
from libnfc_ffi.libnfc_ffi import ffi, libnfc as nfc
ctx = ffi.new("nfc_context**")
print("nfc_init"), nfc.nfc_init(ctx)
c = ctx[0]
conn_strings = ffi.new("nfc_connstring[10]")
d = nfc.nfc_list_devices(c, conn_strings, 10)
print("nfc_list_devices", d)
device = nfc.nfc_open(c, conn_strings[0])
print("nfc_open", device)
nfc.nfc_close(device)
print("nfc_close")
| 386 | Python | .py | 12 | 30.666667 | 52 | 0.718157 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,312 | pn532mitm.py | CardToolz_libnfc_mitm_cffi/pn532mitm.py | #!/usr/bin/python3
#
# pn532mitm.py - NXP PN532 Man-In-The_Middle - log conversations between TAG and external reader
#
'''
IP = 172.16.0.1
pi:raspberry
Initiator must be connected to SPI
Target must be connected to UART
config file:
/etc/nfc/libnfc.conf
allow_autoscan = true
allow_intrusive_scan = false
log_level = 1
device.name = "_PN532_SPI"
device.connstring = "pn532_spi:/dev/spidev0.0:1953000"
#device.name = "_PN532_I2c"
#device.connstring = "pn532_i2c:/dev/i2c-1"
device.name = "_PN532_UART"
device.connstring = "pn532_uart:/dev/ttyS0"
'''
from nfc_wrapper import *
from nfc_helper import *
from NFCRelay import *
from libnfc_ffi.libnfc_ffi import libnfc as nfc
from datetime import datetime
import argparse, os, sys
import random
import logging
logger = logging.getLogger(__name__)
fname_main = os.path.basename(__file__)
cwd = os.getcwd()
fname_date = datetime.now().strftime('%H_%M_%S_%d_%m_%Y')
log_fname_default = "%s_%s_log.json" % (fname_main, fname_date)
logs_path = "{}/logs/".format(cwd)
if not os.path.exists(logs_path):
os.mkdir(logs_path)
easy_framing = True
# target_dev_num_default = 1 # make it command line params
# initiator_dev_num_default = 0 # make it command line params
target_dev_num_default = 0 # make it command line params
initiator_dev_num_default = 1 # make it command line params
def data_hook(direction, data, easy_framing):
send_fragmented = False
print ("Data hook, send_fragmented: %s" % send_fragmented)
return send_fragmented, data
def main():
logging.basicConfig(format='%(asctime)s,%(msecs)03d %(levelname)-8s [%(filename)s:%(lineno)d] %(message)s',
datefmt='%Y-%m-%d:%H:%M:%S', level=logging.INFO)
parser = argparse.ArgumentParser(description='LibNFC relay tool')
parser.add_argument("-l", "--list-devices", dest="list_devices", action='store_true', help="List devices and exit")
parser.add_argument("-o", "--log-filename", dest="log_fname", default=log_fname_default, type=str, help="Output JSON log filename")
parser.add_argument("-i", "--initiator", dest="initiator_dev_num", default=initiator_dev_num_default, type=int, help="Reader device number")
parser.add_argument("-t", "--target", dest="target_dev_num", default=target_dev_num_default, type=int, help="Emulator device number")
parser.add_argument("-n", "--no-easy-framing", dest="no_easy_framing", action='store_true', help="Do not use easy framing. transfer data as frames instead of APDUs")
parser.add_argument("-p", "--print-log", dest="print_log", action='store_true', help="Print log to stdout")
parser.add_argument("-H", "--hook-data", dest="hook_data", action='store_true', help="Use data hook function for data processing")
args = parser.parse_args()
log_fname = logs_path + args.log_fname
initiator_dev_num = args.initiator_dev_num
target_dev_num = args.target_dev_num
easy_framing = not args.no_easy_framing
print_log = args.print_log
list_devs = args.list_devices
hook_data = args.hook_data
print (" *** LibNFC relay tool ***")
print ("tag <---> initiator (relay) <---> target (relay) <---> original reader\n")
print ("%s uses LibNFC ver %s" % (fname_main, get_version_str()))
if list_devs:
devs_list = list_devices(True)
if len(devs_list) < 2:
print ("Found ", len(devs_list), "... Needed 2.\nExitng...")
else:
print ("Initiator dev num:", initiator_dev_num_default)
print ("Target dev num:", target_dev_num_default)
nfc.nfc_exit()
sys.exit()
r = NFCRelay(initiator_dev_num, target_dev_num, easy_framing=easy_framing, log_fname=log_fname)
if r is None:
print ("Can't create NFCRelay object with provided device numbers")
nfc.nfc_exit()
sys.exit()
if hook_data:
print ("Using data hook")
r.set_data_hook(data_hook)
r.reader_setup()
if r.pndReader is None:
print ("Can't open reader")
nfc.nfc_exit()
sys.exit()
print ("****** Waiting for source tag/device ******\n")
tag_count = r.reader_get_targets()
if tag_count == 0:
print ("No tag/device found. Exiting...")
nfc.nfc_exit()
sys.exit()
else:
print ("Found ", tag_count, " tag(s)/device(s)")
for target in r.passive_targets_list:
print ("\tTag info:")
print_target(target)
print("Selecting 1st target by default")
r.select_target()
print("Real target:")
print_target(r.real_target)
# r.emulator_prepare_from_target()
# print("Emulated target:")
# print_target(r.emulated_target)
print ("****** Waiting for reader ******\n")
r.emulator_setup()
if r.pndTag is None or r.pndTag.get_last_err() < 0:
err = r.pndTag.get_last_err()
print("Can't open emulator: {}, {}".format(err, sErrorMessages[err]))
# nfc_exit()
sys.exit()
print("Emulated target:")
print_target(r.emulated_target)
print ("Done, relaying frames now...\n")
r.relay_frames()
print("Saving log to file: %s" % log_fname)
r.fl.save()
if print_log:
print ("\n!************** Log Out ***************")
r.log_print()
if __name__ == "__main__":
main() | 5,330 | Python | .py | 126 | 37.087302 | 173 | 0.64682 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,313 | pn532replay.py | CardToolz_libnfc_mitm_cffi/pn532replay.py | #!/usr/bin/python3
#
# pn532mitm.py - NXP PN532 Man-In-The_Middle - log conversations between TAG and external reader
#
'''
IP = 172.16.0.1
pi:raspberry
Initiator must be connected to SPI
Target must be connected to UART
config file:
/etc/nfc/libnfc.conf
allow_autoscan = true
allow_intrusive_scan = false
log_level = 1
device.name = "_PN532_SPI"
device.connstring = "pn532_spi:/dev/spidev0.0:1953000"
#device.name = "_PN532_I2c"
#device.connstring = "pn532_i2c:/dev/i2c-1"
device.name = "_PN532_UART"
device.connstring = "pn532_uart:/dev/ttyS0"
'''
# from nfc_ctypes import *
from nfc_wrapper import *
from nfc_helper import *
from NFCReplay import *
from datetime import datetime
import argparse, os, sys
import random
fname_main = os.path.basename(__file__)
cwd = os.getcwd()
fname_date = datetime.now().strftime('%H_%M_%S_%d_%m_%Y')
log_fname_default = "%s_%s_log.json" % (fname_main, fname_date)
logs_path = "{}/logs/".format(cwd)
if not os.path.exists(logs_path):
os.mkdir(logs_path)
easy_framing = True
# target_dev_num_default = 1 # make it command line params
# initiator_dev_num_default = 0 # make it command line params
target_dev_num_default = 0 # make it command line params
initiator_dev_num_default = 1 # make it command line params
def data_hook(direction, data, easy_framing):
send_fragmented = False
print ("Data hook, send_fragmented: %s, data len: %d" % (send_fragmented, len(data)))
return send_fragmented, data
def main():
parser = argparse.ArgumentParser(description='LibNFC relay tool')
parser.add_argument("-l", "--list-devices", dest="list_devices", action='store_true', help="List devices and exit")
parser.add_argument("-o", "--log-filename", dest="log_fname", default=log_fname_default, type=str, help="Output JSON log filename")
parser.add_argument("-i", "--initiator", dest="initiator_dev_num", default=initiator_dev_num_default, type=int, help="Reader device number")
parser.add_argument("-t", "--target", dest="target_dev_num", default=target_dev_num_default, type=int, help="Emulator device number")
parser.add_argument("-n", "--no-easy-framing", dest="no_easy_framing", action='store_true', help="Do not use easy framing. transfer data as frames instead of APDUs")
parser.add_argument("-p", "--print-log", dest="print_log", action='store_true', help="Print log to stdout")
parser.add_argument("-H", "--hook-data", dest="hook_data", action='store_true', help="Use data hook function for data processing")
args = parser.parse_args()
log_fname = logs_path + args.log_fname
initiator_dev_num = args.initiator_dev_num
target_dev_num = args.target_dev_num
easy_framing = not args.no_easy_framing
print_log = args.print_log
list_devs = args.list_devices
hook_data = args.hook_data
print (" *** LibNFC re(p)lay tool ***")
print ("tag <---> initiator (relay) <---> target (relay) <---> original reader\n")
print ("%s uses LibNFC ver %s" % (fname_main, get_version_str()))
if list_devs:
devs_list = list_devices(True)
if len(devs_list) < 2:
print ("Found ", len(devs_list), "... Needed 2.\nExitng...")
else:
print ("Initiator dev num:", initiator_dev_num_default)
print ("Target dev num:", target_dev_num_default)
nfc_exit()
sys.exit()
initiator_dev_num = -1
r = NFCRelay(initiator_dev_num, target_dev_num, easy_framing=easy_framing, log_fname=log_fname)
if r is None:
print ("Can't create NFCRelay object with provided device numbers")
nfc_exit()
sys.exit()
if hook_data:
print ("Using data hook")
r.set_data_hook(data_hook)
log = "./logs/test_card_log.json"
ret = r.reader_setup(log_fname=log)
if r.pndReader is None:
print ("Can't open reader/source file")
# nfc_exit()
sys.exit()
if ret:
print ("****** Waiting for source tag/device ******\n")
tag_count = r.reader_get_targets()
if tag_count == 0:
print ("No tag/device found. Exiting...")
nfc_exit()
sys.exit()
else:
print ("Found ", tag_count, " tag(s)/device(s)")
for target in r.passive_targets_list:
print ("\tTag info:")
print_target(target)
print("Selecting 1st target by default")
r.reader_select_target()
print("Real target:")
print_target(r.real_target)
r.emulator_prepare_from_target()
else:
print("Using log file: %s as a data source" % log)
r.emulator_prepare_wo_target()
pass
print("Emulated target:")
print_target(r.emulated_target)
print ("****** Waiting for reader ******\n")
r.emulator_setup()
if r.pndTag is None:
print ("Can't open emulator")
nfc_exit()
sys.exit()
print ("Done, relaying frames now...\n")
r.relay_frames()
print("Saving log to file: %s" % log_fname)
r.fl.save()
if print_log:
print ("\n!************** Log Out ***************")
r.log_print()
if __name__ == "__main__":
main() | 5,208 | Python | .py | 126 | 35.626984 | 173 | 0.637084 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,314 | NFCRelay.py | CardToolz_libnfc_mitm_cffi/NFCRelay.py | # from nfc_ctypes import *
from nfc_wrapper import *
from nfc_helper import *
from libnfc_ffi.libnfc_ffi import libnfc as nfc
from time import time, sleep
from enum import Enum
import logging
logger = logging.getLogger(__name__)
hex2str = lambda x: bytearray.fromhex(x.replace(" ", ""))
apple_frame_sequence = [] # Apple specific frame sequence
apple_frame_sequence.append(hex2str('6a 02 c8 01 00 03 00 02 79 00 00 00 00 c2 d8'))
apple_frame_sequence.append(hex2str('52'))
apple_frame_sequence.append(hex2str('93 20'))
apple_frame_sequence.append(hex2str('93 70 00 00 00 00 00 9c d9'))
def time_ms():
return int(time() * 1000)
def data_hook_default(direction, data, easy_framing):
send_fragmented = False
return send_fragmented, data
class MitmState(Enum):
FromReader = 0
ReaderCardHook = 1
TransceiveCard = 2
CardReaderHook = 3
ToReader = 4
FromReaderFragment = 5
class NFCRelay:
def __init__(self, initiator_dev_num, target_dev_num, easy_framing=True, log_fname=None):
self.data_hook = data_hook_default
self.initiator_dev_num = initiator_dev_num
self.target_dev_num = target_dev_num
self.easy_framing = easy_framing
self.pndReader = None # NfcInitiator
self.pndTag = None # NfcTarget
self.passive_targets_list = None
self.real_target = None
self.emulated_target = None
self.target_modulation = None
self.targettype = None
self.timeout = 2000
self.fl = FrameLogger(easy_framing=easy_framing, log_fname=log_fname)
self.apple_transport = False
self.dev_list = list_devices(False)
if len(self.dev_list) < 2:
assert False, "Not enough devices found"
self.initiator_dev = self.dev_list[self.initiator_dev_num]
self.target_dev = self.dev_list[self.target_dev_num]
def __del__(self):
nfc_exit()
def set_data_hook(self, data_hook):
self.data_hook = data_hook
def reader_setup(self):
self.pndReader = NfcInitiator(self.initiator_dev, verbosity=0)
self.pndReader.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
# self.pndReader.configure(NP_AUTO_ISO14443_4, True)
# self.pndReader.configure_int(NP_TIMEOUT_COMMAND, self.timeout)
return True
def reader_get_targets(self, timeout_ms=0):
tags_count = 0
start_time = time_ms()
while (start_time + timeout_ms > time_ms()) or (timeout_ms == 0):
if self.apple_transport: # TODO: fix apple transport activation. Transceive bits is not working
# print ("Apple specific transport(travel) card activation")
# self.pndReader.configure(nfc.nfc.NP_HANDLE_CRC, False)
for x in range(0, 2):
self.pndReader.transceive_bytes(apple_frame_sequence[0])
self.pndReader.transceive_bits(apple_frame_sequence[1], 7)
self.pndReader.transceive_bytes(apple_frame_sequence[2])
# # pndReader.initiator_transceive_bytes(str4) # is not necessary
tags_count, self.passive_targets_list = self.pndReader.list_passive_targets()
print("Passive targets list: {}".format(self.passive_targets_list))
# tags_count = len(self.passive_targets_list)
if tags_count > 0:
break
if (start_time + timeout_ms < time_ms()) and (timeout_ms != 0):
print("Timeout", start_time, time_ms())
return tags_count
def select_target(self, tag_index=0):
if tag_index >= len(self.passive_targets_list):
logger.warning("Wrong tag index")
return False
nt = self.passive_targets_list[tag_index]
initdata = nt.nti.nai.abtUid[0:nt.nti.nai.szUidLen]
ret, self.real_target = self.pndReader.select_passive_target(initdata=initdata)
if ret < nfc.NFC_SUCCESS:
logger.info('libnfc error, trying again')
ret, self.real_target = self.pndReader.select_passive_target()
if ret < nfc.NFC_SUCCESS:
raise IOError("NFC Error whilst selecting target")
# def emulator_prepare_from_target(self):
# target_info = nfc_target_info(self.real_target.nti.nai)
# target_modulation = nfc_modulation(self.relay_modtype, self.relay_baud)
# self.emulated_target = nfc_target(target_info, target_modulation)
# # self.emulated_target = nfc_target()
# self.emulated_target.nti.nai.abtAtqa[0] = 0x03
# self.emulated_target.nti.nai.abtAtqa[1] = 0x44
# self.emulated_target.nti.nai.abtUid[0] = 0x08 # Needed for PN532 emulation start
# self.emulated_target.nti.nai.szUidLen = 0x04; # Shrink UID to eliminate SEGFAULTs
# # self.emulated_target.nti.nai.btSak = 0x20
# # https://de.wikipedia.org/wiki/Answer_to_Select
# # ATS = (05) 75 33 92 03
# # (TL) T0 TA TB TC
# # | | | +-- CID supported, NAD supported
# # | | +----- FWI=9 SFGI=2 => FWT=154ms, SFGT=1.21ms
# # | +-------- DR=2,4 DS=2,4 => supports 106, 212 & 424bps in both directions
# # +----------- TA,TB,TC, FSCI=5 => FSC=64
# # It seems hazardous to tell we support NAD if the tag doesn't support NAD but I don't know how to disable it
# # PC/SC pseudo-ATR = 3B 80 80 01 01 if there is no historical bytes
# self.emulated_target.nti.nai.abtAts[0] = 0x75
# self.emulated_target.nti.nai.abtAts[1] = 0x11 # supports 106 baud only if ATS presented on source
# self.emulated_target.nti.nai.abtAts[2] = 0x92
# self.emulated_target.nti.nai.abtAts[3] = 0x03
# self.emulated_target.nti.nai.szAtsLen = 0x04
# # self.emulated_target.nti.nai.abtAts[0] = 0x78
# # self.emulated_target.nti.nai.abtAts[1] = 0x77
# # self.emulated_target.nti.nai.abtAts[2] = 0x88
# # self.emulated_target.nti.nai.abtAts[3] = 0x02
# # self.emulated_target.nti.nai.abtAts[4] = 0x80
# # self.emulated_target.nti.nai.szAtsLen = 0x05
# # if self.real_target.nti.nai.szAtsLen < 4:
# # self.emulated_target.nti.nai.szAtsLen = 0x04
# # self.emulated_target.nti.nai.szAtsLen = 0x00;
# # print("Target type: " + str(self.targettype))
# # if self.targettype == 0x44:
# # self.relay_modtype = NMT_ISO14443B
# # self.relay_baud = NBR_106
# # elif self.targettype == 0x04:
# # self.relay_modtype = NMT_ISO14443A
# # self.relay_baud = NBR_106
# # elif self.targettype == 0x02:
# # self.relay_modtype = NMT_ISO14443A
# # self.relay_baud = NBR_106
# # elif self.targettype == 0x52:
# # self.relay_modtype = NMT_ISO14443A
# # self.relay_baud = NBR_106
# # self.apple_transport = True
# # else:
# # print("Unsupported target type")
# # return False
# return True
def emulator_setup(self):
self.pndTag = NfcTarget(self.target_dev, self.emulated_target)
if self.pndTag.get_last_err():
logger.warning("Failed to create target")
return False
self.emulated_target = self.pndTag.get_target()
self.pndTag.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
# self.pndTag.configure(nfc.NP_AUTO_ISO14443_4, True)
# self.pndTag.configure_int(nfc.NP_TIMEOUT_COMMAND, self.timeout) # TODO: Does not work
return True
def relay_frames(self, timeout_ms=0):
if self.pndReader is None or self.pndTag is None:
logger.warning("Reader or tag not initialized")
return False
is_done = False
index = 0
state = MitmState.FromReader
fragmented = False
self.fl.clear()
# print("Starting relay")
self.pndTag.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
start_time = time_ms()
try:
while (start_time + timeout_ms > time_ms()) or (timeout_ms == 0) and not is_done:
logger.info("State = {}".format(state))
self.pndTag.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
self.pndReader.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
# sleep(0.1)
if state == MitmState.FromReader:
target_recvd, ret = self.pndTag.receive_bytes(timeout=timeout_ms)
self.fl.add_frame_by_data(index=index, time=time(), data=target_recvd, result=ret, direction=FrameDirection.FromReader)
if ret <= nfc.NFC_SUCCESS:
print ("Receive from reader result: ({}) {}".format(ret, sErrorMessages[ret]))
is_done = True
continue
state = MitmState.ReaderCardHook
elif state == MitmState.ReaderCardHook:
if self.data_hook is not None:
fragmented, target_recvd = self.data_hook(FrameDirection.FromReader, target_recvd, self.easy_framing)
state = MitmState.TransceiveCard
elif state == MitmState.TransceiveCard: # TODO: implement fragmented transceive
self.fl.add_frame_by_data(index=index, time=time(), data=target_recvd, result=ret, direction=FrameDirection.ToCard, easy_framing=self.easy_framing)
reader_recvd, ret = self.pndReader.transceive_bytes(target_recvd)
index += 1
self.fl.add_frame_by_data(index=index, time=time(), data=reader_recvd, result=ret, direction=FrameDirection.FromCard, easy_framing=self.easy_framing)
if ret <= nfc.NFC_SUCCESS:
print ("Tag/device transceive result: ({}) {}".format(ret, sErrorMessages[ret]))
is_done = True
continue
state = MitmState.CardReaderHook
elif state == MitmState.CardReaderHook:
if self.data_hook is not None:
fragmented, reader_recvd = self.data_hook(FrameDirection.FromCard, reader_recvd, self.easy_framing)
state = MitmState.ToReader
elif state == MitmState.ToReader:
if fragmented:
ret = self.target_send_fragmented(index=index, data=reader_recvd)
# state = MitmState.FromReaderFragment
state = MitmState.FromReader
# print("fragmented send is done")
else:
ret = self.pndTag.send_bytes(reader_recvd)
self.fl.add_frame_by_data(index=index, time=time(), data=reader_recvd, result=ret, direction=FrameDirection.ToReader, easy_framing=self.easy_framing)
state = MitmState.FromReader
index += 1
# print("ToReader next state: {}".format(state))
if ret <= nfc.NFC_SUCCESS:
print ("Send to reader result: ({}) {}".format(ret, sErrorMessages[ret]))
is_done = True
continue
elif state == MitmState.FromReaderFragment:
# print("FromReaderFragment")
target_recvd, ret = self.target_receive_fragmented(timeout=timeout_ms)
self.fl.add_frame_by_data(index=index, time=time(), data=target_recvd, result=ret, direction=FrameDirection.FromReader, easy_framing=self.easy_framing)
# print_frame(self.fl.get_frame_list()[-1])
if ret <= nfc.NFC_SUCCESS:
print ("Receive from reader result: ({}) {}".format(ret, sErrorMessages[ret]))
is_done = True
continue
# self.pndTag.configure(NP_EASY_FRAMING, self.easy_framing)
# sleep(0.05)
state = MitmState.ReaderCardHook
else:
print("Unknown state")
is_done = True
continue
except AssertionError as error:
logger.error('???? WTF with the radio frontend ????')
logger.error(error)
def log_print(self):
self.fl.print()
def target_receive_fragmented(self, timeout=0):
chunks = []
is_last_chunk = False
pcb = ISO14443_PCB()
pcb.asbyte = 0xA2
print("XUIIII")
self.pndTag.set_property_bool(nfc.NP_EASY_FRAMING, False)
while not is_last_chunk:
recvd, ret = self.pndTag.target_receive_bytes(timeout)
# print("Received frame: {}".format(recvd))
if ret <= nfc.NFC_SUCCESS:
print ("Receive from reader result: ({}) {}".format(ret, sErrorMessages[ret]))
return b'', ret
# pcb.asbyte = recvd[0]
chunks.append(recvd[1:])
is_last_chunk = pcb.iblock.chaining == 0
if not is_last_chunk:
pcb.iblock.block_num = pcb.iblock.block_num ^ 1
self.pndTag.target_send_bytes(bytearray([pcb.asbyte]))
# print("Sent frame: {}".format(bytearray([pcb.asbyte])))
# self.pndTag.configure(NP_EASY_FRAMING, self.easy_framing)
data = b''.join(chunks)
return data, len(data)
def target_send_fragmented(self, index, data, fragment_size=200):
# if fragment_size > 128:
# print("Fragment size can't be more than 0x80")
# return False
if len(data) > fragment_size:
data_chunks = [data[i:i + fragment_size] for i in range(0, len(data), fragment_size)]
else:
data_chunks = [data]
# print("Source data: {}".format(data))
# print("Data chunks: {}".format(data_chunks))
# print("Sending {} fragments".format(len(data_chunks)))
self.easy_framing = False
self.pndTag.set_property_bool(nfc.NP_EASY_FRAMING, self.easy_framing)
pcb = ISO14443_PCB(asbyte=0x12) # 0x12/0x13 for phone testing
# block_num = 0
for chunk in data_chunks:
is_last_chunk = chunk == data_chunks[-1]
pcb.iblock.block_num = pcb.iblock.block_num ^ 1
if is_last_chunk: # last chunk
pcb.iblock.chaining = 0
# else:
frame = bytearray([pcb.asbyte]) + chunk
# print("Sending frame: {}".format(frame))
# self.pndTag.configure(NP_EASY_FRAMING, False)
ret = self.pndTag.send_bytes(frame)
self.fl.add_frame_by_data(index=index, time=time(), data=frame, result=ret, direction=FrameDirection.ToReader, easy_framing=False)
# print_frame(self.fl.get_frame_list()[-1])
if ret <= nfc.NFC_SUCCESS:
print ("Send to reader result: ({}) {}".format(ret, sErrorMessages[ret]))
return ret
# pcb.iblock.block_num = pcb.iblock.block_num ^ 1
# self.pndTag.configure(NP_EASY_FRAMING, False)
if not is_last_chunk:
# sleep(0.1)
recvd, ret = self.pndTag.receive_bytes(timeout=0)
self.fl.add_frame_by_data(index=index, time=time(), data=recvd, result=ret, direction=FrameDirection.FromReader, easy_framing=False)
# print_frame(self.fl.get_frame_list()[-1])
if ret <= nfc.NFC_SUCCESS:
print ("Receive from reader result: ({}) {}".format(ret, sErrorMessages[ret]))
return ret
else:
pass
# sleep fo 0.1 sec to avoid "RF transmission error" on PN532
sleep(0.05)
return len(data_chunks)
| 16,067 | Python | .py | 299 | 41.866221 | 173 | 0.587245 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,315 | libnfc_ffi.py | CardToolz_libnfc_mitm_cffi/libnfc_ffi/libnfc_ffi.py | #!/usr/bin/python3
from cffi import FFI
ffi = FFI()
def fetch_nfc_functions(hfile):
lines = []
with open(hfile) as f:
for ln in f:
if ln.startswith("NFC_EXPORT"):
ln = ln.replace("NFC_EXPORT", "extern")
ln = ln.replace("ATTRIBUTE_NONNULL(1)", "")
# ln = ln.replace("typedef ", "")
lines.append(ln)
return "".join(lines)
def fetch_nfc_types(hfile):
with open(hfile) as f:
data = f.read()
data = data[data.index("#endif") + len("#endif"): data.index("# pragma pack()")]
data = data.replace("NFC_BUFSIZE_CONNSTRING", "1024")
# data = data.replace("typedef ", "")
return data
def fetch_nfc_constants(hfile):
lines = []
with open(hfile) as f:
for ln in f:
if ln.startswith("#define"):
lines.append(ln)
return "".join(lines)
def ffi_print_declarations(ffi):
for key in ffi._parser._declarations:
print(key, ffi._parser._declarations[key])
cdef_types = fetch_nfc_types("/usr/include/nfc/nfc-types.h")
# print (cdef_types)
cdef_funcs = fetch_nfc_functions("/usr/include/nfc/nfc.h")
cdef_defs = fetch_nfc_constants("/usr/include/nfc/nfc.h")
ffi.cdef(cdef_types, packed=True)
ffi.cdef(cdef_funcs, packed=True)
ffi.cdef(cdef_defs, packed=True)
libnfc = ffi.dlopen("libnfc.so")
if __name__ == "__main__":
print("CFFI binding for libNFC")
ver_str = ffi.string(libnfc.nfc_version()).decode("utf-8")
print("libNFC version:", ver_str)
print("imported types:")
ffi_print_declarations(ffi)
# some constants tst
# print(sErrorMessages[libnfc.NFC_ECHIP])
# print(libnfc.NP_INFINITE_SELECT) | 1,722 | Python | .py | 47 | 30.297872 | 89 | 0.620398 | CardToolz/libnfc_mitm_cffi | 8 | 0 | 1 | LGPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,316 | GUI.py | AlexSanfilippo_ProceduralMeshGeneration/GUI.py | """
started: April 19th, 2024
update: April 19th, 2024
goals:
-[V]render quad on screen
-[V]textured quad
-[V]control scale and position
-[V]transparent textures
-[V]texture atlasing
-[V]Multiple GUI Elements
-static elements.icons
-displays, ie, not buttons
-want GUI class that holds several GUI elements
instantiate GUI class
add element to class
GUI.draw
-draws all elements
-separate draw calls for now
-so just a loop
-[V]buttons
-subclass existing gui element class
-[V]Check if mouse is inside button
-[V]Check if mouse is clicked
-[V]button to spawn ships
-ie, tie button into application
-[V]Change texture on mouse hover (hover events)
-send mouse position into draw
-allow each element to figure itself out
-[V]activate/deactivate button by clicking other buttons
"""
import gc
import math
import glm
from OpenGL.GL import *
import numpy as np
from OpenGL.GL.shaders import compileProgram, compileShader
from TextureLoader import load_texture
vertex_src = """
# version 330
layout(location = 0) in vec2 position;
layout(location = 1) in vec2 texture;
out vec2 texture_coordinates;
void main()
{
texture_coordinates = texture;
gl_Position = vec4(position, 0.f, 1.f);
}
"""
fragment_src = """
#version 330 core
in vec2 texture_coordinates;
out vec4 fragment_color;
uniform sampler2D gui_texture;
uniform vec4 color;
void main()
{
//fragment_color = vec4(0.7, 0.1, 0.5, 0.5f);
fragment_color = color * texture(gui_texture, texture_coordinates);
}
"""
vertex_src_reference = """
# version 330
layout(location = 0) in vec3 a_position;
layout(location = 1) in vec3 a_color;
layout(location = 2) in vec2 a_texture;
uniform mat4 rotation;
out vec3 v_color;
out vec2 v_texture;
void main()
{
gl_Position = rotation * vec4(a_position, 1.0);
v_color = a_color;
v_texture = a_texture;
}
"""
fragment_src_reference = """
# version 330
in vec3 v_color;
in vec2 v_texture;
out vec4 out_color;
uniform sampler2D s_texture;
void main()
{
out_color = texture(s_texture, v_texture); // vec4(v_color, 1.0);
}
"""
#create default shaders and textures
# shader_default = compileProgram(
# compileShader(
# vertex_src,
# GL_VERTEX_SHADER
# ),
# compileShader(
# fragment_src,
# GL_FRAGMENT_SHADER
# ),
# )
class GUI:
"""
Create GUI For application interaction with mouse, or displaying information
"""
def __init__(
self,
screen_size=(800, 400)
):
self.screen_size = screen_size
self.elements = []
self.buttons = []
self.context_id_to_status = {}
self.context_id_to_element = {}
def add_element(
self,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
atlas_size=1,
atlas_coordinate=0,
context_status=True,
context_id='default'
):
element = Element(
shader=shader,
texture=texture,
position=position,
scale=scale,
screen_size=self.screen_size,
atlas_size=atlas_size,
atlas_coordinate=atlas_coordinate,
context_id=context_id,
)
self.elements.append(element)
self.update_context_maps(element=element, status=context_status)
def add_text_element(
self,
font_texture,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
atlas_size=1,
atlas_coordinate=0,
context_status=True,
context_id='default',
text='',
font_color=(1.0, 1.0, 1.0, 1.0),
font_size=1.0,
):
element = Element(
shader=shader,
texture=texture,
position=position,
scale=scale,
screen_size=self.screen_size,
atlas_size=atlas_size,
atlas_coordinate=atlas_coordinate,
context_id=context_id,
)
element.add_text_box(
texture=font_texture,
text=text,
color=font_color,
font_size=font_size,
)
self.elements.append(element)
self.update_context_maps(element=element, status=context_status)
def draw(self):
#TODO: draw all elements at once! (batch rendering)
for element in self.elements + self.buttons:
element.draw()
def add_button(
self,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
atlas_size=1,
atlas_coordinate=(0,),
click_function=None,
context_id='default',
context_status=True,
color=(1.0, 1.0, 1.0, 1.0),
position_mode='center',
**click_function_kwargs,
):
button = Button(
shader=shader,
texture=texture,
position=position,
scale=scale,
screen_size=self.screen_size,
atlas_size=atlas_size,
atlas_coordinate=atlas_coordinate,
click_function=click_function,
context_id=context_id,
color=glm.vec4(color),
position_mode=position_mode,
click_function_kwargs=click_function_kwargs,
)
self.buttons.append(button)
self.update_context_maps(element=button, status=context_status)
def add_text_button(
self,
font_texture,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
atlas_size=1,
atlas_coordinate=(0,),
click_function=None,
context_id='default',
context_status=True,
color=(1.0, 1.0, 1.0, 1.0),
position_mode='center',
text='',
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
**click_function_kwargs,
):
button = Button(
shader=shader,
texture=texture,
position=position,
scale=scale,
screen_size=self.screen_size,
atlas_size=atlas_size,
atlas_coordinate=atlas_coordinate,
click_function=click_function,
context_id=context_id,
color=glm.vec4(color),
position_mode=position_mode,
click_function_kwargs=click_function_kwargs,
)
button.add_text_box(
texture=font_texture,
font_size=font_size,
color=font_color,
text=text
)
self.buttons.append(button)
self.update_context_maps(element=button, status=context_status)
def button_update(self, position_mouse, left_click, right_click):
"""
tell buttons to check for input (mouse hover or click)
:param position_mouse:
:param left_click:
:param right_click:
:return: No Return
"""
position_mouse_normalized = glm.vec2(
(
position_mouse[0]/self.screen_size[0],
position_mouse[1]/self.screen_size[1]
)
)
for button in self.buttons:
button.update(position_mouse=position_mouse_normalized, left_click=left_click, right_click=right_click)
def set_screen_size(self, screen_size):
self.screen_size = screen_size
def update_context_maps(self, element, status=True):
if element.context_id in self.context_id_to_status.keys():
self.context_id_to_status[element.context_id] = status
self.context_id_to_element[element.context_id].append(element)
else:
self.context_id_to_status[element.context_id] = status
self.context_id_to_element[element.context_id] = [element]
def build_elements_list(self):
self.elements = []
self.buttons = []
for context_id, status in self.context_id_to_status.items():
if status:
elements = self.context_id_to_element[context_id]
for element in elements:
if type(element) == Element:
self.elements.append(element)
else:
self.buttons.append(element)
def switch_context_status(self, context_id, status):
"""
turn a context on or off
:param context_id: id of the element to turn on/off
:param status: which status to switch it to
"""
self.context_id_to_status[context_id] = status
self.build_elements_list()
def toggle_context_status(self, context_id):
self.switch_context_status(
context_id=context_id,
status=not(self.context_id_to_status[context_id]),
)
class Element:
def __init__(
self,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
screen_size=(800, 400),
atlas_size=1,
atlas_coordinate=0,
context_id='default',
color=glm.vec4(1.0, 1.0, 1.0, 1.0),
):
self.color_loc = None
self.model_loc = None
self.VBO = None
self.VAO = None
if shader == None:
shader_default = compileProgram(
compileShader(vertex_src, GL_VERTEX_SHADER),
compileShader(fragment_src, GL_FRAGMENT_SHADER)
)
self.shader = shader_default
else:
self.shader = shader
if texture == None:
gui_textures = glGenTextures(2)
load_texture("Textures/debug_diffuse_coordinates.png", gui_textures[0])
self.texture = gui_textures[0]
else:
self.texture = texture
self.position = glm.vec2(position)
self.scale = scale
self.screen_size = screen_size
self.screen_size = screen_size
self.atlas_size = atlas_size
self.atlas_coordinate = atlas_coordinate
self.context_id = context_id
self.color = color
self.vertices_count = 4
self.vertices = self.generate_vertices()
self.buffer_setup()
self.text_box = None
def generate_vertices(self):
text_coords = self.get_texture_coordinates_atlas()
#OpenGL Screen Coordinates go from -1,-1 (lower left) to 1,1 (upper right)
return np.array(
[
# upper left
-1.0 * self.scale[0] + self.position[0], 1.0 * self.scale[1] + self.position[1], text_coords[0].x, text_coords[0].y,
# lower left
-1.0 * self.scale[0] + self.position[0], -1.0 * self.scale[1] + self.position[1], text_coords[1].x, text_coords[1].y,
#upper right
1.0 * self.scale[0] + self.position[0], 1.0 * self.scale[1] + self.position[1], text_coords[2].x, text_coords[2].y,
#lower right
1.0 * self.scale[0] + self.position[0], -1.0 * self.scale[1] + self.position[1], text_coords[3].x, text_coords[3].y,
],
dtype=np.float32
)
def clean_up(self):
glDeleteVertexArrays(1, [self.VAO])
glDeleteBuffers(1, [self.VBO])
del self.model_loc
del self.color_loc
gc.collect()
def buffer_setup(self):
# quad VAO
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# quad position vertices (vertex attribute)
glVertexAttribPointer(0, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * self.vertices_count, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
# quad texture coords
glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * self.vertices_count, ctypes.c_void_p(8))
glEnableVertexAttribArray(1)
self.model_loc = glGetUniformLocation(self.shader, "model")
self.color_loc = glGetUniformLocation(self.shader, "color")
def replace_vertices(self):
"""
like buffer_setup function, but does not allocate a new buffer. Just gives new
vertex info to the existing buffer. Thus, no clean_up call neccessary.
:return: None
"""
glBindVertexArray(self.VAO)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_DYNAMIC_DRAW)
glBindVertexArray(0)
glBindBuffer(GL_ARRAY_BUFFER, 0)
# # quad position vertices (vertex attribute)
# glVertexAttribPointer(0, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * self.vertices_count, ctypes.c_void_p(0))
# glEnableVertexAttribArray(0)
# # quad texture coords
# glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * self.vertices_count, ctypes.c_void_p(8))
# glEnableVertexAttribArray(1)
#
# self.model_loc = glGetUniformLocation(self.shader, "model")
# self.color_loc = glGetUniformLocation(self.shader, "color")
def draw(self):
glUseProgram(self.shader)
glBindVertexArray(self.VAO)
glActiveTexture(GL_TEXTURE0)
glBindTexture(GL_TEXTURE_2D, self.texture)
# rotate, translate, and scale
# model = glm.mat4(1.0)
# model = glm.translate(model, self.position)
# todo: this rotation smells fishy
# model = glm.rotate(model, self.rotation_magnitude.x, self.rotation_axis)
# model = glm.scale(model, self.scale)
# glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, glm.value_ptr(model))
glUniform4fv(self.color_loc, 1, glm.value_ptr(self.color))
glEnable(GL_BLEND)
glDisable(GL_DEPTH_TEST)
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA)
glDrawArrays(GL_TRIANGLE_STRIP, 0, self.vertices_count)
glDisable(GL_BLEND)
glEnable(GL_DEPTH_TEST)
if self.text_box:
self.text_box.draw()
def update_text(self, text=None, color=None):
self.text_box.update_text(text=text, color=color)
def get_texture_coordinates_atlas(self):
texture_coordinates = [
glm.vec2(0.0, 1.0),
glm.vec2(0.0, 0.0),
glm.vec2(1.0, 1.0),
glm.vec2(1.0, 0.0)
]
# texture atlas index into row and column indices
column_index = float(self.atlas_coordinate % self.atlas_size)
row_index = 1.0 - math.floor(self.atlas_coordinate / self.atlas_size)
# row and column indices into lower and upper bounds of texture coords (4 total)
# column (x-axis)lower and upper
lower_texture_coord_x_axis = 1.0 / self.atlas_size * column_index
upper_texture_coord_x_axis = 1.0 / self.atlas_size * (column_index + 1.0)
# assuming square textures
subtexture_magnitude = upper_texture_coord_x_axis - lower_texture_coord_x_axis
lower_texture_coord_y_axis = 1.0 / self.atlas_size * row_index
upper_texture_coord_y_axis = 1.0 / self.atlas_size * (row_index + 1.0)
# update the current vertices' text coords
for text_coord in texture_coordinates:
text_coord[0] = \
lower_texture_coord_x_axis \
+ text_coord[0] \
* subtexture_magnitude
text_coord[1] = \
lower_texture_coord_y_axis \
+ text_coord[1] \
* subtexture_magnitude
return texture_coordinates
def add_text_box(self, texture, text="DEFAULT TEXT", color=(1.0, 1.0, 1.0, 1.0), font_size=1.0, width=1.0, centered=True):
self.text_box = TextBox(
texture=texture,
position=(self.position.x, self.position.y),
scale=self.scale,
screen_size=self.screen_size,
context_id=self.context_id,
text=text,
font_size=font_size,
width=self.scale[0]*2*width,
color=glm.vec4(color),
centered=centered,
)
class Button(Element):
"""
Clickable GUI Elements
"""
def __init__(
self,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
screen_size=(800, 400),
atlas_size=1,
atlas_coordinate=(0, 0),
click_function=None,
context_id='default',
color=(1.0, 1.0, 1.0, 1.0),
position_mode='center',
click_function_kwargs=None,
):
"""
:param shader:
:param texture:
:param position: tuple of length 3.
:param scale: between 0 and 1, tuple of length 2
:param screen_size: OpenGL Window dimensions
:param atlas_size: length of atlas (square atlas only)
:param atlas_coordinate: texture atlas coordinate. tuple of 2 items
:param click_function: function called upon click
:param context_id: context for turning elements on/off
:param color: pass vec4 color to shader
:param position_mode: position by default is center. options: bottom_left, bottom_right, top_left, top_right
:param click_function_kwargs: args to pass to function.
"""
position = list(position)
if position_mode == 'center':
pass
elif position_mode == 'top_left':
position[0] += scale[0]
position[1] -= scale[1]
elif position_mode == 'top_right':
position[0] -= scale[0]
position[1] -= scale[1]
elif position_mode == 'bottom_left':
position[0] += scale[0]
position[1] += scale[1]
elif position_mode == 'bottom_right':
position[0] -= scale[0]
position[1] += scale[1]
super().__init__(
shader=shader,
texture=texture,
position=position,
scale=scale,
screen_size=screen_size,
atlas_size=atlas_size,
atlas_coordinate=atlas_coordinate[0],
context_id=context_id,
color=color
)
self.generate_bounds()
self.click_function = click_function
self.atlas_coordinate_on = atlas_coordinate[0]
self.atlas_coordinate_off = atlas_coordinate[1]
self.click_function_kwargs = click_function_kwargs
def generate_bounds(self):
"""
Define vertical and horizontal limits that allow for testing if cursor is
inside or outside of the button
"""
self.bounds = {
'vertical_lower': 1.0 - ((self.scale[1] + self.position[1]) * 0.5 + 0.5),
'vertical_upper': 1.0 - ((-1.0 * self.scale[1] + self.position[1]) * 0.5 + 0.5),
'horizontal_upper': (self.scale[0] + self.position[0]) * 0.5 + 0.5,
'horizontal_lower': (-1.0 * self.scale[0] + self.position[0]) * 0.5 + 0.5,
}
def check_mouse_hover(self, position_mouse):
"""
check if the mouse is inside this element
:return: Bool
"""
# position_mouse.x = 0.475
# position_mouse.y = 0.049
if position_mouse.x < self.bounds['horizontal_upper'] \
and position_mouse.x > self.bounds['horizontal_lower'] \
and position_mouse.y > self.bounds['vertical_lower'] \
and position_mouse.y < self.bounds['vertical_upper']:
return True
else:
return False
def check_mouse_click(self, left_click):
"""
Check if the mouse is clicked while inside this element
:return:
"""
return left_click
def update(self, position_mouse, left_click=False, right_click=False):
"""
check for mouse input and react
:return:
"""
if self.check_mouse_hover(position_mouse=position_mouse):
self.atlas_coordinate = self.atlas_coordinate_off
self.vertices = self.generate_vertices()
# self.clean_up()
# self.buffer_setup()
self.replace_vertices()
if self.check_mouse_click(left_click):
if self.click_function:
if self.click_function_kwargs:
self.click_function(*list(self.click_function_kwargs.values()))
else:
self.click_function()
else:
print("WARNING: No click function!")
else:
self.atlas_coordinate = self.atlas_coordinate_on
self.vertices = self.generate_vertices()
# self.clean_up()
# self.buffer_setup()
self.replace_vertices()
class Character(Element):
"""
Render a single quad with a given character
"""
def __init__(
self,
shader=None,
texture=None,
position=(0.0, 0.0),
scale=(0.5, 0.5),
screen_size=(800, 400),
context_id='default',
text=None,
font_size=1.0,
width=1.0,
color=(1.0, 1.0, 1.0, 1.0),
centered=False,
):
"""
:param shader:
:param texture:
:param position:
:param scale: Not used here, but leftover from Element
:param screen_size: Pixel Width, Height of screen
:param context_id: For opening/closing GUI contexts
:param text: The text to display
:param font_size: Functions as multiplier.
:param width: Width as fraction of screen [0.0, 1.0]
"""
if shader == None:
shader_default = compileProgram(
compileShader(vertex_src, GL_VERTEX_SHADER),
compileShader(fragment_src, GL_FRAGMENT_SHADER)
)
self.shader = shader_default
else:
self.shader = shader
if texture == None:
gui_textures = glGenTextures(2)
load_texture("Textures/debug_diffuse_coordinates.png", gui_textures[0])
self.texture = gui_textures[0]
else:
self.texture = texture
self.position = glm.vec2((position[0] - scale[0], position[1] + scale[1]))
self.scale = scale
self.screen_size = screen_size
self.screen_size = screen_size
self.context_id = context_id
self.text = text
self.font_size = font_size
self.vertices_count = 4
self.color = glm.vec4(color)
self.centered = centered
self.characters = self.get_characters()
self.width = width
self.box_right = self.position.x + width
self.vertices = self.generate_vertices()
self.buffer_setup()
def get_characters(self):
file_font = open('C://Users//LENOVO//PycharmProjects//ProceduralMeshGeneration//Fonts//my_font.fnt', 'r')
characters = dict()
line_count = 0
for line in file_font:
line = line[:-1]
if line_count < 4:
line_count += 1
continue
else:
line_as_list = line.split()
line_type = line_as_list[0]
line_values = line_as_list[1:]
if line_type == 'kerning':
break
character_values = dict()
for pair in line_values[1:]:
pair = pair.split(sep='=')
character_values[pair[0]] = int(pair[1])
char_id = line_values[0].split(sep='=')[1]
characters[chr(int(char_id))] = character_values
return characters
def generate_vertices(self):
char_info = self.characters[self.text]
width = char_info['width']/512
height = char_info['height']/512
text_coords_upper_left = glm.vec2(char_info['x'], 1.0 - char_info['y'])/512
text_coords_upper_right = glm.vec2(text_coords_upper_left.x + width, text_coords_upper_left.y)
text_coords_lower_right = glm.vec2(text_coords_upper_left.x + width, text_coords_upper_left.y - height)
text_coords_lower_left = glm.vec2(text_coords_upper_left.x, text_coords_upper_left.y - height)
text_coords = [text_coords_upper_left, text_coords_lower_left, text_coords_upper_right, text_coords_lower_right]
#OpenGL Screen Coordinates go from -1,-1 (lower left) to 1,1 (upper right)
return np.array(
[
# upper left
-1.0 * width * self.font_size + self.position[0], 1.0 * height * self.font_size + self.position[1], text_coords[0].x, text_coords[0].y,
# lower left
-1.0 * width * self.font_size + self.position[0], -1.0 * height * self.font_size + self.position[1], text_coords[1].x, text_coords[1].y,
#upper right
1.0 * width * self.font_size + self.position[0], 1.0 * height * self.font_size + self.position[1], text_coords[2].x, text_coords[2].y,
#lower right
1.0 * width * self.font_size + self.position[0], -1.0 * height * self.font_size + self.position[1], text_coords[3].x, text_coords[3].y,
],
dtype=np.float32
)
class TextBox(Character):
"""
Much like Character, except takes a whole string and produces multiple quads
-Uses OpenGL screen Coordinates, ie (-1,-1) in South West, (1,1) in North East
-so a width of 2.0 would have text box cover whole screen
"""
def generate_vertices(self):
vertices = []
vertices_line = []
self.box_left = self.position[0]
box_center_horizontal = (self.box_right + self.box_left) / 2.0
position_initial = self.position.__copy__()
text = self.text
# for index, char in enumerate(text):
index_char = 0
loop_count = 0
history = {}
history_word = {}
while index_char < len(text):
# loop_count += 1
# if loop_count > 100:
# break
if index_char not in set(history.keys()):
history[index_char] = 0
loop_count += 1
char = text[index_char]
char_info = self.characters[char]
width = (char_info['width'] / 512)
height = (char_info['height'] / 512)
x_offset = (char_info['xoffset'] / 512)
y_offset = (char_info['yoffset'] / 512)
x_advance = (char_info['xadvance'] / 512)
text_coords_upper_left = glm.vec2(char_info['x'], 1.0 - char_info['y']) / 512
text_coords_upper_right = glm.vec2(text_coords_upper_left.x + width, text_coords_upper_left.y)
text_coords_lower_right = glm.vec2(text_coords_upper_left.x + width, text_coords_upper_left.y - height)
text_coords_lower_left = glm.vec2(text_coords_upper_left.x, text_coords_upper_left.y - height)
text_coords = [text_coords_upper_left, text_coords_lower_left, text_coords_upper_right, text_coords_lower_right]
width = (char_info['width'] / 512) * self.font_size
height = (char_info['height'] / 512) * self.font_size
x_offset = (char_info['xoffset'] / 512) * self.font_size
y_offset = (char_info['yoffset'] / 512) * self.font_size
x_advance = (char_info['xadvance'] / 512) * self.font_size
upper_left = [(self.position[0] + x_offset), (self.position[1] - y_offset), text_coords[0].x, text_coords[0].y, ]
lower_left = [(self.position[0] + x_offset), (self.position[1] - y_offset - height), text_coords[1].x, text_coords[1].y, ]
upper_right = [(self.position[0] + width + x_offset), (self.position[1] - y_offset), text_coords[2].x, text_coords[2].y, ]
lower_right = [(self.position[0] + width + x_offset), (self.position[1] - y_offset - height), text_coords[3].x, text_coords[3].y, ]
positions = upper_left + lower_left + upper_right + lower_right + upper_right + lower_left
vertices_line += positions
self.position[0] += x_advance
#Check if need newline and handle
if self.position[0] > self.box_right:
if char == ' ':
self.position[0] = self.box_left
self.position[1] -= 0.25 * self.font_size
else:
#if haven't tried newlining at this index yet
index_back_check = index_char
char_back_check = char
while char_back_check != ' ':
char_back_check = text[index_back_check]
#remove char's quad
vertices_line = vertices_line[:-4*6]
#move onto previous char
index_back_check -= 1
#go back to beginning of word
index_char = index_back_check + 1
index_first_letter = index_char + 1
if index_first_letter in set(history_word.keys()):
history_word[index_first_letter] += 1
else:
history_word[index_first_letter] = 1
#move cursor down to next line, and to far left side
if history_word[index_first_letter] <= 1:
self.position[0] = self.box_left
self.position[1] -= 0.25 * self.font_size
else:
#if we've already tried newlining this word, just have to split the word
raise Exception(f'A word at index {index_first_letter} in the text is longer than the text box.')
# index_char = index_char_original - 1
# self.position[0] = self.box_left
# self.position[1] -= 0.25 * self.font_size
if self.centered:
self.center_text_horizontal(box_center_horizontal, vertices_line)
vertices += vertices_line
vertices_line.clear()
self.vertices_count += 0
index_char += 1
self.position = position_initial
if self.centered:
self.center_text_horizontal(box_center_horizontal, vertices_line)
vertices += vertices_line
self.center_text_vertically(vertices, vertices_line)
else:
vertices += vertices_line
return np.array(
vertices,
dtype=np.float32
)
def center_text_vertically(self, vertices, vertices_line):
box_bottom = self.position.y - (self.scale[1] * 2)
box_center_vertical = (self.position.y + box_bottom) / 2
text_center_vertical = (vertices[1] + vertices_line[-3]) / 2
shift = box_center_vertical - text_center_vertical
for index, value in enumerate(vertices):
if index % 4 == 1:
vertices[index] += shift
def center_text_horizontal(self, box_center_horizontal, vertices_line):
text_center_horizontal = (vertices_line[0] + vertices_line[-8]) / 2.0
shift = box_center_horizontal - text_center_horizontal
for index, value in enumerate(vertices_line):
if index % 4 == 0:
vertices_line[index] += shift
def update_text(self, text=None, color=None):
if text:
self.text = text
if color:
self.color = glm.vec4(color)
self.vertices = self.generate_vertices()
# self.clean_up()
# self.buffer_setup()
self.replace_vertices()
def draw(self):
glUseProgram(self.shader)
glBindVertexArray(self.VAO)
glActiveTexture(GL_TEXTURE0)
glBindTexture(GL_TEXTURE_2D, self.texture)
# rotate, translate, and scale
# model = glm.mat4(1.0)
# model = glm.translate(model, self.position)
# todo: this rotation smells fishy
# model = glm.rotate(model, self.rotation_magnitude.x, self.rotation_axis)
# model = glm.scale(model, self.scale)
# glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, glm.value_ptr(model))
glUniform4fv(self.color_loc, 1, glm.value_ptr(self.color))
glEnable(GL_BLEND)
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA)
glDisable(GL_DEPTH_TEST)
# glDrawArrays(GL_TRIANGLE_STRIP, 0, self.vertices_count)
# glDrawArrays(GL_TRIANGLES, 0, self.vertices_count)
glDrawArrays(GL_TRIANGLES, 0, int(len(self.vertices) / 3))
glDisable(GL_BLEND)
glEnable(GL_DEPTH_TEST) | 33,024 | Python | .py | 821 | 29.811206 | 152 | 0.575968 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,317 | SimplePhysics.py | AlexSanfilippo_ProceduralMeshGeneration/SimplePhysics.py | """
Anything physics-related that we are coding ourself (verses using a library like pybullet)
"""
from glm import vec3, normalize, vec4, dot
import logging
logger = logging.getLogger(__name__)
logging.basicConfig()
# logging.root.setLevel(logging.NOTSET)
#Ray-plane intersection
#given
# ray_origin = vec3([10.0, 10.0, 10.0])
# ray_direction = normalize(vec3((0.0, -1.0, 0.0)))
# plane = vec4(0.0, 1.0, 0.0, 0.0)
# ray_origin = vec3((2.0, 3.0, 4.0))
# ray_direction = vec3((0.577, 0.577, 0.577))
# plane = vec4(1.0, 0.0, 0.0, -7.0)
def find_intersection_ray_plane(ray_origin, ray_direction, plane):
"""
Find the intersection between a ray and plane
created with reference to: https://education.siggraph.org/static/HyperGraph/raytrace/rayplane_intersection.htm
:param ray_origin: origin of the ray
:param ray_direction: NORMALIZED direction of the ray
:param plane: defined as vec4(a,b,c,d), where abc are the normal, and d is the distance to (0,0,0)
:return:
"""
point_intersection = vec3(0.0, 0.0, 0.0)
ray_direction = normalize(ray_direction)
unit_normal = plane.xyz
V_d = dot(unit_normal, ray_direction)
if V_d == 0:
logger.info('Ray is parallel to the plane.')
elif V_d > 0:
logger.info('Ray is pointing away from the plane.')
else:
V_0 = -(dot(unit_normal, ray_origin) + plane.w)
t = V_0 / V_d
if t < 0:
logger.info('Ray intersects plane behind origin, ie, no intersection of interest')
else:
point_intersection = ray_origin + ray_direction * t
logger.info(f'{point_intersection=}')
return point_intersection
# find_intersection_ray_plane(ray_origin=ray_origin, ray_direction=ray_direction, plane=plane)
| 1,768 | Python | .py | 43 | 36.44186 | 114 | 0.674052 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,318 | Skybox.py | AlexSanfilippo_ProceduralMeshGeneration/Skybox.py | """
17th July 2024
Following LearnOpenGL.com's tutorial on skyboxes and cubemaps
"""
from OpenGL.GL import *
from TextureLoader import load_cube_map
import numpy as np
import glm
from OpenGL.GL.shaders import compileProgram, compileShader
vertex_src = """
#version 330
layout (location = 0) in vec3 aPos;
out vec3 TexCoords;
uniform mat4 projection;
uniform mat4 view;
void main()
{
TexCoords = aPos;
vec4 pos = projection * view * vec4(aPos, 1.0);
gl_Position = pos;
}
"""
fragment_src = """
#version 330
out vec4 FragColor;
in vec3 TexCoords;
uniform samplerCube skybox;
void main()
{
FragColor = texture(skybox, TexCoords);
}
"""
class Skybox:
"""
Create a skybox using a cubemap
textures: a list of 6 textures paths in the order:
right, left, top, bottom, front, back
"""
def __init__(self, texture_paths, scale=1000):
self.vertices = np.array(
[
# positions
-1.0, 1.0, -1.0,
-1.0, -1.0, -1.0,
1.0, -1.0, -1.0,
1.0, -1.0, -1.0,
1.0, 1.0, -1.0,
-1.0, 1.0, -1.0,
-1.0, -1.0, 1.0,
-1.0, -1.0, -1.0,
-1.0, 1.0, -1.0,
-1.0, 1.0, -1.0,
-1.0, 1.0, 1.0,
-1.0, -1.0, 1.0,
1.0, -1.0, -1.0,
1.0, -1.0, 1.0,
1.0, 1.0, 1.0,
1.0, 1.0, 1.0,
1.0, 1.0, -1.0,
1.0, -1.0, -1.0,
-1.0, -1.0, 1.0,
-1.0, 1.0, 1.0,
1.0, 1.0, 1.0,
1.0, 1.0, 1.0,
1.0, -1.0, 1.0,
-1.0, -1.0, 1.0,
-1.0, 1.0, -1.0,
1.0, 1.0, -1.0,
1.0, 1.0, 1.0,
1.0, 1.0, 1.0,
-1.0, 1.0, 1.0,
-1.0, 1.0, -1.0,
-1.0, -1.0, -1.0,
-1.0, -1.0, 1.0,
1.0, -1.0, -1.0,
1.0, -1.0, -1.0,
-1.0, -1.0, 1.0,
1.0, -1.0, 1.0,
],
dtype=np.float32
) * scale
self.textures = self.create_cube_map(texture_paths)
self.shader = self.create_shader()
self.setup_buffers()
def create_shader(self):
shader = compileProgram(
compileShader(vertex_src, GL_VERTEX_SHADER),
compileShader(fragment_src, GL_FRAGMENT_SHADER)
)
return shader
def setup_buffers(self):
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# position vertices (vertex attribute)
glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 3, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
def create_cube_map(self, texture_paths):
textures = glGenTextures(1)
paths = texture_paths
load_cube_map(paths, textures)
return textures
def draw(self, view, projection):
glUseProgram(self.shader)
glBindVertexArray(self.VAO)
glActiveTexture(GL_TEXTURE0)
glBindTexture(GL_TEXTURE_CUBE_MAP, self.textures)
proj_loc = glGetUniformLocation(self.shader, "projection")
glUniformMatrix4fv(proj_loc, 1, GL_FALSE, projection)
#remove translation
view_no_transform = view.copy()
view = np.array(glm.mat4x4(glm.mat3x3(glm.mat4x4(view_no_transform.transpose()))).to_list())
#view matrix to shader
view_loc = glGetUniformLocation(self.shader, "view")
glUniformMatrix4fv(view_loc, 1, GL_FALSE, view)
glDrawArrays(GL_TRIANGLES, 0, int(len(self.vertices) / 3))
| 3,963 | Python | .py | 118 | 23.881356 | 103 | 0.517287 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,319 | FPSCounter.py | AlexSanfilippo_ProceduralMeshGeneration/FPSCounter.py | import glfw
class FPSCounter:
"""
calculates frames per second and seconds per frame
"""
def __init__(self, frame_interval=100.0, spf_interval=10.0, mute=True):
self.frame_interval = frame_interval
self.refresh_time = True
self.st = glfw.get_time()
self.st_spf = glfw.get_time()
self.frame_count = 0
self.seconds_per_frame = 1.0/260.0
self.spf_interval = spf_interval
self.mute = mute
self.fps = 10000.0
def update(self):
if self.refresh_time:
self.st = glfw.get_time()
self.refresh_time = False
self.frame_count += 1
if self.frame_count == self.frame_interval:
self.refresh_time = True
et = glfw.get_time()
self.fps = self.frame_interval / (et - self.st)
if not self.mute:
print(self.fps, " FPS")
self.frame_count = 0
return self.update_spf()
def update_spf(self):
if self.frame_count % self.spf_interval == 0:
et = glfw.get_time()
self.seconds_per_frame = (et - self.st_spf) / (self.spf_interval)
self.st_spf = glfw.get_time()
return self.seconds_per_frame
def get_fps(self):
return self.fps
| 1,292 | Python | .py | 36 | 26.611111 | 77 | 0.5704 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,320 | ProceduralMesh.py | AlexSanfilippo_ProceduralMeshGeneration/ProceduralMesh.py | """
1st December, 2023
Classes for creation of basic proc-gen shapes. Shapes to create will be:
-[V]Triangle
-[V]Quad
-[V]Cube
-[V]Dynamic Size on cube using Dimensions [x,y,z]
-[V]Proper control over scaling, rotation, and transposition
-goal:rotating cube
-[V]hard-code normals on cube
-[V]refactor
-put draw, init, etc. into PrimativeMesh class
-new functions to set vertices and indices
-these will be overridden in the children
-----Stop here for proc jam, continue later
-[V]Proper normal vector calculation
-grab/rewrite my old code from primatives.h
-properly lit rotating cube
-[V]N-Sided Polygon
-[V]N-sided Prism
-[]N-sided pyramid
-[]Sphere
-[]Ellipsoid
-[]Torus
-for now, need some basic shapes for procjam 2023. Will start with triangle, quad, and cube
-for all meshes, we will assume they have
-position, texture coordinate, and normal vector as vertex info
-specular and diffuse textures
-use indexed drawing (indices and vertices)
"""
import gc
import glfw
import glm
from OpenGL.GL import *
from OpenGL.GL.shaders import compileProgram, compileShader
from TextureLoader import load_texture
import pyrr
import numpy as np
import math
from math import sin, cos, pi, sqrt, floor, ceil
from random import random
import random as r
vertex_src = """
# version 330
layout(location = 0) in vec3 a_position;
layout(location = 1) in vec2 a_texture;
layout(location = 2) in vec3 a_normal;
uniform mat4 model;
uniform mat4 projection;
uniform mat4 view;
out vec2 tex_coords; //texture coordinates
out vec3 normal;
out vec3 frag_pos;
void main()
{
tex_coords = a_texture;
frag_pos = vec3(model * vec4(a_position, 1.0));
normal = mat3(transpose(inverse(model))) * a_normal;
gl_Position = projection * view * vec4(frag_pos, 1.0);
}
"""
fragment_src = """
#version 330 core
out vec4 frag_color;
struct Material {
sampler2D diffuse;
sampler2D specular;
float shininess;
};
struct DirLight {
vec3 direction;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
struct PointLight {
vec3 position;
float constant;
float linear;
float quadratic;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
struct SpotLight {
vec3 position;
vec3 direction;
float cut_off;
float outer_cut_off;
float constant;
float linear;
float quadratic;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
#define NR_POINT_LIGHTS 1
in vec3 frag_pos;
in vec3 normal;
in vec2 tex_coords;
uniform vec3 view_pos;
uniform DirLight dir_light;
uniform PointLight point_lights[NR_POINT_LIGHTS];
uniform SpotLight spot_light;
uniform Material material;
// function prototypes
vec3 CalcDirLight(DirLight light, vec3 normal, vec3 view_dir);
vec3 CalcPointLight(PointLight light, vec3 normal, vec3 frag_pos, vec3 view_dir);
vec3 CalcSpotLight(SpotLight light, vec3 normal, vec3 frag_pos, vec3 view_dir);
void main()
{
// properties
vec3 norm = normalize(normal);
vec3 view_dir = normalize(view_pos - frag_pos);
// == =====================================================
// Our lighting is set up in 3 phases: directional, point lights and an optional flashlight
// For each phase, a calculate function is defined that calculates the corresponding color
// per lamp. In the main() function we take all the calculated colors and sum them up for
// this fragment's final color.
// == =====================================================
vec3 result;
// phase 1: directional lighting
result = CalcDirLight(dir_light, norm, view_dir);
// phase 2: point lights
for(int i = 0; i < NR_POINT_LIGHTS; i++)
result += CalcPointLight(point_lights[i], norm, frag_pos, view_dir);
// phase 3: spot light
//result += CalcSpotLight(spot_light, norm, frag_pos, view_dir);
frag_color = vec4(result, 1.0);
}
// calculates the color when using a directional light.
vec3 CalcDirLight(DirLight light, vec3 normal, vec3 view_dir)
{
vec3 light_dir = normalize(-light.direction);
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
return (ambient + diffuse + specular);
}
// calculates the color when using a point light.
vec3 CalcPointLight(PointLight light, vec3 normal, vec3 frag_pos, vec3 view_dir)
{
vec3 light_dir = normalize(light.position - frag_pos);
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// attenuation
float distance = length(light.position - frag_pos);
float attenuation = 1.0 / (light.constant + light.linear * distance + light.quadratic * (distance * distance));
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
ambient *= attenuation;
diffuse *= attenuation;
specular *= attenuation;
return (ambient + diffuse + specular);
}
// calculates the color when using a spot light.
vec3 CalcSpotLight(SpotLight light, vec3 normal, vec3 frag_pos, vec3 view_dir)
{
vec3 light_dir = normalize(light.position - frag_pos);
//view_dir = vec3(0.0, -1.0, 0.0); //tp
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// attenuation
float distance = length(light.position - frag_pos);
float attenuation = 1.0 / (light.constant + light.linear * distance + light.quadratic * (distance * distance));
// spotlight intensity
float theta = dot(light_dir, normalize(-light.direction));
float epsilon = light.cut_off - light.outer_cut_off;
float intensity = clamp((theta - light.outer_cut_off) / epsilon, 0.0, 1.0);
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
ambient *= attenuation * intensity;
diffuse *= attenuation * intensity;
specular *= attenuation * intensity;
return (ambient + diffuse + specular);
}
"""
def serialize_faces(func):
"""
Convert vertices of our model from our abstract Faces format to serialized data (a Numpy array) that
OpenGL can draw.
:return: numpy array of vertices (position, texture, and normals)
"""
def inner(*args, **kwargs):
faces = func(*args, **kwargs)
vertices = []
for face in faces:
vertices += face.face_to_list()
faces = np.array(
vertices,
dtype=np.float32
).flatten()
return faces
return inner
class PrimativeMesh:
"""
abstract parent of all primative meshes
-draws quad by default
"""
def __init__(self,
shader,
# material properties
diffuse,
specular,
shininess=32.0,
# mesh properties
dimensions=(5.0, 5.0),
position=(0.0, 0.0, 0.0),
rotation_magnitude=(0, 0, 0),
rotation_axis=(0.0, 0.0, 1.0),
scale=(1.0, 1.0, 1.0),
):
self.model_loc = None
self.VAO = None
self.VBO = None
self.shader = shader
self.diffuse = diffuse
self.specular = specular
self.shininess = shininess
self.position = glm.vec3(position)
self.rotation_magnitude = glm.vec3(rotation_magnitude)
self.rotation_axis = glm.vec3(rotation_axis)
self.scale = glm.vec3(scale)
self.dimensions = dimensions
self.vertices = self.generate_vertices()
self.buffer_setup()
def buffer_setup(self):
# quad VAO
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# quad position vertices (vertex attribute)
glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
# quad texture coords
glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(12))
glEnableVertexAttribArray(1)
# quad normals
glVertexAttribPointer(2, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(20))
glEnableVertexAttribArray(2)
self.model_loc = glGetUniformLocation(self.shader, "model")
def clean_up(self):
"""
Trying to resolve memory leaks
:return:
"""
glDeleteVertexArrays(1, [self.VAO])
glDeleteBuffers(1, [self.VBO])
def generate_vertices(self):
return np.array([
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 0.0, 1.0, 1.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 0.0, 0.0, 1.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 0.0, 1.0, 1.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 0.0, 1.0, 0.0, 0.0, 0.0, 1.0],
dtype=np.float32
)
def draw(self, view):
glUseProgram(self.shader)
glBindVertexArray(self.VAO)
glActiveTexture(GL_TEXTURE0)
glBindTexture(GL_TEXTURE_2D, self.diffuse)
glActiveTexture(GL_TEXTURE1)
glBindTexture(GL_TEXTURE_2D, self.specular)
# rotate, translate, and scale
model = glm.mat4(1.0)
model = glm.translate(model, self.position)
# todo: this rotation smells fishy
model = glm.rotate(model, self.rotation_magnitude.x, self.rotation_axis)
model = glm.scale(model, self.scale)
glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, glm.value_ptr(model))
glDrawArrays(GL_TRIANGLES, 0, int(len(self.vertices) / 3))
def set_diffuse(self, diffuse):
self.diffuse = diffuse
def set_specular(self, specular):
self.specular = specular
def get_diffuse(self):
return self.diffuse
def get_specular(self):
return self.specular
def get_position(self):
return self.position
def set_position(self, position):
self.position = position
# self.translation = pyrr.matrix44.create_from_translation(pyrr.Vector3(self.position))
def set_rotation_magnitude(self, magnitude):
self.rotation_magnitude = magnitude
def rotate_over_time(self, speed=0.5, axis=1):
"""
my_model.rotate_over_time(speed=0.1)
:param speed: how fast to spin
:param axis: x,y,z == 0,1,2
:return: None
"""
self.rotation_magnitude[axis] = speed * glfw.get_time()
def extrude_from_other_face(self, other, direction=[0.0, 1.0, 0.0], distance=1.0, flip_base=True, radius=1.0):
"""
just copy outer vertices, then run outer_Vertices_to_vertices()
Very different from (and likely outmoding) Face's method of same name
:param other: starting face
:param direction: normal vector to extrude in direcion of
:param distance: multiply direction by for magnitude of extrude
:param flip_base: turn starting face around
:param radius: (only if position == world origin) scale face
:return: a new face in a list
"""
face_extruded = Face()
face_extruded.outer_vertices = other.outer_vertices.copy()
face_extruded.calculate_sides()
face_extruded.vertices = self.outer_vertices_to_vertices(face=face_extruded,
number_of_sides=face_extruded.sides,
reverse_texture_coords=True)
face_extruded.normal = face_extruded.calculate_normal()
if flip_base:
pass
return face_extruded
def outer_vertices_to_vertices(self, face, number_of_sides, reverse_texture_coords=False):
"""
Generates the real vertices (with normals, texture coords) of a face
:param reverse_texture_coords: reverse order of list of position vertices
:param face: the face we want to generate vertices for. must have its outer vertices already
:param number_of_sides: eg, 4 for square, 5 for pentagon
:return: the vertices to be drawn by opengl
"""
# texture_coordinates = face.generate_texture_coordinates_polygon()
texture_coordinates = face.generate_texture_coordinates_polygon()
if reverse_texture_coords:
texture_coordinates.reverse()
vertices = []
texture_index = 1
for i in range(int(number_of_sides) - 2):
triangle_vertices = [Vertex(), Vertex(), Vertex()]
triangle_vertices[0].positions = glm.vec3(
[
face.outer_vertices[0],
face.outer_vertices[1],
face.outer_vertices[2]
]
)
triangle_vertices[0].texture_coordinates = glm.vec2(
[
texture_coordinates[0].x,
texture_coordinates[0].y
]
)
triangle_vertices[0].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[1].positions = glm.vec3([
face.outer_vertices[3 * i + 0 + 3 * 1],
face.outer_vertices[3 * i + 1 + 3 * 1],
face.outer_vertices[3 * i + 2 + 3 * 1]
])
triangle_vertices[1].texture_coordinates = glm.vec2([
texture_coordinates[texture_index].x,
texture_coordinates[texture_index].y
])
triangle_vertices[1].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[2].positions = glm.vec3([
face.outer_vertices[3 * i + 0 + 3 * 2],
face.outer_vertices[3 * i + 1 + 3 * 2],
face.outer_vertices[3 * i + 2 + 3 * 2]
])
triangle_vertices[2].texture_coordinates = glm.vec2([
texture_coordinates[texture_index + 1].x,
texture_coordinates[texture_index + 1].y
])
triangle_vertices[2].normals = glm.vec3([0.0, 1.0, 0.0])
for vertex in triangle_vertices:
vertex.normals = calculate_normal([
triangle_vertices[0].positions,
triangle_vertices[1].positions,
triangle_vertices[2].positions
])
vertices += triangle_vertices
texture_index += 1
return vertices
def _generate_polygonal_face(
self,
number_of_sides,
radius=1.0,
transform_x=1.0,
transform_z=1.0,
offset=(0.0, -1.0, 0.0)
):
face = Face()
face.outer_vertices = self.generate_outer_vertices(
number_of_sides=number_of_sides,
initial_angle=pi / number_of_sides,
radius=radius,
transform_x=transform_x,
transform_z=transform_z,
)
face.offset_outer_vertices(offset=glm.vec3(offset))
face.vertices = face.outer_vertices_to_vertices(reverse_texture_coordinates=True)
return face
def _generate_irregular_polygonal_face(
self,
number_of_sides,
radii=1.0,
transform_x=1.0,
transform_z=1.0,
offset=(0.0, -1.0, 0.0)
):
face = Face()
face.outer_vertices = self.generate_outer_vertices(
number_of_sides=number_of_sides,
initial_angle=pi / number_of_sides,
radius=radii,
transform_x=transform_x,
transform_z=transform_z,
)
face.offset_outer_vertices(offset=glm.vec3(offset))
face.vertices = face.outer_vertices_to_vertices(reverse_texture_coordinates=True)
return face
def generate_outer_vertices(
self,
number_of_sides=4,
initial_angle=None,
radius=1.0,
transform_x=1.0,
transform_z=1.0,
):
"""
generates the vertices of a simple polygon-no repeats
used for excuding and stitching, not for drawing
"""
outer_vertices = []
angle_of_rotation = 2 * pi / number_of_sides
if initial_angle == None:
initial_angle = pi / number_of_sides
if type(radius) is list:
for i in range(1, int(number_of_sides) + 1):
x_coord = cos(initial_angle + angle_of_rotation * -i) * radius[i - 1] * transform_x
z_coord = sin(initial_angle + angle_of_rotation * -i) * radius[i - 1] * transform_z
outer_vertices += [x_coord, 0.0, z_coord]
else:
for i in range(1, int(number_of_sides) + 1):
x_coord = cos(initial_angle + angle_of_rotation * -i) * radius * transform_x
z_coord = sin(initial_angle + angle_of_rotation * -i) * radius * transform_z
outer_vertices += [x_coord, 0.0, z_coord]
return outer_vertices
def bevel_cut(self, original_face, bevel_depths=[-0.2, -0.4], border_sizes=[0.6, 0.0], depth=2, direction=False):
"""
Cuts an n-sided hole, offsets a new face, and stitches to create a connected-depression
:bevel_depths: how much to push the excude in/out (negative for in)
:border_sizes: elm of (0,1), 0 means no border, aka, cut is same size as original polygonal face
:return: list of new faces. new normal-facing face (bottom face) is at index [:-(nsides + 1)]
TODO:
-Refactoring for neatness/clarity
-wrong texture coords for irregular convex polygons
"""
if depth == 0:
return [original_face]
original_face.calculate_radius()
radius = original_face.radius
updated_faces = []
test_cut_face = Face()
original_outer_vertices = list_to_vec3_list(original_face.outer_vertices)
for i in range(depth):
# 1. cut a similar polygon from the starting face
cut_outer_vertices = []
for point in original_outer_vertices:
mean_outer_point = calculate_mean_point(original_outer_vertices)
border_direction = glm.normalize(mean_outer_point - point)
border_diagonal_length = radius * border_sizes[i]
cut_point = point + border_direction * border_diagonal_length
cut_outer_vertices.append(cut_point)
# 2. build new face uses new outer vertices
test_cut_face.outer_vertices = vec3_list_to_list(cut_outer_vertices)
test_cut_face.calculate_sides()
test_cut_face.calculate_outer_vertices_as_vec3()
if direction == False:
face_normal = calculate_normal([
test_cut_face.outer_vertices_vec3[0],
test_cut_face.outer_vertices_vec3[1],
test_cut_face.outer_vertices_vec3[2]]
)
else:
face_normal = direction
test_cut_face.offset_outer_vertices(offset=face_normal * bevel_depths[i])
test_cut_face.vertices = test_cut_face.outer_vertices_to_vertices(
reverse_texture_coordinates=True,
regular=False,
)
if i == depth - 1:
updated_faces.append(test_cut_face)
# 3. stitch old faces to new face
updated_faces += self.stitch_faces(original_face, test_cut_face, test_cut_face.sides)
# 4. new face becomes "original" in preperation for next loop
original_face.outer_vertices = test_cut_face.outer_vertices
original_face.vertices = test_cut_face.vertices
original_face.calculate_radius()
radius = original_face.radius
original_outer_vertices = list_to_vec3_list(test_cut_face.outer_vertices.copy())
return updated_faces
def bevel_polygon_corner(self, face, subject_vertex_index=0, bevel_ratio=0.5):
"""
Takes in a face, transforms outer vertices of polygon by beveling a corner
-use case: base polygon before excude
:subject_vertex_index: index of outer vertex to bevel
:bevel_ratio: domain [0,1].
:return: the modified face
"""
older_outer_vertices = list_to_vec3_list(face.outer_vertices.copy())
subject_vertex = older_outer_vertices[subject_vertex_index]
number_of_sides = len(older_outer_vertices)
forward_vertex = older_outer_vertices[get_next_index(number_of_sides, subject_vertex_index)]
backward_vertex = older_outer_vertices[get_previous_index(number_of_sides, subject_vertex_index)]
forward_new_point = subject_vertex - bevel_ratio * (subject_vertex - forward_vertex)
backward_new_point = subject_vertex - bevel_ratio * (subject_vertex - backward_vertex)
updated_outer_vertices = older_outer_vertices.copy()
del updated_outer_vertices[subject_vertex_index]
updated_outer_vertices.insert(subject_vertex_index, forward_new_point)
updated_outer_vertices.insert(subject_vertex_index, backward_new_point)
face.outer_vertices = vec3_list_to_list(updated_outer_vertices)
self.sides += 1
return face
def pyrimidize_face(self, original_face, center_point_offset=0.0):
"""
Changes how an N-sided face is divided into triangles. Using a center point, and two corner points,
N triangles are created for the faces.
:param original_face: face to transform into N new faces
:param center_point_offset:
:return: N faces
"""
# get normal of the face
original_face.calculate_outer_vertices_as_vec3()
original_face.normal = original_face.calculate_normal()
original_normal = original_face.normal
# calculate center of original face
original_outer_vertices_as_vec3s = original_face.outer_vertices_vec3
center_point = calculate_mean_point(original_outer_vertices_as_vec3s)
# adjust center by the normal and center_point_offset
center_point += original_normal * center_point_offset
# construct the N new triangles outer vertices
original_face.calculate_sides()
num_sides = original_face.sides
original_outer_vertices_as_vec3s = original_face.outer_vertices_vec3
new_triangles_outer_vertices = []
for index, original_outer_vertex in enumerate(original_outer_vertices_as_vec3s):
triangle_vertices = []
next_index = index + 1
if index == num_sides - 1:
next_index = 0
triangle_vertices += [original_outer_vertices_as_vec3s[next_index], center_point, original_outer_vertex]
new_triangles_outer_vertices.append(triangle_vertices)
# create N new faces (using outer vertices)
triangle_faces = []
for index in range(num_sides):
triangle_faces.append(Face())
# -convert vec3 outers into list outers
triangle_faces[index].outer_vertices = vec3_list_to_list(new_triangles_outer_vertices[index])
triangle_faces[index].outer_vertices_vec3 = new_triangles_outer_vertices[index]
# generate vertices of N new faces.
# triangle_faces[index].vertices = triangle_faces[index].outer_vertices
triangle_faces[index].vertices = triangle_faces[index].outer_vertices_to_vertices()
triangle_faces[index].apply_hardset_triangle_texture_coords()
# return the N new quad faces
return triangle_faces
# make n new triangular faces
def divide_face_into_quads(self, original_face, center_point_offset=0.0):
"""
subdivide a face into quads. each new quad has as its vertices 2 midpoints, 1 corner, and the center of
the original face. Only works on quad faces?
:param original_face: a face object
:param center_point_offset: move center along face normal
:return: 4 new faces
"""
# get normal of the face
original_face.calculate_outer_vertices_as_vec3()
original_face.normal = original_face.calculate_normal()
original_normal = original_face.normal
# calculate center of original face
original_outer_vertices_as_vec3s = original_face.outer_vertices_vec3
center_point = calculate_mean_point(original_outer_vertices_as_vec3s)
# adjust center by the normal and center_point_offset
center_point += original_normal * center_point_offset
# find the N midpoints (N = number of sides of face polygon)
original_face.calculate_sides()
num_sides = original_face.sides
midpoints = [glm.vec3(0.0, 0.0, 0.0)] * num_sides
for index in range(num_sides):
overflowing_index = 1
if index == num_sides - 1:
overflowing_index = -index
midpoints[index] = calculate_mean_point(
[original_outer_vertices_as_vec3s[index],
original_outer_vertices_as_vec3s[index + overflowing_index]
]
)
# construct the N quads outer vertices.
all_quads_outer_vertices = []
for index, outer_point in enumerate(original_outer_vertices_as_vec3s):
current_quad_outer_vertices = []
if index > 0:
backwards_midpoint_index = index - 1
else:
backwards_midpoint_index = num_sides - 1
current_quad_outer_vertices.append(outer_point)
current_quad_outer_vertices.append(midpoints[index])
current_quad_outer_vertices.append(center_point)
current_quad_outer_vertices.append(midpoints[backwards_midpoint_index])
current_quad_outer_vertices = rotate_list(current_list=current_quad_outer_vertices, steps=-index)
all_quads_outer_vertices.append(current_quad_outer_vertices)
# create N new faces (using outer vertices)
quad_faces = []
for index in range(num_sides):
quad_faces.append(Face())
# -convert vec3 outers into list outers
quad_faces[index].outer_vertices = vec3_list_to_list(all_quads_outer_vertices[index])
quad_faces[index].outer_vertices_vec3 = all_quads_outer_vertices[index]
# generate vertices of N new faces.
quad_faces[index].vertices = quad_faces[index].outer_vertices_to_vertices()
# (option) fix texture coords of N new faces
# quad_faces[index].apply_hardset_quad_texture_coords(sides=4)
# return the N new quad faces
return quad_faces
def bezier_cut(self, face, intervals=0, offset=0):
"""
Constructs a bevel curve from the outer vertices of a quad-face, and uses the curve to cut a hole
in the face, creating two new faces
:param face:
:param iterations:
:return: list of faces
"""
face.calculate_outer_vertices_as_vec3()
original_face_normal = face.calculate_normal()
from collections import deque
d = deque(face.outer_vertices_vec3)
d.rotate(2)
d.reverse()
control_points = list(d)
bezier_points = bezier_cubic(points=control_points, intervals=intervals)
vertices_border = []
outer_vertices_cut = []
even_odd_index = 0
if intervals % 2 == 0:
even_odd_index = 1
# first triangle
vertices_border.append(face.outer_vertices_vec3[0])
vertices_border.append(face.outer_vertices_vec3[1])
vertices_border.append(bezier_points[1])
midpoint = len(bezier_points) / 2
midpoint_indices = [math.floor(midpoint), math.ceil(midpoint)]
# for index, bezier_point in enumerate(bezier_points[1:midpoint_indices[1]+1]):
for index, bezier_point in enumerate(bezier_points[1:midpoint_indices[even_odd_index]]):
vertices_border.append(bezier_point)
vertices_border.append(bezier_points[index + 2])
vertices_border.append(face.outer_vertices_vec3[0])
# keystone_triangle (even vertices case)
vertices_border.append(bezier_points[midpoint_indices[even_odd_index]])
vertices_border.append(face.outer_vertices_vec3[3])
vertices_border.append(face.outer_vertices_vec3[0])
for index, bezier_point in enumerate(bezier_points[midpoint_indices[even_odd_index]:-1]):
vertices_border.append(bezier_point)
vertices_border.append(bezier_points[index + midpoint_indices[even_odd_index] + 1])
vertices_border.append(face.outer_vertices_vec3[3])
proper_vertices = []
for point in vertices_border:
current_vertex = Vertex()
current_vertex.positions = point
current_vertex.texture_coordinates = calculate_texture_coordinates_from_control_points(
point=point,
control_points=control_points
)
current_vertex.normals = original_face_normal
proper_vertices.append(current_vertex)
face.vertices = proper_vertices
"""Now do the cut's vertices"""
# cut saved into a new face
face_cut = Face()
face_cut.outer_vertices = vec3_list_to_list([point - original_face_normal * offset for point in bezier_points])
midpoint_base = calculate_mean_point([bezier_points[0], bezier_points[intervals + 1]])
vertices_cut = []
for index, bezier_point in enumerate(bezier_points[:intervals + 1]):
vertices_cut.append(midpoint_base - original_face_normal * offset)
vertices_cut.append(bezier_points[index + 1] - original_face_normal * offset)
vertices_cut.append(bezier_point - original_face_normal * offset)
vertices_cut_proper = []
for point in vertices_cut:
current_vertex = Vertex()
current_vertex.positions = point
current_vertex.texture_coordinates = calculate_texture_coordinates_from_control_points(
point=point,
control_points=control_points
)
current_vertex.normals = original_face_normal
vertices_cut_proper.append(current_vertex)
face_cut.vertices = vertices_cut_proper
faces_inner_arch = []
if offset != 0:
face_cut_pre_offset = Face()
face_cut_pre_offset.outer_vertices = vec3_list_to_list(bezier_points)
return [face, face_cut] + self.stitch_faces(face_cut, face_cut_pre_offset, intervals + 2)
return [face, face_cut]
def stitch_faces(self, face_a, face_b, number_of_faces):
"""
takes two (geometrically same) Faces and forms a prism-like mesh by connecting them
with quadrilaterals
:returns: a list of faces
"""
faces = []
for side in range(int(number_of_faces)):
face = Face()
if side == number_of_faces - 1:
face.outer_vertices = [
face_b.outer_vertices[side * 3 + 0], face_b.outer_vertices[side * 3 + 1],
face_b.outer_vertices[side * 3 + 2],
face_a.outer_vertices[side * 3 + 0], face_a.outer_vertices[side * 3 + 1],
face_a.outer_vertices[side * 3 + 2],
face_a.outer_vertices[0 * 3 + 0], face_a.outer_vertices[0 * 3 + 1],
face_a.outer_vertices[0 * 3 + 2],
face_b.outer_vertices[0 * 3 + 0], face_b.outer_vertices[0 * 3 + 1],
face_b.outer_vertices[0 * 3 + 2],
]
else:
face.outer_vertices = [
face_b.outer_vertices[side * 3 + 0], face_b.outer_vertices[side * 3 + 1],
face_b.outer_vertices[side * 3 + 2],
face_a.outer_vertices[side * 3 + 0], face_a.outer_vertices[side * 3 + 1],
face_a.outer_vertices[side * 3 + 2],
face_a.outer_vertices[side * 3 + 3], face_a.outer_vertices[side * 3 + 4],
face_a.outer_vertices[side * 3 + 5],
face_b.outer_vertices[side * 3 + 3], face_b.outer_vertices[side * 3 + 4],
face_b.outer_vertices[side * 3 + 5],
]
# face.vertices = self.outer_vertices_to_vertices(face, 4)
face.vertices = face.outer_vertices_to_vertices(reverse_texture_coordinates=True)
faces.append(face)
return faces
def subdivide_hip_roof(self, face, start_point=[0.5, 0.2], end_point=[0.5, 0.8], vertical_offset=0.0):
"""
divides (quad) face into 2 quads and 2 triangle faces, like the triangular "hip" roof of a house
todo:fix texture coords
:return: list of faces: two square, two triangular
"""
original_outer_vertices = face.outer_vertices.copy()
operatable_outer_vertices = original_outer_vertices.copy()
operatable_outer_vertices_vec3 = list_to_vec3_list(operatable_outer_vertices)
# make sure original face has all required attributes
face.calculate_outer_vertices_as_vec3()
face.normal = face.calculate_normal()
original_normal = face.normal
original_outer_vertices_as_vec3s = face.outer_vertices_vec3
face.calculate_sides()
# calculate the start and end points of the center line
center_line_start_point = (operatable_outer_vertices_vec3[1]
+ start_point[0] * (
operatable_outer_vertices_vec3[0] - operatable_outer_vertices_vec3[1])
+ start_point[1] * (
operatable_outer_vertices_vec3[2] - operatable_outer_vertices_vec3[1])
+ vertical_offset * face.normal)
center_line_end_point = (operatable_outer_vertices_vec3[1]
+ end_point[0] * (
operatable_outer_vertices_vec3[0] - operatable_outer_vertices_vec3[1])
+ end_point[1] * (
operatable_outer_vertices_vec3[2] - operatable_outer_vertices_vec3[1])
+ vertical_offset * face.normal)
# generate outer vertices of new quad faces
roof_faces = []
quad_base_points = [
[0, 3, center_line_start_point, center_line_end_point],
[2, 1, center_line_end_point, center_line_start_point]
]
for base_points in quad_base_points:
current_outer_vertices = []
current_outer_vertices.append(original_outer_vertices_as_vec3s[base_points[0]])
current_outer_vertices.append(base_points[2])
current_outer_vertices.append(base_points[3])
current_outer_vertices.append(original_outer_vertices_as_vec3s[base_points[1]])
new_face = Face()
new_face.outer_vertices_vec3 = current_outer_vertices
new_face.outer_vertices = vec3_list_to_list(current_outer_vertices)
new_face.vertices = new_face.outer_vertices_to_vertices()
new_face.apply_hardset_quad_texture_coords()
roof_faces.append(new_face)
triangle_base_points = [[1, 0, center_line_start_point], [3, 2, center_line_end_point]]
for i in range(2):
current_outer_vertices = []
current_outer_vertices.append(original_outer_vertices_as_vec3s[triangle_base_points[i][0]])
current_outer_vertices.append(triangle_base_points[i][2])
current_outer_vertices.append(original_outer_vertices_as_vec3s[triangle_base_points[i][1]])
new_face = Face()
new_face.outer_vertices_vec3 = current_outer_vertices
new_face.outer_vertices = vec3_list_to_list(current_outer_vertices)
new_face.vertices = new_face.outer_vertices_to_vertices()
# todo:fix these triangle faces outer vertices
new_face.apply_hardset_triangle_texture_coords()
roof_faces.append(new_face)
return roof_faces
def subdivide_quad_lengthwise(self, face, subdivisions=2, widthwise=False):
"""
Takes in a quadrillateral face and divides it into subfaces along the length of the face
:return: list of faces
todo: the widthwise option rotates and distorts the texture
-need a way to do widthwise that does not rely on rotating the list naively
"""
original_outer_vertices = face.outer_vertices.copy()
operatable_outer_vertices = original_outer_vertices.copy()
if widthwise:
operatable_outer_vertices = rotate_list(current_list=operatable_outer_vertices, steps=3)
# todo: rotate back at the end -necc?
if subdivisions in (0, 1):
return [face]
updated_faces = []
# make sure original face has all required attributes
face.calculate_outer_vertices_as_vec3()
face.normal = face.calculate_normal()
original_normal = face.normal
original_outer_vertices_as_vec3s = face.outer_vertices_vec3
face.calculate_sides()
# find the N-subdivisions points along the length sides
operatable_outer_vertices_vec3 = list_to_vec3_list(operatable_outer_vertices)
side_top = [operatable_outer_vertices_vec3[0], operatable_outer_vertices_vec3[3]]
side_bottom = [operatable_outer_vertices_vec3[1], operatable_outer_vertices_vec3[2]]
outer_vertices_top = bezier_linear(points=side_top, intervals=subdivisions)
outer_vertices_bottom = bezier_linear(points=side_bottom, intervals=subdivisions)
# generate outer vertices of new faces
subdivision_faces = []
subdivision_index = 0
while len(subdivision_faces) < subdivisions:
subdivision_face = Face()
subdivision_outer_vertices = [
outer_vertices_top[subdivision_index],
outer_vertices_bottom[subdivision_index],
outer_vertices_bottom[subdivision_index + 1],
outer_vertices_top[subdivision_index + 1]
]
subdivision_outer_vertices = rotate_list(current_list=subdivision_outer_vertices, steps=-1)
subdivision_face.outer_vertices_vec3 = subdivision_outer_vertices
subdivision_face.outer_vertices = vec3_list_to_list(subdivision_outer_vertices)
subdivision_face.vertices = subdivision_face.outer_vertices_to_vertices()
subdivision_face.normal = face.normal
subdivision_faces.append(subdivision_face)
subdivision_index += 1
updated_faces += subdivision_faces
return updated_faces
class PrimativeMeshEmission(PrimativeMesh):
"""
Version of primative mesh that redefines draw() method to allow for
emission texture.
"""
def __init__(
self,
shader,
# material properties
diffuse,
specular,
emission,
shininess=32.0,
# mesh properties
dimensions=[5.0, 5.0],
position=[0.0, 0.0, 0.0],
rotation_magnitude=(0.0, 0.0, 0.0),
rotation_axis=(0.0, 0.0, 1.0),
scale=(1.0, 1.0, 1.0),
):
super().__init__(
shader=shader,
# material properties
diffuse=diffuse,
specular=specular,
shininess=shininess,
# mesh properties
dimensions=dimensions,
position=position,
rotation_magnitude=glm.vec3(rotation_magnitude),
rotation_axis=glm.vec3(rotation_axis),
scale=glm.vec3(scale),
)
self.emission = emission
def clean_up(self):
"""
Trying to resolve memory leaks
:return:
"""
glDeleteVertexArrays(1, [self.VAO])
glDeleteBuffers(1, [self.VBO])
del self.vertices
del self.VAO
del self.VBO
gc.collect()
def draw(self, view):
glUseProgram(self.shader)
glBindVertexArray(self.VAO)
glActiveTexture(GL_TEXTURE0)
glBindTexture(GL_TEXTURE_2D, self.diffuse)
glActiveTexture(GL_TEXTURE1)
glBindTexture(GL_TEXTURE_2D, self.specular)
glActiveTexture(GL_TEXTURE2)
glBindTexture(GL_TEXTURE_2D, self.emission)
# rotate, translate, and scale
model = glm.mat4(1.0)
model = glm.translate(model, glm.vec3(self.position))
model = glm.rotate(model, self.rotation_magnitude.x, glm.vec3([1.0, 0.0, 0.0]))
model = glm.rotate(model, self.rotation_magnitude.y, glm.vec3([0.0, 1.0, 0.0]))
model = glm.rotate(model, self.rotation_magnitude.z, glm.vec3([0.0, 0.0, 1.0]))
model = glm.scale(model, self.scale)
glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, glm.value_ptr(model))
glDrawArrays(GL_TRIANGLES, 0, int(len(self.vertices) / 3))
def set_diffuse(self, diffuse):
self.diffuse = diffuse
def set_specular(self, specular):
self.specular = specular
def set_emission(self, emission):
self.emission = emission
class Spaceship(PrimativeMeshEmission):
cardinal_directions = {
'down': [1.0, 0.0, 0.0],
'up': [1.0, 0.0, 0.0],
'port': [0.0, 0.0, 1.0],
'starboard': [0.0, 0.0, -1.0],
'forward': [0.0, 1.0, 0.0],
'backward': [0.0, -1.0, 0.0],
}
def __init__(
self,
shader,
# material properties
diffuse,
specular,
emission,
shininess=32.0,
# mesh properties
dimensions=[5.0, 5.0],
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
number_of_sides=4,
number_of_segments=4,
transform_x=1.0,
transform_z=1.0,
length_of_segment=1.0,
radius=1.0,
seed=1,
):
self.number_of_sides = number_of_sides
self.number_of_segments = number_of_segments
self.transform_x = transform_x
self.transform_z = transform_z
self.length_of_segment = length_of_segment
self.radius = radius
self.seed = seed
r.seed(self.seed)
super().__init__(
shader=shader,
# material properties
diffuse=diffuse,
specular=specular,
shininess=shininess,
emission=emission,
# mesh properties
dimensions=dimensions,
position=position,
rotation_magnitude=rotation_magnitude,
rotation_axis=rotation_axis,
scale=scale,
)
@property
def texture_atlas_size(self):
return 2
def get_radius_multipliers(self):
"""
Produces a list of radius multipliers to determine the ships overall shape
:return:
"""
RADIUS_MAX = self.radius * (0.1029 * self.number_of_segments) + self.radius
RADIUS_MIN = self.radius * (0.0294 * self.number_of_segments) + 0.5
control_point_count = self.number_of_segments
#set control point radii
control_points = []
control_points_relative = []
last_point = self.radius
for index in range(control_point_count):
if index > 0:
last_point = control_points[index-1]
control_points.append((random() * (RADIUS_MAX - RADIUS_MIN) + RADIUS_MIN))
control_points_relative.append(control_points[index] / last_point)
return control_points_relative
def export_as_obj(self, filename=None):
r.seed(self.seed)
faces = self.generate_faces()
vertices_obj = []
faces_obj = []
count_vertices_mesh = 0
if not filename:
filename = f'spaceship_{self.number_of_sides}_{self.seed}'
with open(f'{filename}.obj', 'w') as file:
file.write('#OBJ File generated by Spaceship Generator\n#Alexander Sanfilippo 2024\n')
file.write(f'g {filename}\n\n')
file.write(f'o spaceship\n')
for face in faces:
faces_current = []
counter = 0
for vertex in face.vertices:
count_vertices_mesh += 1
vertices_obj.append(vertex)
if counter == 2:
face_triangle = []
for i in list(reversed(range(3))):
face_triangle.append(count_vertices_mesh - i)
faces_current.append(face_triangle)
counter = 0
else:
counter += 1
for face_current in faces_current:
faces_obj.append(face_current)
for vertex in vertices_obj:
file.write(f'v {vertex.positions.x} {vertex.positions.y} {vertex.positions.z}\n')
# file.write('\n')
for vertex in vertices_obj:
file.write(f'vt {vertex.texture_coordinates.x} {vertex.texture_coordinates.y}\n')
# file.write('\n')
for vertex in vertices_obj:
file.write(f'vn {vertex.normals.x} {vertex.normals.y} {vertex.normals.z}\n')
# file.write('\n')
# write the faces to the file
for face in faces_obj:
file.write('f')
for vertex in face:
file.write(f' {vertex}/{vertex}/{vertex}')
file.write('\n')
@serialize_faces
def generate_vertices(self):
return self.generate_faces()
def generate_faces(self):
"""
Where the mesh is generated, vertex by vertex.
:return:
"""
faces = [
self._generate_polygonal_face(
number_of_sides=self.number_of_sides,
radius=self.radius,
transform_x=self.transform_x,
transform_z=self.transform_z,
)
]
faces[0].update_texture_coords_using_atlas_index(
texture_atlas_index=1,
texture_atlas_size=self.texture_atlas_size,
)
# generate the latter faces via extruding
segment_faces = []
radius_multipliers = self.get_radius_multipliers()
for i in range(self.number_of_segments):
face_extruded = Face()
radius_multiplier_current_segment = radius_multipliers[i]
face_extruded.extrude_from_other_face(
other=faces[i],
direction=list(faces[i].calculate_normal()),
distance=self.length_of_segment,
flip_base=True,
scale=radius_multiplier_current_segment
)
faces.append(face_extruded)
faces_from_stitch = self.stitch_faces(
face_a=faces[i],
face_b=faces[i + 1],
number_of_faces=self.number_of_sides
)
segment_faces += faces_from_stitch
faces[-1].update_texture_coords_using_atlas_index(
texture_atlas_index=3,
texture_atlas_size=self.texture_atlas_size,
)
faces += self.generate_thrusters(face=faces[-1])
faces += self.add_detail_to_faces(faces_to_alter=segment_faces)
faces += self.generate_nose(face=faces[0])
del faces[0]
return faces
def generate_thrusters(self, face):
thruster_faces = self.bevel_cut(
original_face=face,
bevel_depths=[self.length_of_segment * 0.25, 0.0, self.length_of_segment * -0.5],
border_sizes=[-0.25, 0.25, 0.5],
depth=3
)
# texturing
thruster_backend = thruster_faces.pop(-(self.number_of_sides + 1))
thruster_backend.update_texture_coords_using_atlas_index(
texture_atlas_size=self.texture_atlas_size,
texture_atlas_index=4,
)
for face in thruster_faces:
face.update_texture_coords_using_atlas_index(
texture_atlas_size=self.texture_atlas_size,
texture_atlas_index=2,
)
return thruster_faces + [thruster_backend]
def generate_nose(self, face):
faces_nose = []
face_nose_tip = Face()
face_nose_tip.extrude_from_other_face(
other=face,
direction=list(-face.calculate_normal()),
distance=self.length_of_segment * (0.25 + random() * 1.25),
scale=random() * .2 + .1,
)
faces_from_stitch = self.stitch_faces(
face_a=face_nose_tip,
face_b=face,
number_of_faces=self.number_of_sides
)
face_nose_tip.calculate_outer_vertices_as_vec3()
face_nose_tip.outer_vertices_vec3.reverse()
face_nose_tip.outer_vertices = vec3_list_to_list(face_nose_tip.outer_vertices_vec3)
face_nose_tip.vertices = face_nose_tip.outer_vertices_to_vertices(
reverse_texture_coordinates=True,
)
# texturing
# todo:reenable
face_nose_tip.update_texture_coords_using_atlas_index(
texture_atlas_size=self.texture_atlas_size,
texture_atlas_index=1,
)
for face in faces_from_stitch:
face.update_texture_coords_using_atlas_index(
texture_atlas_size=self.texture_atlas_size,
texture_atlas_index=2,
)
# return new faces
faces_nose += [face_nose_tip] + faces_from_stitch
return faces_nose
def determine_radius(
self,
i,
number_of_segments,
radius_multiplier,
minimum_radius_multipler,
maximum_radius_multipler,
growth=1.25,
shrinkage=0.5,
):
if i < floor(random() * 7):
radius_multiplier = min(growth * radius_multiplier, maximum_radius_multipler)
else:
if radius_multiplier > 1.0:
radius_multiplier = 1.0
radius_multiplier = max(shrinkage * radius_multiplier, minimum_radius_multipler)
return radius_multiplier
def add_detail_to_faces(self, faces_to_alter):
"""
Goes around faces of segmented cylindar and replaces plain quads with detailed meshes
:param faces_to_alter:
:return:
"""
if self.number_of_sides % 3 == 0:
symmetry_type = 'triangular'
elif self.number_of_sides % 2 == 0 and self.number_of_sides != 10:
symmetry_type = 'square'
else:
symmetry_type = 'irregular'
instructions_per_segment = []
for index_segment in range(self.number_of_segments):
if symmetry_type == 'square':
number_of_pairs = int(self.number_of_sides / 2)
instructions_per_pair = []
for i in range(number_of_pairs):
instructions_per_pair.append(random())
instructions_per_segment.append(instructions_per_pair)
elif symmetry_type == 'triangular':
number_of_triplets = int(self.number_of_sides / 3)
instructions_per_pair = []
for i in range(number_of_triplets):
instructions_per_pair.append(random())
instructions_per_segment.append(instructions_per_pair)
else:
instructions_per_segment.append((random(),) * self.number_of_sides)
faces_altered = []
faces_unaltered = []
for index_segment in range(self.number_of_segments):
faces_paired = {}
if symmetry_type == 'square':
number_of_pairs = int(self.number_of_sides / 2)
for index_pair in range(number_of_pairs):
if self.number_of_sides == 8:
if index_pair == 1 or index_pair == 3:
faces_paired[index_pair] = [
faces_to_alter[index_segment * self.number_of_sides + index_pair],
faces_to_alter[index_segment * self.number_of_sides + index_pair + number_of_pairs]
]
elif index_pair == 0:
faces_paired[index_pair] = [
faces_to_alter[index_segment * self.number_of_sides + index_pair],
faces_to_alter[index_segment * self.number_of_sides + 6]
]
else:
faces_paired[index_pair] = [
faces_to_alter[index_segment * self.number_of_sides + 4],
faces_to_alter[index_segment * self.number_of_sides + 2]
]
else:
faces_paired[index_pair] = [
faces_to_alter[index_segment * self.number_of_sides + index_pair],
faces_to_alter[index_segment * self.number_of_sides + index_pair + number_of_pairs]
]
faces_altered_local, faces_unaltered_local = self.detail_faces_by_instruction(
faces_altered=faces_altered,
index_segment=index_segment,
instructions_per_segment=instructions_per_segment,
faces_paired=faces_paired[index_pair],
index_instruction=index_pair,
)
faces_altered += faces_altered_local
faces_unaltered += faces_unaltered_local
if symmetry_type == 'triangular':
number_of_triplets = int(self.number_of_sides / 3)
for index_pair in range(number_of_triplets):
faces_paired[index_pair] = [
faces_to_alter[index_segment * self.number_of_sides + index_pair],
faces_to_alter[index_segment * self.number_of_sides + index_pair + number_of_triplets],
faces_to_alter[index_segment * self.number_of_sides + index_pair + number_of_triplets * 2],
]
faces_altered_local, faces_unaltered_local = self.detail_faces_by_instruction(
faces_altered=faces_altered,
index_segment=index_segment,
instructions_per_segment=instructions_per_segment,
faces_paired=faces_paired[index_pair],
index_instruction=index_pair,
)
faces_altered += faces_altered_local
faces_unaltered += faces_unaltered_local
if symmetry_type == 'irregular':
for index_side in range(self.number_of_sides):
faces_altered_local, faces_unaltered_local = self.detail_faces_by_instruction(
faces_altered=faces_altered,
index_segment=index_segment,
instructions_per_segment=instructions_per_segment,
faces_paired=[faces_to_alter[index_side + index_segment * self.number_of_sides]],
index_instruction=index_side,
)
faces_altered += faces_altered_local
faces_unaltered += faces_unaltered_local
# todo: texturing should happen within detail-applying-functions.
for face in faces_unaltered:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=0,
texture_atlas_size=self.texture_atlas_size,
)
return faces_altered + faces_unaltered
def add_detail_recursive_extrude(self, face, bevel_depths, border_sizes):
first_extrusion = self.extrude_and_stitch(face=face, scale=1.0, distance=floor(self.length_of_segment * 0.25))
faces_updated = []
for face_current in first_extrusion:
faces_one_bevel = self.bevel_cut(
original_face=face_current,
bevel_depths=bevel_depths[1:2],
border_sizes=border_sizes[1:2],
depth=1
)
faces_one_bevel[0].update_texture_coords_using_atlas_index(
texture_atlas_index=1,
texture_atlas_size=self.texture_atlas_size,
)
for face in faces_one_bevel:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=2,
texture_atlas_size=self.texture_atlas_size,
)
faces_updated += faces_one_bevel
return faces_updated
def add_detail_intake(self, face):
local_faces_altered = self.bevel_cut(
original_face=face,
depth=1,
bevel_depths=[self.length_of_segment * 0.2],
border_sizes=[0.0]
)
intake_index = 2
intake_faces = [local_faces_altered[intake_index]]
intake_faces_altered = []
for face in intake_faces:
intake_faces_altered += self.bevel_cut(
original_face=face,
depth=2,
bevel_depths=[0.0, -self.length_of_segment * 0.3],
border_sizes=[0.15, 0.5]
)
del local_faces_altered[intake_index]
for face in local_faces_altered:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=2,
texture_atlas_size=self.texture_atlas_size,
)
for face in intake_faces_altered:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=1,
texture_atlas_size=self.texture_atlas_size
)
intake_faces_altered[-5].update_texture_coords_using_atlas_index(
texture_atlas_index=3,
texture_atlas_size=self.texture_atlas_size
)
return local_faces_altered + intake_faces_altered
def detail_faces_by_instruction(
self,
faces_altered,
index_segment,
instructions_per_segment,
faces_paired,
index_instruction,
):
local_faces_altered = []
local_faces_unaltered = []
for face in faces_paired:
if instructions_per_segment[index_segment][index_instruction] < 0.25:
# pass
local_faces_altered += self.add_detail_cubbies(face)
elif instructions_per_segment[index_segment][index_instruction] < 0.5:
local_faces_altered += self.add_detail_pyrimide(face)
elif instructions_per_segment[index_segment][index_instruction] < 0.73:
local_faces_altered += self.add_detail_recursive_extrude(
face,
bevel_depths=[0.4, 0.33],
border_sizes=[0.0, 0.5]
)
for face in local_faces_altered:
if math.isnan(face.outer_vertices[0]):
print('outer vertex is nan!')
elif instructions_per_segment[index_segment][index_instruction] < 0.75:
local_faces_altered += self.add_detail_intake(
face,
)
elif instructions_per_segment[index_segment][index_instruction] < 0.9:
faces_special_bevel = self.bevel_cut(
original_face=face,
bevel_depths=[0.0, -0.5],
border_sizes=[0.25, 0.0],
depth=2,
)
faces_special_bevel[-5].update_texture_coords_using_atlas_index(texture_atlas_index=3,
texture_atlas_size=self.texture_atlas_size)
for face in faces_special_bevel[:-5]:
face.update_texture_coords_using_atlas_index(texture_atlas_index=2, texture_atlas_size=self.texture_atlas_size)
for face in faces_special_bevel[-4:]:
face.update_texture_coords_using_atlas_index(texture_atlas_index=2, texture_atlas_size=self.texture_atlas_size)
local_faces_altered += faces_special_bevel
else:
local_faces_unaltered += [face]
return local_faces_altered, local_faces_unaltered
def add_detail_pyrimide(self, face):
current_stitch_faces = []
faces_altered_local = []
current_stitch_faces += self.pyrimidize_face(original_face=face, center_point_offset=2.5)
faces_altered_local += current_stitch_faces
for face in faces_altered_local:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=2,
texture_atlas_size=self.texture_atlas_size
)
return faces_altered_local
def add_detail_cubbies(self, face):
current_stitch_faces = []
current_stitch_faces += self.subdivide_quad_lengthwise(face=face, subdivisions=2, widthwise=True)
faces_altered_local = []
for new_stitch_face in current_stitch_faces:
faces_current_cubby = self.bevel_cut(
original_face=new_stitch_face,
bevel_depths=[-0.33],
border_sizes=[0.5],
depth=1,
)
faces_current_cubby[0].update_texture_coords_using_atlas_index(
texture_atlas_index=3, # todo: change to 1
texture_atlas_size=self.texture_atlas_size
)
for face in faces_current_cubby[1:]:
face.update_texture_coords_using_atlas_index(
texture_atlas_index=2,
texture_atlas_size=self.texture_atlas_size,
)
faces_altered_local += faces_current_cubby
return faces_altered_local
def extrude_and_stitch(self, face, number_of_faces=4, distance=1, scale=1.0):
current_stitch_faces = []
face.calculate_outer_vertices_as_vec3()
face_extrude = Face()
face_normal = face.calculate_normal()
face_extrude.extrude_from_other_face(other=face, direction=face_normal, distance=distance, scale=scale)
current_stitch_faces += [face_extrude]
current_stitch_faces += self.stitch_faces(face_a=face, face_b=face_extrude, number_of_faces=number_of_faces)
return current_stitch_faces
class Spaceship3x3(Spaceship):
@property
def texture_atlas_size(self):
return 3
class Polygon(PrimativeMesh):
"""
Generates an N-sided Polygonal face
"""
def __init__(self,
shader,
diffuse,
specular,
shininess=32.0,
dimensions=[5.0, 5.0],
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
sides=3,
transform_x=1.0,
transform_z=1.0,
):
self.sides = sides
self.transform_x = transform_x
self.transform_z = transform_z
super().__init__(
shader=shader,
diffuse=diffuse,
specular=specular,
shininess=shininess,
dimensions=dimensions,
position=position,
rotation_magnitude=rotation_magnitude,
rotation_axis=rotation_axis,
scale=scale,
)
def generate_vertices(self):
"""
Generates the outer vertices and vertices of the triangle of this model
"""
face = self._generate_polygonal_face(
number_of_sides=self.sides,
transform_x=self.transform_x,
transform_z=self.transform_z,
)
vertices = face.face_to_list()
return np.array(
vertices,
dtype=np.float32
).flatten()
class PolygonIrregular(Polygon):
def __init__(
self,
shader,
diffuse,
specular,
shininess=32.0,
dimensions=[5.0, 5.0],
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
sides=3,
transform_x=1.0,
transform_z=1.0,
radii=1.0,
):
self.radii = radii
super().__init__(
shader=shader,
diffuse=diffuse,
specular=specular,
shininess=shininess,
dimensions=dimensions,
position=position,
rotation_magnitude=rotation_magnitude,
rotation_axis=rotation_axis,
scale=scale,
sides=sides,
transform_x=transform_x,
transform_z=transform_z,
)
def generate_vertices(self):
"""
Generates the outer vertices and vertices of the triangle of this model
"""
face = self._generate_irregular_polygonal_face(
number_of_sides=self.sides,
transform_x=self.transform_x,
transform_z=self.transform_z,
radii=self.radii,
)
vertices = face.face_to_list()
return np.array(
vertices,
dtype=np.float32
).flatten()
class QuadMesh(PrimativeMesh):
pass
class TriangleMesh(PrimativeMesh):
"""
simple equilateral triangle. oriented with normal in the Z+ direction, lying upon XY Plane, at origin
"""
def generate_vertices(self):
top_vertex_angle = math.pi / 2.0
return np.array([
self.dimensions[0] * math.cos(top_vertex_angle), self.dimensions[0] * math.sin(top_vertex_angle), 0.0, 1.0,
1.0, 0.0, 0.0, 1.0,
self.dimensions[0] * math.cos(top_vertex_angle + 2.0 * math.pi / 3.0),
self.dimensions[0] * math.sin(top_vertex_angle + 2.0 * math.pi / 3.0), 0.0, 0.0, 1.0, 0.0, 0.0, 1.0,
self.dimensions[0] * math.cos(top_vertex_angle + 4.0 * math.pi / 3.0),
self.dimensions[0] * math.sin(top_vertex_angle + 4.0 * math.pi / 3.0), 0.0, 0.0, 0.0, 0.0, 0.0, 1.0,
],
dtype=np.float32
)
class CubeMeshStatic(PrimativeMesh):
"""
Gives us a cube mesh, but does not proc-gen the cube using our face-and-excude system
my_cube = primatives.CubeMesh(
shader=shader,
diffuse=tile_textures[0],
specular=tile_textures[1],
shininess=32.0,
position=[0.0, 0.0, 10.0],
dimensions=[1.0, 3.0, 5.0],
rotation_axis=[0.5, 1.0, 0.1]
)
How to draw:
view = active_cam.get_view_matrix()
my_cube.draw(view=view)
"""
def generate_vertices(self):
return np.array([
# Z- facing
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 0.0, -1.0,
# Z+ facing
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, 1.0,
# X- facing
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, -1.0, 0.0, 0.0,
# X+ facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 1.0, 0.0, 1.0,
# Y- facing
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, -1.0, 0.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, -1.0, 0.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, -1.0, 0.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, -1.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, -1.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, -1.0, 0.0,
# Y+ facing
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 1.0, 0.0, ],
dtype=np.float32
)
class Skybox(PrimativeMesh):
def generate_vertices(self):
return np.array([
# Z- facing
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, -1.0,
# Z+ facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 0.0, 1.0,
# X- facing
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, -1.0, 0.0, 0.0,
# X+ facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 1.0, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.0, 1.0, 0.0, 1.0,
# Y- facing
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, -1.0, 0.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, -1.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, -1.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, -1.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, -1.0, 0.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, -1.0, 0.0,
# Y+ facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.0, 0.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 1.0, 0.0, 0.0, 1.0, 0.0,],
dtype=np.float32
)
class SkyboxEmission(PrimativeMeshEmission):
def generate_vertices(self):
return np.array([
# Z+ facing
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.34, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.66, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 0.66, 0.0, 0.0, -1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.34, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 0.66, 0.0, 0.0, -1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 1.0, 0.34, 0.0, 0.0, -1.0,
# Z- facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.66, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.34, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.34, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.34, 0.0, 0.0, 1.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.66, 0.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.66, 0.0, 0.0, 1.0,
# X- facing
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.34, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.66, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.66, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.66, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.34, -1.0, 0.0, 0.0,
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.0, 0.34, -1.0, 0.0, 0.0,
# X+ facing
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.34, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.66, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.66, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.34, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.66, 1.0, 0.0, 1.0,
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.75, 0.34, 1.0, 0.0, 1.0,
# Y- facing
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.333, 0.0, -1.0, 0.0, #e
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.5, 0.0, 0.0, -1.0, 0.0, #b
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.25, 0.0, 0.0, -1.0, 0.0, #c
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.25, 0.0, 0.0, -1.0, 0.0, #c
-1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.333, 0.0, -1.0, 0.0, #a
1.0 * self.dimensions[0], -1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.333, 0.0, -1.0, 0.0, #e
# Y+ facing
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.5, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.667, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.25, 1.0, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.25, 0.667, 0.0, 1.0, 0.0,
-1.0 * self.dimensions[0], 1.0 * self.dimensions[1], -1.0 * self.dimensions[2], 0.25, 1.0, 0.0, 1.0, 0.0,
1.0 * self.dimensions[0], 1.0 * self.dimensions[1], 1.0 * self.dimensions[2], 0.5, 0.667, 0.0, 1.0, 0.0, ],
dtype=np.float32
)
class FloorTileMesh(PrimativeMesh):
"""
generates a basic floor-tile for our little proc-gen roguelike
"""
def __init__(self,
shader,
# material properties
diffuse,
specular,
shininess=32.0,
# mesh properties
dimensions=[5.0, 5.0],
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
):
self.shader = shader
self.diffuse = diffuse
self.specular = specular
self.shininess = shininess
self.position = position
self.rotation_magnitude = glm.vec3(rotation_magnitude)
self.rotation_axis = glm.vec3(rotation_axis)
self.scale = glm.vec3(scale)
self.dimensions = dimensions
self.texture_horizontal_offset = [0.0, 0.0]
self.texture_vertical_offset = [0.0, 0.0]
self.vertices = self.generate_vertices()
# quad VAO
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# quad position vertices (vertex attribute)
glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
# quad texture coords
glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(12))
glEnableVertexAttribArray(1)
# quad normals
glVertexAttribPointer(2, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(20))
glEnableVertexAttribArray(2)
self.model_loc = glGetUniformLocation(self.shader, "model")
def generate_vertices(self):
# todo: seperate position from texture coord and normal
# then zip-up at then end
# todo: excude base
# todo: stitch two bases together (Get a prism)
# todo: make new tile class, based on this one, that will
# 1. outline points for a new base ontop of the floor
# 2. excude upwards
# 3. stitch together to get a wall
# todo: organize by surface
# a new class that holds vertices of a flat, polygonal surface
# cut a hole in the surface, re-stitch the triangles
# excude from a hole
# bezier curved-holes should be final touch
# focus on cutting, excuding, bevel(cut + excude + fill)
radius = self.dimensions[0]
number_of_sides = 4.0
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
faces[0].offset_outer_vertices(offset=glm.vec3([0.0, -1.0, 0.0]))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
faces[0].apply_hardset_quad_texture_coords()
# generate the second face via exude
faces.append(Face())
faces[1].extrude_from_other_face(other=faces[0], direction=[0.0, 1.0, 0.0], distance=self.dimensions[1])
# generate the 4 vertical faces
faces += self.stitch_faces(face_a=faces[0], face_b=faces[1], number_of_faces=number_of_sides)
vertices = []
for face in faces:
vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
def generate_outer_vertices(self, number_of_sides=4, initial_angle=None):
"""
generates the vertices of a simple polygon-no repeats
used for excuding and stitching, not for drawing
"""
outer_vertices = []
angle_of_rotation = 2 * pi / number_of_sides
if initial_angle == None:
initial_angle = pi / number_of_sides
for i in range(1, int(number_of_sides) + 1):
x_coord = cos(initial_angle + angle_of_rotation * -i) * self.dimensions[0]
z_coord = sin(initial_angle + angle_of_rotation * -i) * self.dimensions[0]
outer_vertices += [x_coord, 0.0, z_coord]
return outer_vertices
def outer_vertices_to_vertices(self, face, number_of_sides):
"""
Generates the real vertices (with normals, texture coords) of a face from the
outer vertices
:param face: the face we want to generate vertices for. must have its outer vertices already
:param number_of_sides: eg, 4 for square, 5 for pentagon
:return: the vertices to be drawn by opengl
"""
vertices = []
for i in range(int(number_of_sides) - 2):
triangle_vertices = [Vertex(), Vertex(), Vertex()]
triangle_vertices[0].positions = glm.vec3([
face.outer_vertices[0],
face.outer_vertices[1],
face.outer_vertices[2]
])
triangle_vertices[0].texture_coordinates = glm.vec2([
face.outer_vertices[0] / 2.0 + 0.5,
face.outer_vertices[2] / 2.0 + 0.5
])
triangle_vertices[0].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[1].positions = glm.vec3([
face.outer_vertices[3 * i + 0 + 3 * 1],
face.outer_vertices[3 * i + 1 + 3 * 1],
face.outer_vertices[3 * i + 2 + 3 * 1]
])
triangle_vertices[1].texture_coordinates = glm.vec2([
face.outer_vertices[3 * i + 0 + 3 * 1] / 2.0 + 0.5,
face.outer_vertices[3 * i + 2 + 3 * 1] / 2.0 + 0.5
])
triangle_vertices[1].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[2].positions = glm.vec3([
face.outer_vertices[3 * i + 0 + 3 * 2],
face.outer_vertices[3 * i + 1 + 3 * 2],
face.outer_vertices[3 * i + 2 + 3 * 2]
])
triangle_vertices[2].texture_coordinates = glm.vec2([
face.outer_vertices[3 * i + 0 + 3 * 2] / 2.0 + 0.5,
face.outer_vertices[3 * i + 2 + 3 * 2] / 2.0 + 0.5
])
triangle_vertices[2].normals = glm.vec3([0.0, 1.0, 0.0])
for vertex in triangle_vertices:
vertex.normals = calculate_normal([
triangle_vertices[0].positions,
triangle_vertices[1].positions,
triangle_vertices[2].positions
])
vertices += triangle_vertices
return vertices
class Face:
"""
convert Face into a 1-dimensional numpy array that opengl can draw
"""
def __init__(self, vertices=[], outer_vertices=[]):
self.vertices = vertices
self.outer_vertices = outer_vertices # non-repeating vertices, to use for extruding, stitching, etc
self.tile_width = 2.0 # TODO: make this a parameter
self.texture_horizontal_offset = [0.0, 0.0]
self.texture_vertical_offset = [0.0, 0.0]
self.normal = glm.vec3([0.0, 1.0, 0.0])
self.outer_vertices_vec3 = None
self.radius = 1.0 # todo calculate this from outer_vertices or something
self.sides = 0
self.texture_atlas_index = 0
def calculate_sides(self):
self.calculate_outer_vertices_as_vec3()
self.sides = len(self.outer_vertices_vec3)
def offset_vertices_all(self, offset=glm.vec3([0.0, 0.0, 0.0])):
"""
offset the outer vertices and inner vertices
:return: None
"""
self.offset_outer_vertices(offset=offset)
for vertex in self.vertices:
vertex.positions += offset
def calculate_normal(self, reverse=False):
"""
uses outer vertices to calculate normal.
:return: glm.vec3 representing the normal to this face's plane
"""
try:
triangle = [
self.outer_vertices_vec3[0],
self.outer_vertices_vec3[1],
self.outer_vertices_vec3[2]
]
if reverse:
triangle = [
self.outer_vertices_vec3[0],
self.outer_vertices_vec3[2],
self.outer_vertices_vec3[1]
]
normal = calculate_normal(
triangle
)
except TypeError:
self.calculate_outer_vertices_as_vec3()
triangle = [
self.outer_vertices_vec3[0],
self.outer_vertices_vec3[1],
self.outer_vertices_vec3[2]
]
if reverse:
triangle = [
self.outer_vertices_vec3[0],
self.outer_vertices_vec3[2],
self.outer_vertices_vec3[1]
]
normal = calculate_normal(
triangle
)
return normal
def calculate_outer_vertices_as_vec3(self):
"""
does not return a value, but assigns to attribute
:return: nothing
"""
self.outer_vertices_vec3 = list_to_vec3_list(self.outer_vertices)
def calculate_radius(self):
if self.outer_vertices_vec3 == None:
self.calculate_outer_vertices_as_vec3()
self.radius = glm.distance(calculate_mean_point(self.outer_vertices_vec3), self.outer_vertices_vec3[0])
def set_texture_offsets(self, vertical, horizontal):
self.texture_horizontal_offset = horizontal
self.texture_vertical_offset = vertical
def face_to_list(self):
vertices_list = []
for vertex in self.vertices:
vertices_list.append(vertex.vertex_to_list())
return vertices_list
def extrude_from_other_face(
self,
other,
direction=(0.0, 1.0, 0.0),
distance=1.0,
flip_base=True,
scale=None,
):
"""
takes in another face, copies points with positional offset. Only effects outer vertices, so call to
generate outer vertices is required as next step
todo: this should live in primative mesh. needs some refactoring to do so
"""
offset = glm.vec3(direction) * distance
all_vertices = []
for vertex in other.vertices:
updated_vertex = Vertex()
if scale is None:
updated_vertex.positions = glm.vec3(vertex.positions) + offset
else:
radius_adjusted_vertices = (
glm.vec3(vertex.positions).x * scale,
glm.vec3(vertex.positions).y,
glm.vec3(vertex.positions).z * scale
)
updated_vertex.positions = glm.vec3(radius_adjusted_vertices) + offset
updated_vertex.texture_coordinates = glm.vec2(vertex.texture_coordinates)
updated_vertex.normals = glm.vec3(vertex.normals)
if flip_base:
vertex.normals *= -1.0
all_vertices.append(updated_vertex)
if flip_base:
other.vertices.reverse()
self.vertices = all_vertices
outer_vertices = [0.0] * len(other.outer_vertices)
for vertex_index in range(int(len(other.outer_vertices) / 3)):
if scale is None:
outer_vertices[vertex_index * 3] = other.outer_vertices[vertex_index * 3] + offset.x
outer_vertices[vertex_index * 3 + 1] = other.outer_vertices[vertex_index * 3 + 1] + offset.y
outer_vertices[vertex_index * 3 + 2] = other.outer_vertices[vertex_index * 3 + 2] + offset.z
else:
outer_vertices[vertex_index * 3] = other.outer_vertices[vertex_index * 3] * scale + offset.x
outer_vertices[vertex_index * 3 + 1] = other.outer_vertices[vertex_index * 3 + 1] + offset.y
outer_vertices[vertex_index * 3 + 2] = other.outer_vertices[vertex_index * 3 + 2] * scale + offset.z
self.outer_vertices = outer_vertices.copy()
self.calculate_sides()
self.vertices = self.outer_vertices_to_vertices(reverse_texture_coordinates=True)
def extrude_from_other_face_2(self, other, direction=[0.0, 1.0, 0.0], distance=1.0, flip_base=True, radius=1.0):
"""
Trying a new extrude process WIP
"""
offset = glm.vec3(direction) * distance
all_vertices = []
other.calculate_outer_vertices_as_vec3()
outer_vertices_other = other.outer_vertices_vec3.copy()
if flip_base:
outer_vertices_other.reverse()
self.outer_vertices_vec3 = outer_vertices_other.copy()
for vertex in self.outer_vertices_vec3:
vertex.x *= radius
vertex.z *= radius
vertex.x += offset.x
vertex.y += offset.y
vertex.z += offset.z
if flip_base:
self.outer_vertices = vec3_list_to_list(
rotate_list(self.outer_vertices_vec3, math.ceil(len(self.outer_vertices_vec3) / 2)))
else:
self.outer_vertices = vec3_list_to_list(self.outer_vertices_vec3)
self.vertices = self.outer_vertices_to_vertices()
# todo: this should work....
for vertex, other_vertex in zip(self.vertices, other.vertices):
vertex.texture_coordinates = glm.vec2(other_vertex.texture_coordinates)
def update_texture_coords_using_atlas_index(self, texture_atlas_index, texture_atlas_size):
"""
uses the texture atlas index to modify the per-vertex texture coordinates of this face
:param texture_atlas_index: left->right, top->bottom
:param texture_atlas_size: the length of the texture atlas (always square)
:return:
"""
self.texture_atlas_index = texture_atlas_index
# texture atlas index into row and column indices
column_index = float(texture_atlas_index % texture_atlas_size)
row_index = (texture_atlas_size - 1) - math.floor(texture_atlas_index / texture_atlas_size)
# row and column indices into lower and upper bounds of texture coords (4 total)
# column (x-axis)lower and upper
lower_texture_coord_x_axis = 1.0 / texture_atlas_size * column_index
upper_texture_coord_x_axis = 1.0 / texture_atlas_size * (column_index + 1.0)
# assuming square textures
subtexture_magnitude = upper_texture_coord_x_axis - lower_texture_coord_x_axis
lower_texture_coord_y_axis = 1.0 / texture_atlas_size * row_index
upper_texture_coord_y_axis = 1.0 / texture_atlas_size * (row_index + 1.0)
# update the current vertices' text coords
for vertex in self.vertices:
vertex.texture_coordinates[0] = lower_texture_coord_x_axis + vertex.texture_coordinates[
0] * subtexture_magnitude
vertex.texture_coordinates[1] = lower_texture_coord_y_axis + vertex.texture_coordinates[
1] * subtexture_magnitude
def apply_hardset_quad_texture_coords(self):
"""
A brute-force fix to getting correct texture coords on the slab
in future: a function that takes number of sides, and generates dictionary like below.
outer_to_vertices function then reads this dictionary
-found better solution in bezier_cut()
"""
horizontal_offset = self.texture_horizontal_offset
vertical_offset = self.texture_vertical_offset
quad_texture_coords = {
1: glm.vec2([0.0 + horizontal_offset[0], 0.0 + vertical_offset[0]]),
0: glm.vec2([0.0 + horizontal_offset[0], 1.0 + vertical_offset[1]]),
3: glm.vec2([1.0 + horizontal_offset[1], 1.0 + vertical_offset[1]]),
2: glm.vec2([1.0 + horizontal_offset[1], 0.0 + vertical_offset[0]]),
}
self.vertices[0].texture_coordinates = glm.vec2(quad_texture_coords[0])
self.vertices[1].texture_coordinates = glm.vec2(quad_texture_coords[1])
self.vertices[2].texture_coordinates = glm.vec2(quad_texture_coords[2])
self.vertices[3].texture_coordinates = glm.vec2(quad_texture_coords[0])
self.vertices[4].texture_coordinates = glm.vec2(quad_texture_coords[2])
self.vertices[5].texture_coordinates = glm.vec2(quad_texture_coords[3])
def apply_hardset_triangle_texture_coords(self):
"""
another hacky solution of calculating texture coordinates
"""
self.vertices[0].texture_coordinates = glm.vec2([0.0, 0.0])
self.vertices[1].texture_coordinates = glm.vec2([1.0, 0.0])
self.vertices[2].texture_coordinates = glm.vec2([0.5, 1.0])
def offset_outer_vertices(self, offset=glm.vec3([0.0, 0.0, 0.0])):
for i in range(len(self.outer_vertices)):
if i % 3 == 0:
self.outer_vertices[i] += offset.x
if i % 3 == 1:
self.outer_vertices[i] += offset.y
if i % 3 == 2:
self.outer_vertices[i] += offset.z
# print(self.outer_vertices)
def generate_texture_coordinates_regular_polygon(self, sides, reverse=True):
"""
A new means of generating texture coordinates for any (regular) polygon.
Even works for polygons not on XZ plane at origin
:return: a list of glm.vec2() coordinates, one for each point of shape
"""
angle_delta = -2.0 * pi / sides
angle_initial = (pi / sides) - (pi / 2.0)
radius = 0.5
if sides == 4:
radius = 1.0 / math.sqrt(2.0)
angle_initial = (pi / sides) + angle_delta
texture_coordinates = []
for index_point in range(sides):
coordinate = glm.vec2([0.0, 0.0])
coordinate.x = (cos(angle_initial - angle_delta * index_point) * radius) + 0.5
coordinate.y = (sin(angle_initial - angle_delta * index_point) * radius) + 0.5
texture_coordinates.append(coordinate)
if reverse:
texture_coordinates.reverse()
return texture_coordinates
def generate_texture_coordinates_polygon(self, reverse=False):
"""
Generate texture coordinates for any convex polygon (even irregular)
in 3D space.
:param reverse: reverse the texture coordinates at the end
:return: list of coordinates as vec2
"""
texture_coordinates = []
self.calculate_outer_vertices_as_vec3()
sides = len(self.outer_vertices_vec3)
# 1 get center point of polygon
center_point = calculate_mean_point(points=self.outer_vertices_vec3)
# get distances from center point to each vertex
# also, find the largest distance
vertex_distances = []
largest_distance = 0
for point in self.outer_vertices_vec3:
distance = glm.distance(point, center_point)
vertex_distances.append(distance)
if distance > largest_distance:
largest_distance = distance
# scale the distances by the largest
for index, distance in enumerate(vertex_distances):
vertex_distances[index] = distance / (largest_distance * 2)
# Get the angles between the vertices
vertex_angles = []
angle_sum = 0
for index in range(sides - 1):
a = self.outer_vertices_vec3[index]
b = self.outer_vertices_vec3[index + 1]
c = center_point
angle = self.calculate_angle(a, b, c)
vertex_angles.append(angle)
angle_sum += angle
vertex_angles.append(2 * pi - angle_sum)
is_even = False
if sides % 2 == 0:
is_even = True
# calculate initial angle offset (rotate the whole triangle)
if is_even:
if (sides / 2) % 2 == 0:
angle_initial = pi / 2 + vertex_angles[0] + vertex_angles[1] / 2
if sides == 8:
angle_initial = pi + vertex_angles[2] + vertex_angles[1] / 2
elif sides == 12:
angle_initial = pi + vertex_angles[2] / 2 + vertex_angles[0] + vertex_angles[1]
else:
# 6
angle_initial = pi + vertex_angles[0]
if sides == 10:
angle_initial = pi + vertex_angles[0] + vertex_angles[1]
else:
angle_initial = pi + ((pi - vertex_angles[-1]) / 2)
vertex_distances = rotate_list(vertex_distances, steps=1)
# go around the unit circle, angle by angle, and place a point at the distance away
for index in range(0, sides):
angle_current = sum(vertex_angles[:index + 1])
coordinate = glm.vec2((0.0, 0.0))
coordinate.x = (cos(angle_initial - angle_current) * vertex_distances[index]) + 0.5
coordinate.y = (sin(angle_initial - angle_current) * vertex_distances[index]) + 0.5
texture_coordinates.append(coordinate)
if is_even:
texture_coordinates = rotate_list(texture_coordinates, steps=sides - 1)
else:
texture_coordinates = rotate_list(texture_coordinates, steps=sides - 1)
if reverse:
texture_coordinates.reverse()
return texture_coordinates
def calculate_angle(self, a, b, c):
CA = a - c
CB = b - c
angle = glm.acos(glm.dot(CA, CB) / (glm.length(CA) * glm.length(CB)))
return angle
def outer_vertices_to_vertices(self, reverse_texture_coordinates=False, regular=False):
"""
Generates the real vertices (with normals, texture coords) of a face
:param regular: polygon is regular or irregular
:param reverse_texture_coordinates: Flips texture coordinates about the textures Y-Axis
:param number_of_sides: eg, 4 for square, 5 for pentagon
:return: vertices ready to be drawn by opengl
"""
if not self.sides:
self.calculate_sides()
number_of_sides = self.sides
if not self.outer_vertices_vec3:
self.calculate_outer_vertices_as_vec3()
if regular:
texture_coordinates = self.generate_texture_coordinates_regular_polygon(sides=number_of_sides)
else:
texture_coordinates = self.generate_texture_coordinates_polygon(reverse=reverse_texture_coordinates)
vertices = []
texture_index = 1
for i in range(int(number_of_sides) - 2):
triangle_vertices = [Vertex(), Vertex(), Vertex()]
triangle_vertices[0].positions = glm.vec3((
self.outer_vertices[0],
self.outer_vertices[1],
self.outer_vertices[2]
))
triangle_vertices[0].texture_coordinates = glm.vec2((
texture_coordinates[0].x,
texture_coordinates[0].y
))
triangle_vertices[0].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[1].positions = glm.vec3((
self.outer_vertices[3 * i + 0 + 3 * 1],
self.outer_vertices[3 * i + 1 + 3 * 1],
self.outer_vertices[3 * i + 2 + 3 * 1]
))
triangle_vertices[1].texture_coordinates = glm.vec2((
texture_coordinates[texture_index].x,
texture_coordinates[texture_index].y
))
triangle_vertices[1].normals = glm.vec3([0.0, 1.0, 0.0])
triangle_vertices[2].positions = glm.vec3((
self.outer_vertices[3 * i + 0 + 3 * 2],
self.outer_vertices[3 * i + 1 + 3 * 2],
self.outer_vertices[3 * i + 2 + 3 * 2]
))
triangle_vertices[2].texture_coordinates = glm.vec2((
texture_coordinates[texture_index + 1].x,
texture_coordinates[texture_index + 1].y
))
triangle_vertices[2].normals = glm.vec3((0.0, 1.0, 0.0))
for vertex in triangle_vertices:
vertex.normals = calculate_normal([
triangle_vertices[0].positions,
triangle_vertices[1].positions,
triangle_vertices[2].positions
])
vertices += triangle_vertices
texture_index += 1
return vertices
def flip_winding_order(self):
"""
flips this winding order of this face, mirroring vertices outer vertices
"""
self.outer_vertices_vec3.reverse()
self.outer_vertices = vec3_list_to_list(self.outer_vertices_vec3)
self.outer_vertices_to_vertices()
class Vertex():
def __init__(self):
self.positions = glm.vec3()
self.normals = glm.vec3()
self.texture_coordinates = glm.vec2()
def vertex_to_list(self):
"""
convert positions, normals, and texture coords into one big 1-dimensional list
"""
vertex_as_list = []
for position in self.positions:
vertex_as_list.append(position)
for texture_coord in self.texture_coordinates:
vertex_as_list.append(texture_coord)
for normal in self.normals:
vertex_as_list.append(normal)
return vertex_as_list
class WallTileMesh(FloorTileMesh):
"""
straight walls for wfc world
"""
def generate_vertices(self):
radius = self.dimensions[0]
number_of_sides = 4.0
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
faces[0].offset_outer_vertices(offset=glm.vec3([0.0, -1.0, 0.0]))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
faces[0].apply_hardset_quad_texture_coords()
# generate the second face via exude
faces.append(Face())
faces[1].extrude_from_other_face(other=faces[0], direction=[0.0, 1.0, 0.0], distance=self.dimensions[1])
# generate the 4 vertical faces
faces += self.stitch_faces(face_a=faces[0], face_b=faces[1], number_of_faces=number_of_sides)
# generate the top of the wall
x_border = 0.0
z_border = 0.5
y_border = 2.0
y_wall_bottom = self.dimensions[1]
top_of_wall = Face()
bottom_of_wall = Face()
# TODO: refactor this big time
top_of_wall.outer_vertices = faces[0].outer_vertices.copy()
top_of_wall.outer_vertices[0] -= x_border
top_of_wall.outer_vertices[1] += y_border
top_of_wall.outer_vertices[2] += z_border
top_of_wall.outer_vertices[3] += x_border
top_of_wall.outer_vertices[4] += y_border
top_of_wall.outer_vertices[5] += z_border
top_of_wall.outer_vertices[6] += x_border
top_of_wall.outer_vertices[7] += y_border
top_of_wall.outer_vertices[8] -= z_border
top_of_wall.outer_vertices[9] -= x_border
top_of_wall.outer_vertices[10] += y_border
top_of_wall.outer_vertices[11] -= z_border
top_of_wall.vertices = self.outer_vertices_to_vertices(number_of_sides=4, face=top_of_wall)
# top_of_wall.apply_hardset_quad_texture_coords()
faces.append(top_of_wall)
bottom_of_wall.outer_vertices = faces[0].outer_vertices.copy()
bottom_of_wall.outer_vertices[0] -= x_border
bottom_of_wall.outer_vertices[1] += y_wall_bottom
bottom_of_wall.outer_vertices[2] += z_border
bottom_of_wall.outer_vertices[3] += x_border
bottom_of_wall.outer_vertices[4] += y_wall_bottom
bottom_of_wall.outer_vertices[5] += z_border
bottom_of_wall.outer_vertices[6] += x_border
bottom_of_wall.outer_vertices[7] += y_wall_bottom
bottom_of_wall.outer_vertices[8] -= z_border
bottom_of_wall.outer_vertices[9] -= x_border
bottom_of_wall.outer_vertices[10] += y_wall_bottom
bottom_of_wall.outer_vertices[11] -= z_border
bottom_of_wall.vertices = self.outer_vertices_to_vertices(number_of_sides=4, face=bottom_of_wall)
# wall vertical quads
faces += self.stitch_faces(face_a=bottom_of_wall, face_b=top_of_wall, number_of_faces=number_of_sides)
vertices = []
beveled_faces = []
for face in faces:
beveled_faces += self.bevel_cut(face, bevel_depths=[0.0, -0.2], border_sizes=[0.6, 0.2])
for face in beveled_faces:
vertices += face.face_to_list()
# for face in faces:
# vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
class DoorwayTileMesh(FloorTileMesh):
"""
straight walls for wfc world
"""
def generate_vertices(self):
radius = self.dimensions[0]
number_of_sides = 4.0
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
faces[0].offset_outer_vertices(offset=glm.vec3([0.0, -1.0, 0.0]))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
# generate the second face via exude
faces.append(Face())
faces[1].extrude_from_other_face(other=faces[0], direction=[0.0, 1.0, 0.0], distance=self.dimensions[1])
# generate the 4 vertical faces
faces += self.stitch_faces(face_a=faces[0], face_b=faces[1], number_of_faces=number_of_sides)
# generate the top of the wall
x_border = 0.0
z_border = 0.5
y_border = 2.0
y_wall_bottom = self.dimensions[1]
top_of_wall = Face()
bottom_of_wall = Face()
# TODO: refactor this big time
top_of_wall.outer_vertices = faces[0].outer_vertices.copy()
top_of_wall.outer_vertices[0] -= x_border
top_of_wall.outer_vertices[1] += y_border
top_of_wall.outer_vertices[2] += z_border
top_of_wall.outer_vertices[3] += x_border
top_of_wall.outer_vertices[4] += y_border
top_of_wall.outer_vertices[5] += z_border
top_of_wall.outer_vertices[6] += x_border
top_of_wall.outer_vertices[7] += y_border
top_of_wall.outer_vertices[8] -= z_border
top_of_wall.outer_vertices[9] -= x_border
top_of_wall.outer_vertices[10] += y_border
top_of_wall.outer_vertices[11] -= z_border
top_of_wall.vertices = self.outer_vertices_to_vertices(number_of_sides=4, face=top_of_wall)
faces.append(top_of_wall)
bottom_of_wall.outer_vertices = faces[0].outer_vertices.copy()
bottom_of_wall.outer_vertices[0] -= x_border
bottom_of_wall.outer_vertices[1] += y_wall_bottom
bottom_of_wall.outer_vertices[2] += z_border
bottom_of_wall.outer_vertices[3] += x_border
bottom_of_wall.outer_vertices[4] += y_wall_bottom
bottom_of_wall.outer_vertices[5] += z_border
bottom_of_wall.outer_vertices[6] += x_border
bottom_of_wall.outer_vertices[7] += y_wall_bottom
bottom_of_wall.outer_vertices[8] -= z_border
bottom_of_wall.outer_vertices[9] -= x_border
bottom_of_wall.outer_vertices[10] += y_wall_bottom
bottom_of_wall.outer_vertices[11] -= z_border
bottom_of_wall.vertices = self.outer_vertices_to_vertices(number_of_sides=4, face=bottom_of_wall)
# wall vertical faces
faces_vertical = []
faces_vertical += self.stitch_faces(face_a=bottom_of_wall, face_b=top_of_wall, number_of_faces=number_of_sides)
# ? order of faces?
# -need to to get the two z-facing faces (parallel to x-axis)
# should be 0 and 2 index
doorway_faces = []
doorway_faces += self.subdivide_into_doorway(
face_z_negative=faces_vertical[0],
face_z_positive=faces_vertical[2],
doorway_width=0.4,
doorway_height=.8
)
# delete faces from which we cut a door
del faces_vertical[0]
del faces_vertical[1]
# add all faces to list, and numpy-ify so opengl can draw
faces += faces_vertical + doorway_faces
vertices = []
beveled_faces = []
for face in faces:
beveled_faces += self.bevel_cut(face, bevel_depths=[0.0, -0.2], border_sizes=[0.6, 0.2])
for face in beveled_faces:
vertices += face.face_to_list()
# for face in faces:
# vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
def subdivide_into_vertical_thirds(self, face, center_width=0.4, remove_center=False, z_positive=False):
"""
use: making doors/doorways
divides a square face into 3 vertical rectangles. returns 3 new faces
:param center_width: width of center rectangle
:return subfaces, a list of 3(2 for hole) faces
"""
subfaces = [] # return
center_face = Face()
left_face = Face()
right_face = Face()
center_width_normalized = center_width
center_width *= face.tile_width
if z_positive:
upper_left = glm.vec3(face.outer_vertices[0], face.outer_vertices[1], face.outer_vertices[2])
lower_left = glm.vec3(face.outer_vertices[3], face.outer_vertices[4], face.outer_vertices[5])
lower_right = glm.vec3(face.outer_vertices[6], face.outer_vertices[7], face.outer_vertices[8])
upper_right = glm.vec3(face.outer_vertices[9], face.outer_vertices[10], face.outer_vertices[11])
else:
upper_left = glm.vec3(face.outer_vertices[0], face.outer_vertices[1], face.outer_vertices[2])
lower_left = glm.vec3(face.outer_vertices[3], face.outer_vertices[4], face.outer_vertices[5])
lower_right = glm.vec3(face.outer_vertices[6], face.outer_vertices[7], face.outer_vertices[8])
upper_right = glm.vec3(face.outer_vertices[9], face.outer_vertices[10], face.outer_vertices[11])
# center quad's outer vertices
midpoint_horizontal = (lower_right.x + lower_left.x) / 2.0
face_plane_z = lower_left.z
face_upper_y = upper_left.y
face_lower_y = lower_left.y
center_lower_left = glm.vec3(midpoint_horizontal - center_width / 2.0, face_lower_y, face_plane_z)
center_lower_right = glm.vec3(midpoint_horizontal + center_width / 2.0, face_lower_y, face_plane_z)
center_upper_right = glm.vec3(midpoint_horizontal + center_width / 2.0, face_upper_y, face_plane_z)
center_upper_left = glm.vec3(midpoint_horizontal - center_width / 2.0, face_upper_y, face_plane_z)
if z_positive:
center_face.outer_vertices = list(center_upper_left) + list(center_lower_left) + \
list(center_lower_right) + list(center_upper_right)
else:
center_face.outer_vertices = list(center_upper_right) + list(center_lower_right) + list(
center_lower_left) + list(
center_upper_left)
center_face.vertices = center_face.outer_vertices_to_vertices()
border_width = (1.0 - center_width_normalized) / 2
center_face.set_texture_offsets(horizontal=[border_width, -border_width], vertical=[0.0, 0.0])
center_face.apply_hardset_quad_texture_coords()
if not remove_center:
subfaces.append(center_face)
# subfaces += self.subdivide_into_horizontal_halves(center_face, bottom_height=0.9, z_positive=False)
# now make the 2 border faces
# left
if z_positive:
left_face.outer_vertices = list(upper_left) + list(lower_left) + \
list(center_lower_left) + list(center_upper_left)
left_face.vertices = left_face.outer_vertices_to_vertices()
left_face.set_texture_offsets(horizontal=[0, -(border_width + center_width_normalized)],
vertical=[0.0, 0.0])
left_face.apply_hardset_quad_texture_coords()
else: # right
left_face.outer_vertices = list(center_upper_left) + list(center_lower_left) + \
list(lower_right) + list(upper_right)
left_face.vertices = left_face.outer_vertices_to_vertices()
left_face.set_texture_offsets(horizontal=[border_width + center_width_normalized, 0], vertical=[0.0, 0.0])
left_face.apply_hardset_quad_texture_coords()
subfaces.append(left_face)
# right
if z_positive:
right_face.outer_vertices = list(center_upper_right) + list(center_lower_right) + \
list(lower_right) + list(upper_right)
right_face.vertices = right_face.outer_vertices_to_vertices()
right_face.set_texture_offsets(horizontal=[border_width + center_width_normalized, 0], vertical=[0.0, 0.0])
right_face.apply_hardset_quad_texture_coords()
else: # left
right_face.outer_vertices = list(upper_left) + list(lower_left) + \
list(center_lower_right) + list(center_upper_right)
right_face.vertices = right_face.outer_vertices_to_vertices()
right_face.set_texture_offsets(horizontal=[0, -(border_width + center_width_normalized)],
vertical=[0.0, 0.0])
right_face.apply_hardset_quad_texture_coords()
subfaces.append(right_face)
return subfaces
def subdivide_into_horizontal_halves(self, face, bottom_height=0.5, z_positive=False):
"""
use: making doors/doorways
divides a square face into 2 quad faces. returns 2 new faces
:param bottom_height: height of lower rectangle
:return subfaces, a list of 2 faces
"""
subfaces = [] # return
bottom_face = Face()
top_face = Face()
bottom_height_normalized = bottom_height
# bottom_height *= face.tile_width
# if z_positive:
upper_right = glm.vec3(face.outer_vertices[0], face.outer_vertices[1], face.outer_vertices[2])
lower_right = glm.vec3(face.outer_vertices[3], face.outer_vertices[4], face.outer_vertices[5])
lower_left = glm.vec3(face.outer_vertices[6], face.outer_vertices[7], face.outer_vertices[8])
upper_left = glm.vec3(face.outer_vertices[9], face.outer_vertices[10], face.outer_vertices[11])
bottom_height = bottom_height * (upper_right.y - lower_right.y)
# else:
# lower_right = glm.vec3(face.outer_vertices[0], face.outer_vertices[1], face.outer_vertices[2])
# upper_right = glm.vec3(face.outer_vertices[3], face.outer_vertices[4], face.outer_vertices[5])
# upper_left = glm.vec3(face.outer_vertices[6], face.outer_vertices[7], face.outer_vertices[8])
# lower_left = glm.vec3(face.outer_vertices[9], face.outer_vertices[10], face.outer_vertices[11])
# middle line points
middle_left = lower_left + glm.vec3(0.0, bottom_height, 0.0)
middle_right = lower_right + glm.vec3(0.0, bottom_height, 0.0)
# bottom quads outer vertices
bottom_lower_left = lower_left
bottom_lower_right = lower_right
bottom_upper_left = middle_left
bottom_upper_right = middle_right
if z_positive:
bottom_face.outer_vertices = list(bottom_upper_left) + list(bottom_lower_left) + \
list(bottom_lower_right) + list(bottom_upper_right)
else:
bottom_face.outer_vertices = list(bottom_upper_right) + list(bottom_lower_right) + \
list(bottom_lower_left) + list(bottom_upper_left)
bottom_face.vertices = bottom_face.outer_vertices_to_vertices()
bottom_face.set_texture_offsets(horizontal=face.texture_horizontal_offset,
vertical=[0.0, -(1.0 - bottom_height_normalized)])
bottom_face.apply_hardset_quad_texture_coords()
subfaces.append(bottom_face)
# top quads outer vertices
top_lower_left = middle_left
top_lower_right = middle_right
top_upper_left = upper_left
top_upper_right = upper_right
if z_positive:
top_face.outer_vertices = list(top_upper_left) + list(top_lower_left) + \
list(top_lower_right) + list(top_upper_right)
else:
top_face.outer_vertices = list(top_upper_right) + list(top_lower_right) + \
list(top_lower_left) + list(top_upper_left)
top_face.vertices = top_face.outer_vertices_to_vertices()
top_face.set_texture_offsets(horizontal=face.texture_horizontal_offset,
vertical=[bottom_height_normalized, 0.0])
top_face.apply_hardset_quad_texture_coords()
subfaces.append(top_face)
return subfaces
def subdivide_into_doorway(self, face_z_positive, face_z_negative, doorway_width=0.5, doorway_height=0.8):
"""
Takes 2 parallel faces and returns the 9 faces of a doorway
TODO:
just make it take 2 faces in and return one with the doorway-cut
:param face_z_positive: face with normal towards z+
:param face_z_negative: face with normal towards z-
:param center_width: width of door
:return:
"""
faces = []
third_faces_z_positive = self.subdivide_into_vertical_thirds(
face=face_z_positive,
center_width=doorway_width,
remove_center=False,
z_positive=True
)
faces.append(third_faces_z_positive[1]) # left of door
faces.append(third_faces_z_positive[2]) # right of door
door_faces_z_positive = self.subdivide_into_horizontal_halves(
face=third_faces_z_positive[0],
bottom_height=doorway_height,
z_positive=False
) # subdivide the center face
faces.append(door_faces_z_positive[1]) # above door
third_faces_z_negative = self.subdivide_into_vertical_thirds(
face=face_z_negative,
center_width=doorway_width,
remove_center=False,
z_positive=False
)
faces.append(third_faces_z_negative[1]) # left of door
faces.append(third_faces_z_negative[2]) # right of door
door_faces_z_negative = self.subdivide_into_horizontal_halves(
face=third_faces_z_negative[0],
bottom_height=doorway_height,
z_positive=False
) # subdivide the center face
faces.append(door_faces_z_negative[1]) # above door
# left side of the inner door hallway
door_hole_z_negative = door_faces_z_negative[0].outer_vertices
door_hole_z_positive = door_faces_z_positive[0].outer_vertices
# get our outer vertices z+
z_positive_upper_left = glm.vec3(door_hole_z_positive[0], door_hole_z_positive[1], door_hole_z_positive[2])
z_positive_lower_left = glm.vec3(door_hole_z_positive[3], door_hole_z_positive[4], door_hole_z_positive[5])
z_positive_lower_right = glm.vec3(door_hole_z_positive[6], door_hole_z_positive[7], door_hole_z_positive[8])
z_positive_upper_right = glm.vec3(door_hole_z_positive[9], door_hole_z_positive[10], door_hole_z_positive[11])
# get our outer vertices z-
z_negative_upper_right = glm.vec3(door_hole_z_negative[0], door_hole_z_negative[1], door_hole_z_negative[2])
z_negative_lower_right = glm.vec3(door_hole_z_negative[3], door_hole_z_negative[4], door_hole_z_negative[5])
z_negative_lower_left = glm.vec3(door_hole_z_negative[6], door_hole_z_negative[7], door_hole_z_negative[8])
z_negative_upper_left = glm.vec3(door_hole_z_negative[9], door_hole_z_negative[10], door_hole_z_negative[11])
# construct left side
left_hall = Face()
left_hall.outer_vertices = list(z_positive_upper_left) + list(z_positive_lower_left) + list(
z_negative_lower_left) + list(z_negative_upper_left)
left_hall.vertices = left_hall.outer_vertices_to_vertices()
left_hall.set_texture_offsets(vertical=[0.0, 0.0], horizontal=[0.0, 0.0])
left_hall.apply_hardset_quad_texture_coords()
faces.append(left_hall)
# construct left side
right_hall = Face()
right_hall.outer_vertices = list(z_negative_upper_right) + list(z_negative_lower_right) + \
list(z_positive_lower_right) + list(z_positive_upper_right)
right_hall.vertices = right_hall.outer_vertices_to_vertices()
right_hall.set_texture_offsets(vertical=[0.0, 0.0], horizontal=[0.0, 0.0])
right_hall.apply_hardset_quad_texture_coords()
faces.append(right_hall)
# construct ceiling
ceiling_hall = Face()
ceiling_hall.outer_vertices = list(z_negative_upper_right) + list(z_positive_upper_right) + \
list(z_positive_upper_left) + list(z_negative_upper_left)
ceiling_hall.vertices = ceiling_hall.outer_vertices_to_vertices()
ceiling_hall.set_texture_offsets(vertical=[0.0, 0.0], horizontal=[0.0, 0.0])
ceiling_hall.apply_hardset_quad_texture_coords()
faces.append(ceiling_hall)
return faces
class Prism(PrimativeMesh):
"""
generates a prism based on an N-sided regular polygon.
"""
def __init__(self,
shader,
# material properties
diffuse,
specular,
shininess=32.0,
# mesh properties
dimensions=[1.0, 1.0], # radius, height
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
sides=3
):
self.shader = shader
self.diffuse = diffuse
self.specular = specular
self.shininess = shininess
self.position = glm.vec3(position)
self.rotation_magnitude = glm.vec3(rotation_magnitude)
self.rotation_axis = glm.vec3(rotation_axis)
self.scale = glm.vec3(scale)
self.dimensions = dimensions
self.texture_horizontal_offset = [0.0, 0.0]
self.texture_vertical_offset = [0.0, 0.0]
self.sides = sides
self.vertices = self.generate_vertices()
self.setup()
def setup(self):
# quad VAO
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# quad position vertices (vertex attribute)
glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
# quad texture coords
glVertexAttribPointer(1, 2, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(12))
glEnableVertexAttribArray(1)
# quad normals
glVertexAttribPointer(2, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 8, ctypes.c_void_p(20))
glEnableVertexAttribArray(2)
self.model_loc = glGetUniformLocation(self.shader, "model")
def generate_vertices(self):
"""
Generates the outer vertices and vertices of the triangle of this model
"""
radius = self.dimensions[0]
height = 0
try:
height = self.dimensions[1]
except IndexError:
height = math.sqrt(radius * radius * 2)
number_of_sides = self.sides
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
faces = [self._generate_polygonal_face(number_of_sides=number_of_sides)]
# generate the second face via exude
faces.append(Face())
faces[1].extrude_from_other_face(other=faces[0], direction=[0.0, 1.0, 0.0], distance=height)
# generate the 4 vertical faces
faces += self.stitch_faces(face_a=faces[0], face_b=faces[1], number_of_faces=number_of_sides)
vertices = []
for face in faces:
vertices += face.face_to_list()
return np.array(
vertices,
dtype=np.float32
).flatten()
class NPrismBezierCut(Prism):
def generate_vertices(self):
"""
Testing our bezier cut by applying it to an N-sided Prisms quad faces
"""
radius = self.dimensions[0]
height = 0
try:
height = self.dimensions[1]
except IndexError:
height = math.sqrt(radius * radius * 2)
number_of_sides = self.sides
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
faces = [Face()]
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
# faces[0].offset_outer_vertices(offset=glm.vec3([0.0, -1.0, 0.0]))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
# # faces[0].apply_hardset_quad_texture_coords()
# # generate the second face via exude
faces.append(Face())
faces[1].extrude_from_other_face(other=faces[0], direction=[0.0, 1.0, 0.0], distance=height)
# # generate the 4 vertical faces
faces += self.stitch_faces(face_a=faces[0], face_b=faces[1], number_of_faces=number_of_sides)
# quad_faces = self.divide_face_into_quads(original_face=faces[2], center_point_offset=0.0)
quad_faces = self.subdivide_quad_lengthwise(face=faces[2], subdivisions=3, widthwise=True)
for face in quad_faces:
faces_bevel_cut = self.bevel_cut(original_face=face, bevel_depths=[-0.0, -1.5], border_sizes=[0.1, 0.3],
depth=2, direction=glm.vec3(.5, 0.0, -1.0))
faces_from_bezier_cut = self.bezier_cut(face=faces_bevel_cut[-5], intervals=8, offset=0.2)
del faces_bevel_cut[-5]
for face in faces_from_bezier_cut:
faces.insert(3, face)
for face in faces_bevel_cut:
faces.insert(3, face)
del faces[2]
hip_faces = self.subdivide_hip_roof(faces[-1], start_point=[0.1, 0.1], end_point=[0.1, 0.9],
vertical_offset=1.0)
del faces[-1]
for face in hip_faces:
faces.insert(3, face)
vertices = []
index = 0
# faces[-3].update_texture_coords_using_atlas_index(texture_atlas_index=3, texture_atlas_size=2)
for face in faces:
face.update_texture_coords_using_atlas_index(texture_atlas_index=index % 4, texture_atlas_size=2)
vertices += face.face_to_list()
index += 1
return np.array(vertices,
dtype=np.float32
).flatten()
class SegmentedPrism(Prism):
"""
Makes potion bottles (segmented cylindar with different radius per segment)
"""
def __init__(self,
shader,
# material properties
diffuse,
specular,
shininess=32.0,
# mesh properties
dimensions=[1.0, 1.0], # radius, height
position=[0.0, 0.0, 0.0],
rotation_magnitude=0,
rotation_axis=glm.vec3([0.0, 0.0, 1.0]),
scale=glm.vec3([1.0, 1.0, 1.0]),
sides=3,
segments=1,
outer_vertices_offset=[0.0, -1.0, 0.0],
bevel_depths=[-0.2, -0.4],
border_sizes=[0.6, 0.0],
depth=2
):
self.shader = shader
self.diffuse = diffuse
self.specular = specular
self.shininess = shininess
self.position = glm.vec3(position)
self.rotation_magnitude = glm.vec3(rotation_magnitude)
self.rotation_axis = glm.vec3(rotation_axis)
self.scale = glm.vec3(scale)
self.dimensions = dimensions
self.texture_horizontal_offset = [0.0, 0.0]
self.texture_vertical_offset = [0.0, 0.0]
self.sides = sides
self.segments = segments
self.bevel_depths = bevel_depths
self.border_sizes = border_sizes
self.depth = depth
self.outer_vertices_offset = outer_vertices_offset
self.vertices = self.generate_vertices()
self.setup()
def generate_vertices(self):
radius = self.dimensions[0]
height = self.dimensions[1]
segment_height = height / self.segments
number_of_sides = self.sides
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
faces[0].offset_outer_vertices(offset=glm.vec3(self.outer_vertices_offset))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
# generate the second face via exude
base_faces = [faces[0]]
radius = 1.0
for index in range(1, self.segments + 1):
base_faces.append(Face())
if index < self.segments / 4:
radius *= 1.0 + 0.1 * index
else:
radius *= 0.9 # + .1*index
base_faces[index].extrude_from_other_face(other=base_faces[index - 1], direction=[0.0, 1.0, 0.0],
distance=segment_height, radius=radius)
faces += self.stitch_faces(face_a=base_faces[index - 1], face_b=base_faces[index],
number_of_faces=number_of_sides)
if index == self.segments:
faces.append(base_faces[index])
vertices = []
for face in faces:
vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
class SegmentedPrismBevelCutTest(SegmentedPrism):
"""
testing our bevel cut on an a multi-faced, non-axis-aligned mesh based
"""
def generate_vertices(self):
radius = self.dimensions[0]
height = self.dimensions[1]
segment_height = height / self.segments
number_of_sides = self.sides
initial_angle = pi / number_of_sides
angle_of_rotation = 2 * pi / number_of_sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(number_of_sides=number_of_sides,
initial_angle=initial_angle)
faces[0].offset_outer_vertices(offset=glm.vec3(self.outer_vertices_offset))
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=number_of_sides, face=faces[0])
# generate the later N faces via exude
base_faces = [faces[0]]
radius = 1.0
for index in range(1, self.segments + 1):
base_faces.append(Face())
if index < self.segments / 4:
radius *= 1.0 + 0.1 * index
else:
radius *= 0.9 # + .1*index
base_faces[index].extrude_from_other_face(other=base_faces[index - 1], direction=[0.0, 1.0, 0.0],
distance=segment_height, radius=radius)
base_faces[index].radius = radius
faces += self.stitch_faces(face_a=base_faces[index - 1], face_b=base_faces[index],
number_of_faces=number_of_sides)
if index == self.segments:
faces.append(base_faces[index])
vertices = []
# apply bevel cut
bevel_cut_faces = []
# face_to_bevel_cut = faces.pop(len(faces)-1)
# bevel_cut_faces = self.bevel_cut(face_to_bevel_cut)
# faces += bevel_cut_faces
faces[0].calculate_outer_vertices_as_vec3()
faces[0].flip_winding_order()
for index, face in enumerate(faces):
face_to_cut = faces[index]
bevel_cut_faces += self.bevel_cut(
face_to_cut,
bevel_depths=self.bevel_depths,
border_sizes=self.border_sizes,
depth=self.depth,
)
for face in bevel_cut_faces:
vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
class SegmentedPrismBevelPolygonCornerTest(SegmentedPrism):
"""
using our bevel cut function (bevel_polygon_corner) on a base polygon.
Also testing the divide_face_into_quads
"""
def generate_vertices(self):
radius = self.dimensions[0]
height = self.dimensions[1]
segment_height = height / self.segments
initial_angle = pi / self.sides
angle_of_rotation = 2 * pi / self.sides
vertices = []
faces = []
faces.append(Face())
faces[0].outer_vertices = self.generate_outer_vertices(
number_of_sides=self.sides,
initial_angle=initial_angle
)
faces[0].offset_outer_vertices(offset=glm.vec3(self.outer_vertices_offset))
"""NEW beveling on base polygon testing-working"""
for i in range(1):
faces[0] = self.bevel_polygon_corner(faces[0], subject_vertex_index=0, bevel_ratio=0.25)
for i in range(1):
faces[0] = self.bevel_polygon_corner(faces[0], subject_vertex_index=3, bevel_ratio=0.25)
faces[0].vertices = self.outer_vertices_to_vertices(number_of_sides=self.sides, face=faces[0])
# generate the later N faces via exude
base_faces = [faces[0]]
radius = 1.0
for index in range(1, self.segments + 1):
base_faces.append(Face())
if index < self.segments / 4:
radius *= 1.0 + 0.1 * index
else:
radius *= 0.9 # + .1*index
base_faces[index].extrude_from_other_face(other=base_faces[index - 1], direction=[0.0, 1.0, 0.0],
distance=segment_height, radius=radius)
base_faces[index].radius = radius
faces += self.stitch_faces(face_a=base_faces[index - 1], face_b=base_faces[index],
number_of_faces=self.sides)
if index == self.segments:
faces.append(base_faces[index])
"""NEW quad subdivide of faces function testing"""
faces_post_divide = []
faces[0].calculate_outer_vertices_as_vec3()
faces[0].flip_winding_order()
for face in faces:
# faces_post_divide += self.divide_face_into_quads(face, center_point_offset=0.2)
faces_post_divide += self.pyrimidize_face(face, center_point_offset=0.2)
faces = faces_post_divide
vertices = []
bevel_cut_faces = []
faces[0].calculate_outer_vertices_as_vec3()
faces[0].flip_winding_order()
for index, face in enumerate(faces):
face_to_cut = faces[index]
bevel_cut_faces += self.bevel_cut(
face_to_cut,
bevel_depths=self.bevel_depths,
border_sizes=self.border_sizes,
depth=self.depth,
)
for face in bevel_cut_faces:
vertices += face.face_to_list()
return np.array(vertices,
dtype=np.float32
).flatten()
def calculate_normal(vertices=[]):
"""
calculates the normal vector
:param vertices: a list of 3 glm vec3's
:return: a vec3 with the normal to this triangle
"""
return glm.normalize(glm.cross(vertices[1] - vertices[0], vertices[2] - vertices[0]))
def calculate_mean_point(points):
"""
Calculate the mean point of a polygon
:param points: list of points of a polygon as glm.vec3s
:return: mean, a glm.vec3
"""
sum_point = glm.vec3([0.0, 0.0, 0.0])
for point in points:
sum_point += point
return sum_point / len(points)
def list_to_vec3_list(current_list):
"""
convert a list of points (serialized fully, as a list) to a list of glm.vec3's
:param current_list:
:return:
"""
reformated_list = []
index = 0
while index < len(current_list):
reformated_list.append(
glm.vec3(
current_list[index],
current_list[index + 1],
current_list[index + 2],
)
)
index += 3
return reformated_list
def vec3_list_to_list(vec3_list):
"""
convert a list of points-as-vec3's to a serialized list
:return: a list
"""
serial_list = []
for point in vec3_list:
serial_list += list(point)
return serial_list
def bezier_linear(points=[], intervals=0):
"""
Linearly interpolate between two points
:param points: two points to interpolate between. glm.vec3
:param intervals: number of steps to take between point_0 and point_1
:return: a list of points (including start and end control points)
"""
bezier_points = points.copy()
for step in range(1, intervals + 1):
time = step * (1.0 / intervals)
new_point = points[0] + time * (points[1] - points[0])
bezier_points.insert(-1, new_point)
return bezier_points
def bezier_quadratic(points=[], intervals=0):
"""
Quadratically interpolate between two points
:param points: three points to interpolate between. glm.vec3
:param intervals: number of steps to take between point_0 and point_1
:return: a list of points
"""
bezier_points = [points[0], points[2]]
control_points = points.copy()
for step in range(1, intervals + 1):
time = step * (1.0 / (intervals + 1))
new_point = control_points[1] + pow((1.0 - time), 2) * (control_points[0] - control_points[1]) + pow(time, 2) * \
(control_points[2] - control_points[1])
bezier_points.insert(-1, new_point)
return bezier_points
def bezier_cubic(points=[], intervals=0):
"""
cubically interpolate between two points
:param points: three points to interpolate between. glm.vec3
:param intervals: number of steps to take between point_0 and point_1
:return: a list of points
EXAMPLE
bezier_points = [
glm.vec3(0.0, 10.0, 0.0),
glm.vec3(0.0, 25.0, 0.0),
glm.vec3(45.0, 25.0, 0.0),
glm.vec3(45.0, 10.0, 0.0)
]
bezier_points = primatives.bezier_cubic(points=bezier_points, intervals=20)
IN DRAW LOOP:
for point in bezier_points:
bezier_model.position = point
bezier_model.draw(view=view)
"""
bezier_points = [points[0], points[3]]
control_points = points.copy()
fractional_interval = 1.0 / (intervals + 1)
for step in range(1, intervals + 1):
time = step * (fractional_interval)
reverse_time = 1.0 - time
new_point = pow(reverse_time, 3) * points[0] + 3.0 * pow(reverse_time, 2) * time * points[1] + \
3.0 * reverse_time * pow(time, 2) * points[2] + pow(time, 3) * points[3]
bezier_points.insert(-1, new_point)
return bezier_points
def calculate_texture_coordinates_from_control_points(point, control_points):
"""
given the 4 control points, calculate the proper texture coord.
Assumes the 4 control points fall on exact corners of the texture
:return: glm.vec2(width, height), each in range [0,1]
"""
control_points = rotate_list(control_points, 2)
height = glm.distance(control_points[0], control_points[1])
width = glm.distance(control_points[1], control_points[2])
height_delta = calculate_triangle_height(points=[control_points[2], point, control_points[1]])
width_delta = calculate_triangle_height(points=[control_points[0], point, control_points[1]])
texture_coordinates = glm.vec2(
glm.clamp(1.0 - width_delta / width, 0.0, 1.0),
glm.clamp(height_delta / height, 0.0, 1.0)
)
return texture_coordinates
def calculate_triangle_height(points=[]):
"""
Given 3 points of a triangle, return the height
The points must be given counter-clockwise, with the first and third point forming the base
TODO: Refactor
:param points:
:return: the height of the triangle
"""
triangle_sides = triangle_points_to_side_lengths(points=points)
area = herons_formula(triangle_sides)
base = glm.distance(points[0], points[2])
height = 2.0 * area / base
return height
def triangle_points_to_side_lengths(points):
"""
Given the 3 points of a triangle, calculate the lengths of the 3 sides
:param points:
:return: list of sides
"""
sides = []
sides.append(glm.distance(points[0], points[1]))
sides.append(glm.distance(points[1], points[2]))
sides.append(glm.distance(points[2], points[0]))
return sides
def herons_formula(sides):
"""
Uses Heron's Formula to calculate the area of a triangle given the three sides
:param sides:
:return: Area of the triangle
"""
return 0.25 * math.sqrt(
4.0 * sides[0] * sides[0] * sides[1] * sides[1] - pow(sides[0] * sides[0] + sides[1] * sides[1] - \
sides[2] * sides[2], 2))
def rotate_list(current_list, steps):
"""
rotates a list by n steps. negative steps puts last element first.
EXAMPLE
fruits = ['a', 'b', 'c']
updated_fruits = rotate(fruits, -1)
updated_fruits is now ['c', 'b', 'a']
:param current_list: list to operate on. Original not changed
:param steps: number of elements to shift over
:return: a new list
"""
return current_list[steps:] + current_list[:steps]
def get_next_index(size, index):
if index + 1 > size - 1:
return 0
else:
return index + 1
def get_previous_index(size, index):
if index - 1 < 0:
return size - 1
else:
return index - 1
| 150,202 | Python | .py | 3,084 | 37.401102 | 131 | 0.586866 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,321 | MeshWorkshop.py | AlexSanfilippo_ProceduralMeshGeneration/MeshWorkshop.py | """
Main script of the spaceship generator. run this to launch the OpenGL window.
Contains high-level code pertaining creation and drawing of:
Model
GUI
Skybox
As well as OpenGL callback functions
"""
import glfw
from OpenGL.GL import *
from OpenGL.GL.shaders import compileProgram, compileShader
import pyrr
from TextureLoader import load_texture
from Camera import Camera, SimulationCamera
from math import pi
import random as r
import glm
from PIL import Image, ImageOps
import PointLightCube as plc
from ProceduralMesh import Spaceship3x3
import FPSCounter
from GUI import GUI
import Skybox as CubeMapSkybox
from tkinter import filedialog
import gc
DEV_BUILD = True
images = []
follow_cam = SimulationCamera(camera_pos=[6.0, 247.0, 242.0])
cam = Camera(camera_pos=[0.0, 20.0, 20.0])
use_follow_cam = False
def capture_screenshot(filename='screenshot.jpg'):
print(f'saving screenshot as {filename}')
fbo_array = glReadPixels(0, 0, WIDTH, HEIGHT, GL_RGB, GL_UNSIGNED_BYTE)
screenshot = Image.frombytes("RGB", (WIDTH, HEIGHT), fbo_array)
screenshot = ImageOps.flip(screenshot)
screenshot.save(
filename,
)
WIDTH, HEIGHT = 1728, 972
lastX, lastY = WIDTH / 2, HEIGHT / 2
first_mouse = True
left, right, forward, backward, make_new_surface = False, False, False, False, False
player_left, player_right, player_forward, player_backward = False, False, False, False
yaw_counterclockwise, yaw_clockwise = False, False
up, down = False, False
write_to_gif = False
wrote_to_gif = False
make_new_ship = False
pause = False
ship_texture_cycle_id = 0
# the keyboard input callback
def key_input_clb(window, key, scancode, action, mode):
global left, right, forward, backward, make_new_surface, player_left, player_right, player_forward, \
player_backward, yaw_counterclockwise, yaw_clockwise, write_to_gif, make_new_ship,\
pause, ship_texture_cycle_id, up, down, wrote_to_gif
if key == glfw.KEY_ESCAPE and action == glfw.PRESS:
glfw.set_window_should_close(window, True)
if key == glfw.KEY_W and action == glfw.PRESS:
forward = True
elif key == glfw.KEY_W and action == glfw.RELEASE:
forward = False
if key == glfw.KEY_S and action == glfw.PRESS:
backward = True
elif key == glfw.KEY_S and action == glfw.RELEASE:
backward = False
if key == glfw.KEY_A and action == glfw.PRESS:
left = True
elif key == glfw.KEY_A and action == glfw.RELEASE:
left = False
if key == glfw.KEY_D and action == glfw.PRESS:
right = True
elif key == glfw.KEY_D and action == glfw.RELEASE:
right = False
if key == glfw.KEY_Q and action == glfw.PRESS:
yaw_clockwise = True
elif key == glfw.KEY_Q and action == glfw.RELEASE:
yaw_clockwise = False
if key == glfw.KEY_E and action == glfw.PRESS:
yaw_counterclockwise = True
elif key == glfw.KEY_E and action == glfw.RELEASE:
yaw_counterclockwise = False
if key == glfw.KEY_TAB and action == glfw.PRESS:
up = True
elif key == glfw.KEY_TAB and action == glfw.RELEASE:
up = False
if key == glfw.KEY_LEFT_SHIFT and action == glfw.PRESS:
down = True
elif key == glfw.KEY_LEFT_SHIFT and action == glfw.RELEASE:
down = False
if key == glfw.KEY_1 and action == glfw.PRESS:
glPolygonMode(GL_FRONT_AND_BACK, GL_LINE)
if key == glfw.KEY_2 and action == glfw.PRESS:
glPolygonMode(GL_FRONT_AND_BACK, GL_FILL)
if key == glfw.KEY_9 and action == glfw.PRESS:
if DEV_BUILD:
write_to_gif = not write_to_gif
wrote_to_gif = True
if key == glfw.KEY_SPACE and action == glfw.PRESS:
generate_next_ship()
if key == glfw.KEY_P and action == glfw.PRESS:
pause = not pause
if key == glfw.KEY_T and action == glfw.PRESS:
cycle_ship_texture()
if key == glfw.KEY_V and action == glfw.PRESS:
switch_camera_mode()
if key == glfw.KEY_K and action == glfw.PRESS:
if DEV_BUILD:
print_camera_position()
if key == glfw.KEY_F12 and action == glfw.PRESS:
capture_screenshot()
def cycle_ship_texture(forward=True):
global ship_texture_cycle_id
MIN = 0
MAX = 4
if forward:
ship_texture_cycle_id = min(ship_texture_cycle_id + 1, MAX)
else:
ship_texture_cycle_id = max(ship_texture_cycle_id - 1, MIN)
if ship_texture_cycle_id == 0:
spaceship_parameters['diffuse'] = texture_dictionary['penguin_diffuse']
spaceship_parameters['specular'] = texture_dictionary['penguin_specular']
spaceship_parameters['emission'] = texture_dictionary['penguin_emission']
update_spaceship_texture()
elif ship_texture_cycle_id == 1:
spaceship_parameters['diffuse'] = texture_dictionary['ship_a_diffuse']
spaceship_parameters['specular'] = texture_dictionary['ship_a_specular']
spaceship_parameters['emission'] = texture_dictionary['ship_a_emission']
update_spaceship_texture()
elif ship_texture_cycle_id == 2:
spaceship_parameters['diffuse'] = texture_dictionary['whoa_diffuse']
spaceship_parameters['specular'] = texture_dictionary['whoa_specular']
spaceship_parameters['emission'] = texture_dictionary['atlas_debug_emission']
update_spaceship_texture()
elif ship_texture_cycle_id == 3:
spaceship_parameters['diffuse'] = texture_dictionary['blue_metal_diffuse']
spaceship_parameters['specular'] = texture_dictionary['blue_metal_specular']
spaceship_parameters['emission'] = texture_dictionary['blue_metal_emission']
update_spaceship_texture()
elif ship_texture_cycle_id == 4:
spaceship_parameters['diffuse'] = texture_dictionary['atlas_debug_diffuse']
spaceship_parameters['specular'] = texture_dictionary['atlas_debug_specular']
spaceship_parameters['emission'] = texture_dictionary['atlas_debug_emission']
update_spaceship_texture()
def mouse_button_callback(window, button, action, mods):
global make_new_ship
left_click = button == glfw.MOUSE_BUTTON_LEFT and action == glfw.PRESS
right_click = button == glfw.MOUSE_BUTTON_RIGHT and action == glfw.PRESS
if button == left_click:
mpos = glfw.get_cursor_pos(window)
gui.button_update(position_mouse=glfw.get_cursor_pos(window), left_click=left_click, right_click=right_click)
def update_spaceship_texture():
meshes[0].set_diffuse(spaceship_parameters['diffuse'])
meshes[0].set_specular(spaceship_parameters['specular'])
meshes[0].set_emission(spaceship_parameters['emission'])
def do_movement(speed=1.0):
"""
do the camera movement, call this function in the main loop
:param speed:
:return:
"""
if left:
active_camera.process_keyboard("LEFT", speed)
if right:
active_camera.process_keyboard("RIGHT", speed)
if forward:
active_camera.process_keyboard("FORWARD", speed)
if backward:
active_camera.process_keyboard("BACKWARD", speed)
if yaw_clockwise:
follow_cam.process_keyboard("YAW_CLOCKWISE", speed)
if yaw_counterclockwise:
follow_cam.process_keyboard("YAW_COUNTERCLOCKWISE", speed)
if up:
active_camera.process_keyboard("UP", speed)
if down:
active_camera.process_keyboard("DOWN", speed)
def mouse_look_clb(window, xpos, ypos):
global first_mouse, lastX, lastY
if first_mouse:
lastX = xpos
lastY = ypos
first_mouse = False
xoffset = xpos - lastX
yoffset = lastY - ypos
lastX = xpos
lastY = ypos
cam.process_mouse_movement(xoffset, yoffset)
def scroll_callback(window, xoffset, yoffset):
follow_cam.process_mouse_scroll(xoffset, yoffset)
vertex_src = """
# version 330
layout(location = 0) in vec3 a_position;
layout(location = 1) in vec2 a_texture;
layout(location = 2) in vec3 a_normal;
uniform mat4 model;
uniform mat4 projection;
uniform mat4 view;
out vec2 tex_coords; //texture coordinates
out vec3 normal;
out vec3 frag_pos;
void main()
{
tex_coords = a_texture;
frag_pos = vec3(model * vec4(a_position, 1.0));
normal = mat3(transpose(inverse(model))) * a_normal;
gl_Position = projection * view * vec4(frag_pos, 1.0);
}
"""
fragment_src = """
#version 330 core
out vec4 frag_color;
struct Material {
sampler2D diffuse;
sampler2D specular;
sampler2D emission;
float shininess;
};
struct DirLight {
vec3 direction;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
struct PointLight {
vec3 position;
float constant;
float linear;
float quadratic;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
struct SpotLight {
vec3 position;
vec3 direction;
float cut_off;
float outer_cut_off;
float constant;
float linear;
float quadratic;
vec3 ambient;
vec3 diffuse;
vec3 specular;
};
#define NR_POINT_LIGHTS 2
in vec3 frag_pos;
in vec3 normal;
in vec2 tex_coords;
uniform vec3 view_pos;
uniform DirLight dir_light;
uniform PointLight point_lights[NR_POINT_LIGHTS];
uniform SpotLight spot_light;
uniform Material material;
// function prototypes
vec3 CalcDirLight(DirLight light, vec3 normal, vec3 view_dir, vec2 tex_coords);
vec3 CalcPointLight(PointLight light, vec3 normal, vec3 frag_pos, vec3 view_dir, vec2 tex_coords);
vec3 CalcSpotLight(SpotLight light, vec3 normal, vec3 frag_pos, vec3 view_dir);
void main()
{
vec2 tex_coords_alternate = tex_coords;
/*
float q_coord = 0.1f;
if (tex_coords_alternate.x > 0.9f && tex_coords_alternate.y > 0.9f){
tex_coords_alternate = tex_coords_alternate * q_coord;
}
if (tex_coords_alternate == vec2(0.f, 1.f)){
tex_coords_alternate = tex_coords_alternate * q_coord;
}
*/
// properties
vec3 norm = normalize(normal);
vec3 view_dir = normalize(view_pos - frag_pos);
// == =====================================================
// Our lighting is set up in 3 phases: directional, point lights and an optional flashlight
// For each phase, a calculate function is defined that calculates the corresponding color
// per lamp. In the main() function we take all the calculated colors and sum them up for
// this fragment's final color.
// == =====================================================
vec3 result;
// phase 1: directional lighting
result = CalcDirLight(dir_light, norm, view_dir, tex_coords_alternate);
// phase 2: point lights
for(int i = 0; i < NR_POINT_LIGHTS; i++)
result += CalcPointLight(point_lights[i], norm, frag_pos, view_dir, tex_coords_alternate);
// phase 3: spotlight
//result += CalcSpotLight(spot_light, norm, frag_pos, view_dir);
// emission
vec3 emission = texture(material.emission, tex_coords_alternate).rgb;
result = result + emission;
frag_color = vec4(result, 1.0);
}
// calculates the color when using a directional light.
vec3 CalcDirLight(DirLight light, vec3 normal, vec3 view_dir, vec2 tex_coords)
{
vec3 light_dir = normalize(-light.direction);
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
return (ambient + diffuse + specular);
}
// calculates the color when using a point light.
vec3 CalcPointLight(PointLight light, vec3 normal, vec3 frag_pos, vec3 view_dir, vec2 tex_coords)
{
vec3 light_dir = normalize(light.position - frag_pos);
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// attenuation
float distance = length(light.position - frag_pos);
float attenuation = 1.0 / (light.constant + light.linear * distance + light.quadratic * (distance * distance));
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
ambient *= attenuation;
diffuse *= attenuation;
specular *= attenuation;
return (ambient + diffuse + specular);
}
// calculates the color when using a spot light.
vec3 CalcSpotLight(SpotLight light, vec3 normal, vec3 frag_pos, vec3 view_dir)
{
vec3 light_dir = normalize(light.position - frag_pos);
//view_dir = vec3(0.0, -1.0, 0.0); //tp
// diffuse shading
float diff = max(dot(normal, light_dir), 0.0);
// specular shading
vec3 reflect_dir = reflect(-light_dir, normal);
float spec = pow(max(dot(view_dir, reflect_dir), 0.0), material.shininess);
// attenuation
float distance = length(light.position - frag_pos);
float attenuation = 1.0 / (light.constant + light.linear * distance + light.quadratic * (distance * distance));
// spotlight intensity
float theta = dot(light_dir, normalize(-light.direction));
float epsilon = light.cut_off - light.outer_cut_off;
float intensity = clamp((theta - light.outer_cut_off) / epsilon, 0.0, 1.0);
// combine results
vec3 ambient = light.ambient * vec3(texture(material.diffuse, tex_coords));
vec3 diffuse = light.diffuse * diff * vec3(texture(material.diffuse, tex_coords));
vec3 specular = light.specular * spec * vec3(texture(material.specular, tex_coords));
ambient *= attenuation * intensity;
diffuse *= attenuation * intensity;
specular *= attenuation * intensity;
return (ambient + diffuse + specular);
}
"""
# the window resize callback function
def window_resize_clb(window, width, height):
glViewport(0, 0, width, height)
projection = pyrr.matrix44.create_perspective_projection_matrix(45, width / height, 0.1, 2000)
glUniformMatrix4fv(proj_loc, 1, GL_FALSE, projection)
gui.set_screen_size(screen_size=(width, height))
# initializing glfw library
if not glfw.init():
raise Exception("glfw can not be initialized!")
# creating the window
window = glfw.create_window(WIDTH, HEIGHT, "Spaceship Generator", None, None)
# check if window was created
if not window:
glfw.terminate()
raise Exception("glfw window can not be created!")
# set window's position
glfw.set_window_pos(window, 40, 40)
# set the callback function for window resize
glfw.set_window_size_callback(window, window_resize_clb)
# set the mouse position callback
glfw.set_cursor_pos_callback(window, mouse_look_clb)
# set the keyboard input callback
glfw.set_key_callback(window, key_input_clb)
#set the scroll-wheel input callback
glfw.set_scroll_callback(window, scroll_callback)
#set the mouse callback
glfw.set_mouse_button_callback(window, mouse_button_callback)
# capture the mouse cursor
# glfw.set_input_mode(window, glfw.CURSOR, glfw.CURSOR_DISABLED)
# glfw.set_input_mode(window, glfw.CURSOR, glfw.CURSOR_CAPTURED)
# make the context current
glfw.make_context_current(window)
def switch_camera_mode():
global use_follow_cam, active_camera
use_follow_cam = not use_follow_cam
if use_follow_cam:
glfw.set_input_mode(window, glfw.CURSOR, glfw.CURSOR_NORMAL) #glfw.CURSOR_CAPTURED,
active_camera = follow_cam
else:
glfw.set_input_mode(window, glfw.CURSOR, glfw.CURSOR_DISABLED)
active_camera = cam
switch_camera_mode()
def print_camera_position():
global active_cam
print(f'Camera Position: {active_camera.camera_pos}')
shader = compileProgram(compileShader(vertex_src, GL_VERTEX_SHADER), compileShader(fragment_src, GL_FRAGMENT_SHADER))
#uncomment to enable backfase culling
glEnable(GL_CULL_FACE)
#uncomment to see that cull is working by culling front faces rather than back
# glCullFace(GL_BACK)
# glFrontFace(GL_CW)
glUseProgram(shader)
# set the texture unit (integer) of each sampler2D
glUniform1i(glGetUniformLocation(shader, "material.diffuse"), 0)
glUniform1i(glGetUniformLocation(shader, "material.specular"), 1)
glUniform1i(glGetUniformLocation(shader, "material.emission"), 2)
glClearColor(0, 0.1, 0.1, 1)
glEnable(GL_DEPTH_TEST)
glEnable(GL_BLEND)
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA)
projection_2 = pyrr.matrix44.create_perspective_projection_matrix(45, WIDTH / HEIGHT, 0.1, 2000)
model_loc = glGetUniformLocation(shader, "model")
proj_loc = glGetUniformLocation(shader, "projection")
view_loc = glGetUniformLocation(shader, "view")
view_pos_loc = glGetUniformLocation(shader, "view_pos") # cam pos for specular lighting
glUniformMatrix4fv(proj_loc, 1, GL_FALSE, projection_2)
"""direction light settings"""
direction_diffuse = [0.0]*3
direction_ambient = [0.0]*3
direction_specular = [0.0]*3
# create the light cube
my_plc = plc.PointLightCube(pos=[0.0, 40.0, 0.0], ambient=[0.1]*3, diffuse=[0.9]*3,
specular=[0.9]*3, linear=0.0014, quadratic=.000007)
debug_plcs = [my_plc]
move_whole_compass_z = -80
debug_plcs.append(plc.PointLightCube(
pos=[0.0, 10.0, move_whole_compass_z+5.0],
ambient=[0.0]*3,
diffuse=[0.0]*3,
specular=[0.0]*3
))
debug_plcs.append(plc.PointLightCube(
pos=[0.0, 5.0, move_whole_compass_z+10.0],
ambient=[0.0, 0.0, 1.0],
diffuse=[0.0, 0.0, 1.0],
specular=[0.0, 0.0, 1.0],
))
debug_plcs.append(plc.PointLightCube(
pos=[0.0, 5.0, move_whole_compass_z+0.0],
ambient=[1.0, 0.0, 0.0],
diffuse=[1.0, 0.0, 0.0],
specular=[1.0, 0.0, 0.0],
))
debug_plcs.append(plc.PointLightCube(
pos=[5.0, 5.0, move_whole_compass_z+5.0],
ambient=[0.0, 1.0, 0.0],
diffuse=[0.0, 1.0, 0.0],
specular=[0.0, 1.0, 0.0],
))
debug_plcs.append(plc.PointLightCube(
pos=[-5.0, 5.0, move_whole_compass_z+5.0],
ambient=[1.0, 1.0, 0.0],
diffuse=[1.0, 1.0, 0.0],
specular=[1.0, 1.0, 0.0],
))
seed = 1946
r.seed(seed)
# direction light pass to shader
glUniform3fv(glGetUniformLocation(shader, "dir_light.position"), 1, [0, 20, 10])
glUniform3fv(glGetUniformLocation(shader, "dir_light.diffuse"), 1, direction_diffuse)
glUniform3fv(glGetUniformLocation(shader, "dir_light.ambient"), 1, direction_ambient)
glUniform3fv(glGetUniformLocation(shader, "dir_light.specular"), 1, direction_specular)
textures = glGenTextures(25)
load_texture("Textures/photo_texture_atlas_diffuse_3x3.png", textures[0])
load_texture("Textures/photo_texture_atlas_specular_3x3.png", textures[1])
load_texture("Textures/photo_texture_atlas_emission.png", textures[2])
load_texture("Textures/debug_quad_red.png", textures[3])
load_texture("Textures/penguin_atlas_specular.png", textures[4])
load_texture("Textures/penguin_atlas_specular.png", textures[5])
load_texture("Textures/debug_texture_atlas.png", textures[6])
load_texture("Textures/blue_gold_alt_diffuse.png", textures[7])
load_texture("Textures/blue_gold_specular.png", textures[8])
load_texture("Textures/no_emission.png", textures[9])
load_texture("Textures/debug_diffuse_coordinates.png", textures[10])
load_texture("Textures/penguin_atlas_emission.png", textures[11])
load_texture("Textures/penguin_atlas_specular.png", textures[12])
load_texture("Textures/whoa_atlas_diffuse_3x3.png", textures[13])
load_texture("Textures/whoa_atlas_specular_3x3.png", textures[14])
load_texture("Textures/penguin_atlas_emission.png", textures[15])
load_texture("Fonts/my_font.png", textures[16])
load_texture("Textures/button_atlas_gradient.png", textures[17])
load_texture("Textures/dark_metal_diffuse.png", textures[18])
load_texture("Textures/dark_metal_specular.png", textures[19])
load_texture("Textures/dark_metal_emission.png", textures[20])
load_texture("Textures/SkyboxLearnOpenGL.png", textures[21])
texture_dictionary = {
"pink": textures[3],
"penguin_diffuse": textures[0],
"penguin_specular": textures[1],
"penguin_emission": textures[2],
"blue_metal_diffuse": textures[7],
"blue_metal_specular": textures[8],
"blue_metal_emission": textures[9],
"atlas_debug_diffuse": textures[10],
"atlas_debug_emission": textures[11],
"atlas_debug_specular": textures[12],
"whoa_diffuse": textures[13],
"whoa_specular": textures[14],
"whoa_emission": textures[15],
"font_atlas": textures[16],
"button_atlas": textures[17],
"ship_a_diffuse": textures[18],
"ship_a_specular": textures[19],
"ship_a_emission": textures[20],
"skybox_nebula": textures[21],
}
debug_ship_orders = [
['stuck']
]
spaceship_parameters = {
'number_of_sides': 8,
'number_of_segments': 5,
'transform_x': 1.0,
'transform_z': 2.0,
'scale': 3.3,
'diffuse': texture_dictionary['penguin_diffuse'],
'specular': texture_dictionary['penguin_specular'],
'emission': texture_dictionary['penguin_emission'],
'position': [0.0, -50.0, 0.0],
'length_of_segment': 10,
}
def update_spaceship_parameters(key, value):
spaceship_parameters[key] = value
def increase_light_brightness():
ambient_current = my_plc.ambient[0]
ambient_current = min(ambient_current + 0.1, 1.0)
my_plc.ambient = [ambient_current] * 3
def decrease_light_brightness():
ambient_current = my_plc.ambient[0]
ambient_current = max(ambient_current - 0.1, 0.0)
my_plc.ambient = [ambient_current] * 3
def increase_specular_brightness():
specular_current = my_plc.specular[0]
specular_current = min(specular_current + 0.1, 1.0)
my_plc.specular = [specular_current] * 3
def decrease_specular_brightness():
specular_current = my_plc.specular[0]
specular_current = max(specular_current - 0.1, 0.0)
my_plc.specular = [specular_current] * 3
def increase_diffuse_brightness():
diffuse_current = my_plc.diffuse[0]
diffuse_current = min(diffuse_current + 0.1, 1.0)
my_plc.diffuse = [diffuse_current] * 3
def decrease_diffuse_brightness():
diffuse_current = my_plc.diffuse[0]
diffuse_current = max(diffuse_current - 0.1, 0.0)
my_plc.diffuse = [diffuse_current] * 3
def increase_ship_sides(display):
MAX_SIDES = 10
sides = spaceship_parameters['number_of_sides']
if sides == MAX_SIDES:
return
sides += 1
update_spaceship_parameters(key='number_of_sides', value=sides)
generate_new_ship()
display.update_text(text=str(int(display.text_box.text) + 1))
def decrease_ship_sides(display):
MIN_SIDES = 3
sides = spaceship_parameters['number_of_sides']
if sides == MIN_SIDES:
return
sides -= 1
update_spaceship_parameters(key='number_of_sides', value=sides)
generate_new_ship()
display.update_text(text=str(int(display.text_box.text) - 1))
def increase_ship_segments(display=None):
MAX_SEGMENTS = 20
segments = spaceship_parameters['number_of_segments']
if segments == MAX_SEGMENTS:
return
segments += 1
update_spaceship_parameters(key='number_of_segments', value=segments)
generate_new_ship()
display.update_text(text=str(int(display.text_box.text) + 1))
def decrease_ship_segments(display=None):
MIN_SEGMENTS = 1
segments = spaceship_parameters['number_of_segments']
if segments == MIN_SEGMENTS:
return
segments -= 1
update_spaceship_parameters(key='number_of_segments', value=segments)
generate_new_ship()
display.update_text(text=str(int(display.text_box.text) - 1))
def increase_ship_segment_length(display=None):
MAX_LENGTH = 60.0
length = spaceship_parameters['length_of_segment']
if length >= MAX_LENGTH:
return
delta= 0.5
length += delta
update_spaceship_parameters(key='length_of_segment', value=length)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) + delta, 2)))
def decrease_ship_segment_length(display=None):
MIN_LENGTH = 4.5
length = spaceship_parameters['length_of_segment']
print('length=', length)
if length <= MIN_LENGTH:
return
delta = 0.5
length -= delta
update_spaceship_parameters(key='length_of_segment', value=length)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) - delta, 2)))
def increase_ship_x_scaling(display=None):
MAX_SCALE = 10.0
x_scale = spaceship_parameters['transform_x']
if x_scale >= MAX_SCALE:
return
delta = .1
x_scale += delta
update_spaceship_parameters(key='transform_x', value=x_scale)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) + delta, 2)))
def decrease_ship_x_scaling(display=None):
MIN_SCALE = 0.1
x_scale = spaceship_parameters['transform_x']
if x_scale <= MIN_SCALE:
return
delta = .1
x_scale -= delta
update_spaceship_parameters(key='transform_x', value=x_scale)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) - delta, 2)))
def increase_ship_z_scaling(display=None):
MAX_SCALE = 10.0
z_scale = spaceship_parameters['transform_z']
if z_scale >= MAX_SCALE:
return
delta = .1
z_scale += delta
update_spaceship_parameters(key='transform_z', value=z_scale)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) + delta, 2)))
def decrease_ship_z_scaling(display=None):
MIN_SCALE = 0.1
z_scale = spaceship_parameters['transform_z']
if z_scale <= MIN_SCALE:
return
delta = .1
z_scale -= delta
update_spaceship_parameters(key='transform_z', value=z_scale)
generate_new_ship()
display.update_text(text=str(round(float(display.text_box.text) - delta, 2)))
ships = []
num_ships = 1
for ship in range(num_ships):
spaceship = Spaceship3x3(
shader=shader,
diffuse=spaceship_parameters['diffuse'],
specular=spaceship_parameters['specular'],
emission=spaceship_parameters['emission'],
dimensions=[5.0, 5.0],
position=spaceship_parameters['position'],
rotation_magnitude=[0.0, 0.0, -pi*0.5],
number_of_sides=spaceship_parameters['number_of_sides'],
number_of_segments=spaceship_parameters['number_of_segments'],
transform_x=spaceship_parameters['transform_x'],
transform_z=spaceship_parameters['transform_z'],
length_of_segment=spaceship_parameters['length_of_segment'],
radius=3.0,
scale=spaceship_parameters['scale'],
seed=seed
)
ships.append(spaceship)
my_fps = FPSCounter.FPSCounter(frame_interval=300.0, mute=True)
meshes = []
meshes += ships
def increase_seed():
global seed
seed += 1
def decrease_seed():
global seed
seed -= 1
def generate_next_ship(display=None):
global seed
increase_seed()
generate_new_ship()
if display:
display.update_text(text='seed: ' + str(seed))
def generate_previous_ship(display=None):
global seed
decrease_seed()
generate_new_ship()
if display:
display.update_text(text='seed: ' + str(seed))
def generate_new_ship():
global seed
global meshes
spaceship_next = Spaceship3x3(
shader=shader,
diffuse=spaceship_parameters['diffuse'],
specular=spaceship_parameters['specular'],
emission=spaceship_parameters['emission'],
dimensions=[5.0, 5.0],
position=ships[0].position,
rotation_magnitude=ships[0].rotation_magnitude,
rotation_axis=glm.vec3((0.0, 0.0, 1.0)),
number_of_sides=spaceship_parameters['number_of_sides'],
number_of_segments=spaceship_parameters['number_of_segments'],
transform_x=spaceship_parameters['transform_x'],
transform_z=spaceship_parameters['transform_z'],
length_of_segment=spaceship_parameters['length_of_segment'],
radius=3.0,
scale=spaceship_parameters['scale'],
seed=seed
)
meshes[0].clean_up()
del meshes[0]
gc.collect()
meshes.append(spaceship_next)
"""GUI CREATION"""
gui = GUI(screen_size=(WIDTH, HEIGHT))
gui.add_text_element(
shader=None,
position=(0.75, 0.73),
scale=(0.20, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='button',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Texture',
)
gui.add_text_button(
shader=None,
position=(0.95, 0.73),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=cycle_ship_texture,
context_id='button',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(0.5, 0.73),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=cycle_ship_texture,
forward=False,
context_id='button',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_button(
shader=None,
position=(0.88, 0.95),
scale=(0.12, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=gui.toggle_context_status,
context_id='context_on',
context_status=True,
click_function_context_id='button',
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Options',
)
gui.add_text_element(
shader=None,
position=(-0.25, 0.83),
scale=(0.08, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text=str(spaceship_parameters['number_of_sides']),
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(-0.38, 0.83),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_ship_sides,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(-0.48, 0.83),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_ship_sides,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_element(
shader=None,
# position=(-0.5, -0.0),
position=(-0.78, 0.83), #-.28, +0.83
scale=(0.25, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Sides',
)
gui.add_text_element(
shader=None,
position=(-0.25, 0.73),
scale=(0.08, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text=str(spaceship_parameters['number_of_segments']),
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(-0.48, .73),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_ship_segments,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_button(
shader=None,
position=(-0.38, 0.73),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_ship_segments,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_element(
shader=None,
position=(-0.78, 0.73),
scale=(0.25, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Segments',
)
gui.add_text_element(
shader=None,
position=(-0.25, 0.63),
scale=(0.08, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='10.0',
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_element(
shader=None,
position=(-0.78, 0.63),
scale=(0.25, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Length Scale',
)
gui.add_text_button(
shader=None,
position=(-0.38, 0.63),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_ship_segment_length,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(-0.48, 0.63),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_ship_segment_length,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_element(
shader=None,
position=(-0.78, 0.53),
scale=(0.25, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Height Scale',
)
gui.add_text_element(
shader=None,
position=(-0.25, 0.53),
scale=(0.08, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='1.0',
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(-0.48, 0.53),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_ship_x_scaling,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_button(
shader=None,
position=(-0.38, 0.53),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_ship_x_scaling,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_element(
shader=None,
position=(-0.78, 0.42999999999999994),
scale=(0.25, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Width Scale',
)
gui.add_text_element(
shader=None,
position=(-0.25, 0.42999999999999994),
scale=(0.08, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='ship_settings',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='2.0',
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(-0.48, 0.42999999999999994),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_ship_z_scaling,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_button(
shader=None,
position=(-0.38, 0.42999999999999994),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_ship_z_scaling,
display=displayer,
context_id='ship_settings',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_element(
shader=None,
position=(-0.90, 0.95),
scale=(0.15, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='fps',
context_status=True,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='100',
)
#todo refactor this. Need way to access text/element we want to update
display_fps = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(-0.55, 0.95),
scale=(0.22, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=gui.toggle_context_status,
context_id='context_on',
context_status=True,
click_function_context_id='ship_settings',
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Ship Settings',
)
gui.add_text_element(
shader=None,
position=(0.75, 0.83),
scale=(0.20, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='button',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Seed: ' + str(seed),
)
#todo refactor this. Need way to access text/element we want to update
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(0.95, 0.83),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=generate_next_ship,
display=displayer,
context_id='button',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(0.50, 0.83),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=generate_previous_ship,
display=displayer,
context_id='button',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
gui.add_text_button(
shader=None,
position=(-0.185, 0.95),
scale=(0.15, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=gui.toggle_context_status,
context_id='context_on',
context_status=True,
click_function_context_id='light',
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Light',
)
gui.add_text_element(
shader=None,
position=(0.75, 0.52),
scale=(0.20, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='light',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Ambient',
)
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(0.95, 0.52),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_light_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(0.50, 0.52),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_light_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
#Specular
gui.add_text_element(
shader=None,
position=(0.75, 0.42),
scale=(0.20, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='light',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Specular',
)
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(0.95, 0.42),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_specular_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(0.50, 0.42),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_specular_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
#Diffuse
gui.add_text_element(
shader=None,
position=(0.75, 0.32),
scale=(0.20, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=2,
context_id='light',
context_status=False,
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Diffuse',
)
displayer = gui.elements[-1]
gui.add_text_button(
shader=None,
position=(0.95, 0.32),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=increase_diffuse_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='+',
)
gui.add_text_button(
shader=None,
position=(0.50, 0.32),
scale=(0.05, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=decrease_diffuse_brightness,
context_id='light',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='-',
)
def export_ship():
"""
Open Window to export the ship model as a wavefront obj file with name of choice
"""
filename = filedialog.asksaveasfilename()
meshes[0].export_as_obj(filename=filename)
gui.add_text_button(
shader=None,
position=(0.11, 0.95),
scale=(0.15, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=export_ship,
context_id='context_on',
context_status=True,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Export',
)
cubemaps = {
2: [
"Textures/Skyboxes/mountain_lake/right.jpg",
"Textures/Skyboxes/mountain_lake/left.jpg",
"Textures/Skyboxes/mountain_lake/top.jpg",
"Textures/Skyboxes/mountain_lake/bottom.jpg",
"Textures/Skyboxes/mountain_lake/front.jpg",
"Textures/Skyboxes/mountain_lake/back.jpg",
],
0: [
"Textures/Skyboxes/galaxy/galaxy+X.png",
"Textures/Skyboxes/galaxy/galaxy-X.png",
"Textures/Skyboxes/galaxy/galaxy+Y.png",
"Textures/Skyboxes/galaxy/galaxy-Y.png",
"Textures/Skyboxes/galaxy/galaxy+Z.png",
"Textures/Skyboxes/galaxy/galaxy-Z.png",
],
1: [
"Textures/Skyboxes/nebula/skybox_left.png",
"Textures/Skyboxes/nebula/skybox_right.png",
"Textures/Skyboxes/nebula/skybox_up.png",
"Textures/Skyboxes/nebula/skybox_down.png",
"Textures/Skyboxes/nebula/skybox_front.png",
"Textures/Skyboxes/nebula/skybox_back.png",
]
}
skybox_cube_map = CubeMapSkybox.Skybox(
texture_paths=cubemaps[0],
scale=8000,
)
skybox_index = 0
def change_skybox():
global skybox_cube_map
global skybox_index
MAX_SKYBOX_INDEX = 2
skybox_index += 1
if skybox_index > MAX_SKYBOX_INDEX:
skybox_index = 0
skybox_cube_map = CubeMapSkybox.Skybox(
texture_paths=cubemaps[skybox_index],
scale=8000,
)
gui.add_text_button(
shader=None,
position=(0.75, 0.63),
scale=(0.3, 0.05),
texture=texture_dictionary['button_atlas'],
atlas_size=2,
atlas_coordinate=(2, 1),
click_function=change_skybox,
context_id='button',
context_status=False,
color=(1.0, 1.0, 1.0, 1.0),
font_texture=texture_dictionary['font_atlas'],
font_size=0.35,
font_color=(1.0, 1.0, 1.0, 1.0),
text='Skybox',
)
#must call as final setup of GUI
gui.build_elements_list()
#decouple fps from camera movement with time delta
time_start = 0
time_end = 1/60
time_delta = 1.0
projection = pyrr.matrix44.create_perspective_projection_matrix(45, WIDTH / HEIGHT, 0.1, 20050)
glUniformMatrix4fv(proj_loc, 1, GL_FALSE, projection)
if not DEV_BUILD:
display_fps.update_text(text='v1.0')
def write_fbo_to_gif():
global data, image
data = glReadPixels(0, 0, WIDTH, HEIGHT, GL_RGB, GL_UNSIGNED_BYTE)
image = Image.frombytes("RGB", (WIDTH, HEIGHT), data)
image = ImageOps.flip(image)
images.append(image)
while not glfw.window_should_close(window):
glfw.poll_events()
time_start = glfw.get_time()
do_movement(speed=100 * (time_delta))
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT)
fps = my_fps.update()
# if DEV_BUILD:
# display_fps.update_text(text=str(round(my_fps.get_fps())))
view = active_camera.get_view_matrix()
glUniformMatrix4fv(view_loc, 1, GL_FALSE, view)
for mesh in meshes:
mesh.draw(view=view)
#GUI hover events
#todo: mouse hover is maybe expensive! Consider spatial partition
if use_follow_cam:
gui.button_update(position_mouse=glfw.get_cursor_pos(window), left_click=False, right_click=False)
skybox_cube_map.draw(view=view, projection=projection)
glUseProgram(shader)
# pass cam position for specular light
glUniform3fv(view_pos_loc, 1, list(active_camera.camera_pos))
point_light_position_loc = glGetUniformLocation(shader, "point_light.position")
glUniform1f(glGetUniformLocation(shader, "material.shininess"), 4.0)
glUniform3fv(glGetUniformLocation(shader, "point_lights[0].position"), 1, debug_plcs[0].get_pos())
glUniform3fv(glGetUniformLocation(shader, "point_lights[0].diffuse"), 1, debug_plcs[0].get_diffuse())
glUniform3fv(glGetUniformLocation(shader, "point_lights[0].ambient"), 1, debug_plcs[0].get_ambient())
glUniform3fv(glGetUniformLocation(shader, "point_lights[0].specular"), 1, debug_plcs[0].get_specular())
glUniform1f(glGetUniformLocation(shader, "point_lights[0].constant"), debug_plcs[0].get_constant())
glUniform1f(glGetUniformLocation(shader, "point_lights[0].linear"), debug_plcs[0].get_linear())
glUniform1f(glGetUniformLocation(shader, "point_lights[0].quadratic"), debug_plcs[0].get_quadratic())
#second light for mesh viewing
glUniform3fv(glGetUniformLocation(shader, "point_lights[1].position"), 1, debug_plcs[1].get_pos())
glUniform3fv(glGetUniformLocation(shader, "point_lights[1].diffuse"), 1, debug_plcs[1].get_diffuse())
glUniform3fv(glGetUniformLocation(shader, "point_lights[1].ambient"), 1, debug_plcs[1].get_ambient())
glUniform3fv(glGetUniformLocation(shader, "point_lights[1].specular"), 1, debug_plcs[1].get_specular())
glUniform1f(glGetUniformLocation(shader, "point_lights[1].constant"), debug_plcs[1].get_constant())
glUniform1f(glGetUniformLocation(shader, "point_lights[1].linear"), debug_plcs[1].get_linear())
glUniform1f(glGetUniformLocation(shader, "point_lights[1].quadratic"), debug_plcs[1].get_quadratic())
# spotlight
glUniform3fv(glGetUniformLocation(shader, "spot_light.position"), 1, list(active_camera.camera_pos))
glUniform3fv(glGetUniformLocation(shader, "spot_light.direction"), 1, list(active_camera.camera_front))
glUniform3fv(glGetUniformLocation(shader, "spot_light.diffuse"), 1, [0.0]*3)
glUniform3fv(glGetUniformLocation(shader, "spot_light.ambient"), 1, [0.0]*3)
glUniform3fv(glGetUniformLocation(shader, "spot_light.specular"), 1, [0.0]*3)
glUniform1f(glGetUniformLocation(shader, "spot_light.cut_off"), glm.cos(glm.radians(12.5)))
glUniform1f(glGetUniformLocation(shader, "spot_light.outer_cut_off"), glm.cos(glm.radians(45.0)))
glUniform1f(glGetUniformLocation(shader, "spot_light.constant"), 1.0)
glUniform1f(glGetUniformLocation(shader, "spot_light.linear"), 0.00003)
glUniform1f(glGetUniformLocation(shader, "spot_light.quadratic"), 0.00007)
if use_follow_cam:
gui.draw()
glUseProgram(shader)
if write_to_gif:
write_fbo_to_gif()
glfw.swap_buffers(window)
time_end = glfw.get_time()
time_delta = time_end - time_start
if wrote_to_gif:
name = f'spaceship_generator.gif'
print(f'saving gif as {name}')
images[0].save(
name,
save_all=True,
append_images=images[1:],
optimize=False,
duration=20,
loop=0
)
glfw.terminate()
| 52,869 | Python | .py | 1,509 | 30.404904 | 119 | 0.678685 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,322 | TextureLoader.py | AlexSanfilippo_ProceduralMeshGeneration/TextureLoader.py | from OpenGL.GL import glBindTexture, glTexParameteri, GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, \
GL_TEXTURE_WRAP_T, GL_REPEAT, GL_TEXTURE_MIN_FILTER, GL_TEXTURE_MAG_FILTER, GL_LINEAR, \
glTexImage2D, GL_RGBA, GL_UNSIGNED_BYTE, GL_NEAREST, GL_TEXTURE_CUBE_MAP, GL_TEXTURE_WRAP_R, \
GL_CLAMP_TO_EDGE, glGenTextures, GL_TEXTURE_CUBE_MAP_POSITIVE_X, GL_RGB
from PIL import Image
# for use with GLFW
def load_texture(path, texture):
glBindTexture(GL_TEXTURE_2D, texture)
# Set the texture wrapping parameters
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT)
# Set texture filtering parameters
# glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR)
# glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST)
# load image
image = Image.open(path)
image = image.transpose(Image.FLIP_TOP_BOTTOM)
img_data = image.convert("RGBA").tobytes()
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, image.width, image.height, 0, GL_RGBA, GL_UNSIGNED_BYTE, img_data)
return texture
#todo: convert this function (from learnOpenGL) into python
def load_cube_map(paths, texture_id):
glBindTexture(GL_TEXTURE_CUBE_MAP, texture_id)
# int width, height, nrChannels
for i, path in enumerate(paths):
# unsigned char *data = stbi_load(faces[i].c_str(), &width, &height, &nrChannels, 0)
# load_texture(path, textures[0])
image = Image.open(path)
# image = image.transpose(Image.FLIP_TOP_BOTTOM)
# image = image.transpose(Image.FLIP_LEFT_RIGHT)
img_data = image.convert("RGBA").tobytes()
glTexImage2D(GL_TEXTURE_CUBE_MAP_POSITIVE_X + i, 0, GL_RGBA, image.width, image.height, 0, GL_RGBA, GL_UNSIGNED_BYTE, img_data)
# glTexImage2D(GL_TEXTURE_CUBE_MAP_POSITIVE_X + i, 0, GL_RGB, width, height, 0, GL_RGB, GL_UNSIGNED_BYTE, data)
# stbi_image_free(data)
glTexParameteri(GL_TEXTURE_CUBE_MAP, GL_TEXTURE_MIN_FILTER, GL_LINEAR)
glTexParameteri(GL_TEXTURE_CUBE_MAP, GL_TEXTURE_MAG_FILTER, GL_LINEAR)
glTexParameteri(GL_TEXTURE_CUBE_MAP, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE)
glTexParameteri(GL_TEXTURE_CUBE_MAP, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE)
glTexParameteri(GL_TEXTURE_CUBE_MAP, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE)
return texture_id
# for use with pygame
def load_texture_pygame(path, texture):
import pygame
glBindTexture(GL_TEXTURE_2D, texture)
# Set the texture wrapping parameters
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT)
# Set texture filtering parameters
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR)
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR)
# load image
image = pygame.image.load(path)
image = pygame.transform.flip(image, False, True)
image_width, image_height = image.get_rect().size
img_data = pygame.image.tostring(image, "RGBA")
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, image_width, image_height, 0, GL_RGBA, GL_UNSIGNED_BYTE, img_data)
return texture | 3,315 | Python | .py | 59 | 50.949153 | 135 | 0.722736 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,323 | Camera.py | AlexSanfilippo_ProceduralMeshGeneration/Camera.py | import glfw
from pyrr import Vector3, vector, vector3, matrix44
from math import sin, cos, radians, atan2, sqrt
import glm as glm
from glm import vec3
from SimplePhysics import find_intersection_ray_plane
class Camera:
def __init__(self, camera_pos=[0.0, 0.0, 1.20], yaw=0, pitch=0):
self.camera_pos = Vector3(camera_pos)
self.camera_front = Vector3([0.0, 0.0, 1.0])
self.camera_up = Vector3([0.0, 1.0, 0.0])
self.camera_right = Vector3([1.0, 0.0, 0.0])
self.mouse_sensitivity = 0.25 #0.25 default
self.yaw = yaw
self.pitch = pitch
self.update_camera_vectors()
def get_view_matrix(self):
return matrix44.create_look_at(self.camera_pos, self.camera_pos + self.camera_front, self.camera_up)
def process_mouse_movement(self, xoffset, yoffset, constrain_pitch=True):
xoffset *= self.mouse_sensitivity
yoffset *= self.mouse_sensitivity
self.yaw += xoffset
self.pitch += yoffset
if constrain_pitch:
if self.pitch > 90:
self.pitch = 90
if self.pitch < -90:
self.pitch = -90
self.update_camera_vectors()
def update_camera_vectors(self):
front = Vector3([0.0, 0.0, 0.0])
front.x = cos(radians(self.yaw)) * cos(radians(self.pitch))
front.y = sin(radians(self.pitch))
front.z = sin(radians(self.yaw)) * cos(radians(self.pitch))
self.camera_front = vector.normalise(front)
self.camera_right = vector.normalise(vector3.cross(self.camera_front, Vector3([0, 1, 0])))
self.camera_up = vector.normalise(vector3.cross(self.camera_right, self.camera_front))
# Camera method for the WASD movement
def process_keyboard(self, direction, velocity):
if direction == "FORWARD":
self.camera_pos += self.camera_front * velocity
if direction == "BACKWARD":
self.camera_pos -= self.camera_front * velocity
if direction == "LEFT":
self.camera_pos -= self.camera_right * velocity
if direction == "RIGHT":
self.camera_pos += self.camera_right * velocity
if direction == "PITCH_POS":
self.pitch += velocity
self.update_camera_vectors()
if direction == "DOWN":
self.camera_pos = list(vec3(self.camera_pos) - vec3(0.0, 1.5, 0.0))
self.update_camera_vectors()
if direction == "UP":
self.camera_pos = list(vec3(self.camera_pos) + vec3(0.0, 1.5, 0.0))
self.update_camera_vectors()
def inherit_from_camera(self, other_camera):
self.camera_pos = Vector3(other_camera.camera_pos)
front = Vector3(other_camera.target_position) - self.camera_pos
self.camera_front = vector.normalise(front)
self.camera_right = vector.normalise(vector3.cross(self.camera_front, Vector3([0, 1, 0])))
self.camera_up = vector.normalise(vector3.cross(self.camera_right, self.camera_front))
class FollowCamera(Camera):
def __init__(self, camera_pos=[0.0, 0.0, 0.0], yaw=0, pitch=0, target_position=[0.0, 0.0, 0.0]):
self.camera_pos = vec3(camera_pos)
self.camera_front = vec3([0.0, 0.0, 1.0])
self.camera_up = vec3([0.0, 1.0, 0.0])
self.camera_right = vec3([1.0, 0.0, 0.0])
self.target_position = vec3(target_position)
self.offset = self.calculate_offset()
self.offset_distance = glm.length(self.offset)
self.mouse_sensitivity = 0.25 #0.25 default
self.yaw = yaw #was -90
self.pitch = pitch
self.update_camera_vectors()
self.angle = self.calclate_angle()
def set_position(self, updated_position):
self.camera_pos = vec3(updated_position)
def rotate_camera_over_time(self, speed = 1.0):
movement = glfw.get_time() * speed
distance_multiplier = 0.70
self.set_position([
cos(movement) * self.offset_distance * distance_multiplier,
cos(movement*0.5) * self.offset_distance * distance_multiplier,
# self.camera_pos.y,
# self.camera_pos.z,
sin(movement*.77) * self.offset_distance * distance_multiplier
])
self.look_at_target()
self.update_camera_vectors()
def calclate_angle(self):
'''
finds the angle on the XZ plane between the camera and the target
'''
return atan2(self.camera_pos.y - self.target_position.y, self.camera_pos.x - self.target_position.x)
def calculate_offset(self):
return self.camera_pos - self.target_position
def get_view_matrix(self):
return matrix44.create_look_at(self.camera_pos, self.camera_pos + self.camera_front, self.camera_up)
# return glm.lookAt(vec3(self.camera_pos), self.camera_pos + self.camera_front, vec3(self.camera_up))
def update_position_by_target_position(self, target_position):
self.target_position = target_position
self.camera_pos = self.target_position + self.offset
self.look_at_target()
def look_at_target(self):
"""
change camera pitch and yaw to look at a target object
"""
return glm.lookAt(vec3(self.camera_pos), self.target_position, vec3(self.camera_up))
#what does lookAt return? how to adjust
def update_camera_vectors(self):
front = vec3([0.0, 0.0, 0.0])
look_at = self.look_at_target()
# front = look_at[2].xyz
front = self.target_position - self.camera_pos
self.camera_front = glm.normalize(front)
self.camera_right = glm.normalize(glm.cross(self.camera_front, vec3([0, 1, 0])))
self.camera_up = glm.normalize(glm.cross(self.camera_right, self.camera_front))
def process_keyboard(self, direction, velocity):
if direction == "YAW_CLOCKWISE":
pos_from_center = self.camera_pos - self.target_position
self.angle += 0.01
distance_from_origin = sqrt(pow(pos_from_center.x,2) + pow(pos_from_center.z,2))
pos_from_center.z = sin(self.angle)*distance_from_origin
pos_from_center.x = cos(self.angle)*distance_from_origin
self.camera_pos = pos_from_center + self.target_position
self.update_camera_vectors()
self.offset = self.calculate_offset()
if direction == "YAW_COUNTERCLOCKWISE":
pos_from_center = self.camera_pos - self.target_position
self.angle -= 0.01
distance_from_origin = sqrt(pow(pos_from_center.x, 2) + pow(pos_from_center.z, 2))
pos_from_center.z = sin(self.angle) * distance_from_origin
pos_from_center.x = cos(self.angle) * distance_from_origin
self.camera_pos = pos_from_center + self.target_position
self.update_camera_vectors()
self.offset = self.calculate_offset()
def inherit_from_camera(self, other_camera):
self.camera_pos = vec3(other_camera.camera_pos)
self.offset = self.camera_pos - self.target_position
self.yaw = other_camera.yaw
self.pitch = other_camera.pitch
self.angle = self.calclate_angle()
self.update_camera_vectors()
def process_mouse_scroll(self, xoffset, yoffset):
zoom_speed = 10.0
front = glm.normalize(self.target_position - self.camera_pos)
updated_camera_pos = glm.vec3(self.camera_pos)
updated_camera_pos += front * yoffset * zoom_speed
minimum_distance_camera_target = 1.0
if glm.distance(self.target_position, updated_camera_pos) > minimum_distance_camera_target:
self.camera_pos = updated_camera_pos
self.offset = self.calculate_offset()
else:
# self.camera_pos = (-front * 2.001) + self.target_position
# self.offset = self.calculate_offset()
pass
def process_mouse_movement(self, xoffset, yoffset, constrain_pitch=True):
pass
# xoffset *= self.mouse_sensitivity
# yoffset *= self.mouse_sensitivity
#
# self.yaw += xoffset
# self.pitch += yoffset
#
# if constrain_pitch:
# if self.pitch > 90:
# self.pitch = 90
# if self.pitch < -90:
# self.pitch = -90
#
# self.update_camera_vectors()
class SimulationCamera(FollowCamera):
"""
Camera mode as seen in games likes banished, cities skylines, etc
target to 'follow' (rotate,zoom around) is based on camera position and forward vector
"""
def calculate_target(self):
self.target_position = find_intersection_ray_plane(
ray_origin=self.camera_pos,
ray_direction=self.camera_front,
plane=glm.vec4(0.0, 1.0, 0.0, 0.0)
)
def process_keyboard(self, direction, velocity):
sim_front = vec3(self.camera_front.x, 0.0, self.camera_front.z)
if direction == "FORWARD":
self.camera_pos += sim_front * velocity * 2.5
self.calculate_target()
if direction == "BACKWARD":
self.camera_pos += -sim_front * velocity * 2.5
self.calculate_target()
if direction == "LEFT":
self.camera_pos += -self.camera_right * velocity * 2.5
self.calculate_target()
if direction == "RIGHT":
self.camera_pos += self.camera_right * velocity * 2.5
self.calculate_target()
if direction == "YAW_CLOCKWISE":
pos_from_center = self.camera_pos - self.target_position
self.angle += 0.015 * velocity
distance_from_origin = sqrt(pow(pos_from_center.x, 2) + pow(pos_from_center.z, 2))
pos_from_center.z = sin(self.angle)*distance_from_origin
pos_from_center.x = cos(self.angle)*distance_from_origin
self.camera_pos = pos_from_center + self.target_position
self.update_camera_vectors()
self.offset = self.calculate_offset()
if direction == "YAW_COUNTERCLOCKWISE":
pos_from_center = self.camera_pos - self.target_position
self.angle -= 0.015 * velocity
distance_from_origin = sqrt(pow(pos_from_center.x, 2) + pow(pos_from_center.z, 2))
pos_from_center.z = sin(self.angle) * distance_from_origin
pos_from_center.x = cos(self.angle) * distance_from_origin
self.camera_pos = pos_from_center + self.target_position
self.update_camera_vectors()
self.offset = self.calculate_offset()
if direction == "DOWN":
self.camera_pos -= vec3(0.0, 1.5, 0.0) * velocity * 2.0
self.update_camera_vectors()
if direction == "UP":
self.camera_pos += vec3(0.0, 1.5, 0.0) * velocity * 2.0
self.update_camera_vectors()
| 10,867 | Python | .py | 221 | 39.484163 | 109 | 0.618944 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,324 | PointLightCube.py | AlexSanfilippo_ProceduralMeshGeneration/PointLightCube.py | """
7th August, 2023
Brief: A class for a cube that emits a point light
"""
import glfw
from OpenGL.GL import *
from OpenGL.GL.shaders import compileProgram, compileShader
import pyrr
import numpy as np
#Let's have it contain its own shader
vertex_src = """
# version 330
layout(location = 0) in vec3 a_position;
uniform mat4 model;
uniform mat4 projection;
uniform mat4 view;
void main()
{
gl_Position = projection * view * model * vec4(a_position, 1.0);
//gl_Position = projection * view * vec4(a_position, 1.0);
//gl_Position = vec4(a_position, 1.0); //SAVE FOR GUI LATER
}
"""
fragment_src = """
# version 330
out vec4 out_color;
uniform vec3 light_color;
void main()
{
out_color = vec4(light_color,1.f);
}
"""
class PointLightCube:
def __init__(self,
pos=[0.0, 0.0, 0.0],
ambient=[1.0, 1.0, 1.0],
diffuse=[1.0, 1.0, 1.0],
specular=[1.0, 1.0, 1.0],
scale=1.0,
constant=1.0,
linear=0.014,
quadratic=0.00007
):
self.pos = pos
self.ambient = ambient
self.diffuse = diffuse
self.specular = specular
self.scale = scale
# falloff
self.constant = constant
self.linear = linear
self.quadratic = quadratic
self.vertices = np.array([-0.5, -0.5, 0.5,
0.5, -0.5, 0.5,
0.5, 0.5, 0.5,
-0.5, 0.5, 0.5,
-0.5, -0.5, -0.5,
0.5, -0.5, -0.5,
0.5, 0.5, -0.5,
-0.5, 0.5, -0.5,
0.5, -0.5, -0.5,
0.5, 0.5, -0.5,
0.5, 0.5, 0.5,
0.5, -0.5, 0.5,
-0.5, 0.5, -0.5,
-0.5, -0.5, -0.5,
-0.5, -0.5, 0.5,
-0.5, 0.5, 0.5,
-0.5, -0.5, -0.5,
0.5, -0.5, -0.5,
0.5, -0.5, 0.5,
-0.5, -0.5, 0.5,
0.5, 0.5, -0.5,
-0.5, 0.5, -0.5,
-0.5, 0.5, 0.5,
0.5, 0.5, 0.5], dtype=np.float32) * scale
self.indices = np.array([0, 1, 2, 2, 3, 0,
4, 5, 6, 6, 7, 4,
8, 9, 10, 10, 11, 8,
12, 13, 14, 14, 15, 12,
16, 17, 18, 18, 19, 16,
20, 21, 22, 22, 23, 20], dtype=np.uint32)
# cube VAO
self.VAO = glGenVertexArrays(1)
glBindVertexArray(self.VAO)
self.VBO = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, self.VBO)
glBufferData(GL_ARRAY_BUFFER, self.vertices.nbytes, self.vertices, GL_STATIC_DRAW)
# Element Buffer Object
self.EBO = glGenBuffers(1)
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, self.EBO)
glBufferData(GL_ELEMENT_ARRAY_BUFFER, self.indices.nbytes, self.indices, GL_STATIC_DRAW)
# cube vertices (vertex attribute)
glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, self.vertices.itemsize * 3, ctypes.c_void_p(0))
glEnableVertexAttribArray(0)
self.shader = compileProgram(compileShader(vertex_src, GL_VERTEX_SHADER), \
compileShader(fragment_src, GL_FRAGMENT_SHADER))
self.light_color_loc = glGetUniformLocation(self.shader, "light_color")
self.model_loc = glGetUniformLocation(self.shader, "model")
self.proj_loc = glGetUniformLocation(self.shader, "projection")
self.view_loc = glGetUniformLocation(self.shader, "view")
self.projection = pyrr.matrix44.create_perspective_projection_matrix(45, 1280 / 720, 0.1, 1000)
# glUniformMatrix4fv(self.proj_loc, 1, GL_FALSE, self.projection)
self.translation = pyrr.matrix44.create_from_translation(pyrr.Vector3(pos))
self.rotation = pyrr.matrix44.create_from_quaternion([1.0, 1.0, 1.0, 1.0])
def draw(self,view):
glBindVertexArray(self.VAO)
glUseProgram(self.shader)
glUniformMatrix4fv(self.view_loc, 1, GL_FALSE, view)
# glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, self.pos)
model_mat = pyrr.matrix44.create_identity()
model_mat = pyrr.matrix44.multiply(self.rotation, self.translation) #.translation * self.rotation
glUniformMatrix4fv(self.model_loc, 1, GL_FALSE, model_mat)
glUniformMatrix4fv(self.proj_loc, 1, GL_FALSE, self.projection)
glUniform3fv(self.light_color_loc, 1, self.diffuse)
glDrawElements(GL_TRIANGLES, len(self.indices), GL_UNSIGNED_INT, None)
# glDrawArrays(GL_TRIANGLES, 0, 36);
def set_ambient(self, ambient):
self.ambient = ambient
def set_diffuse(self, diffuse):
self.diffuse = diffuse
def set_specular(self, specular):
self.specular = specular
def get_ambient(self):
return self.ambient
def get_diffuse(self):
return self.diffuse
def get_specular(self):
return self.specular
def get_pos(self):
return self.pos
def get_constant(self):
return self.constant
def get_linear(self):
return self.linear
def get_quadratic(self):
return self.quadratic
def set_pos(self, pos):
self.pos = pos
self.translation = pyrr.matrix44.create_from_translation(pyrr.Vector3(self.pos))
# rotate the PointLightCube, given a quaternion from bullet
def set_orientation(self, orientation):
# cube_pos = pyrr.matrix44.create_from_translation(pyrr.Vector3([6, 4, 0]))
#opt A
# rot_y = pyrr.Matrix44.from_y_rotation(orientation)
#opt B
# rot_y = pyrr.matrix44.create_from_inverse_of_quaternion(orientation)
# opt C
rot_y = pyrr.Matrix44.from_quaternion(orientation)
#pass to shader
self.rotation = rot_y #pyrr.matrix44.multiply(rot_y, self.translation)
# TP: make sure we can rotate the object this way
def rotate_cube(self):
rot_y = pyrr.Matrix44.from_y_rotation(0.8 * glfw.get_time())
self.rotation = rot_y #pyrr.matrix44.multiply(rot_y, self.translation)
| 6,275 | Python | .py | 155 | 30.283871 | 105 | 0.571947 | AlexSanfilippo/ProceduralMeshGeneration | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,325 | setup.py | Mauropieroni_fastPTA/setup.py | from setuptools import setup, find_packages
from fastPTA import __author__, __version__, __url__
with open("README.md", "r") as fh:
long_description = fh.read()
with open("requirements.txt") as f:
required_packages = f.read().splitlines()
setup(
name="fastPTA",
description="Code for fast PTA forecasts",
keywords="PTAs, GWs",
version=__version__,
author=__author__,
url=__url__,
long_description=long_description,
long_description_content_type="text/markdown",
packages=find_packages(),
install_requires=required_packages,
package_data={"fastPTA": ["defaults/*"]},
)
| 625 | Python | .py | 19 | 28.894737 | 52 | 0.68386 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,326 | HD_constraints.ipynb | Mauropieroni_fastPTA/examples/HD_constraints.ipynb | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {},
"outputs": [],
"source": [
"# Global\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Local\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"import fastPTA.plotting_functions as pf\n",
"from fastPTA.Fisher_code import compute_fisher"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Constants to be used in the analysis"
]
},
{
"cell_type": "code",
"execution_count": 77,
"metadata": {},
"outputs": [],
"source": [
"# Total observation time in years\n",
"T_obs_yrs = 16.03\n",
"\n",
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# Number of pulsars in the analysis\n",
"n_pulsars = 25\n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Parameters for the HD computations:\n",
"# Method to compute the HD, either \"Legendre\" or \"Binned\"\n",
"HD_basis = \"Legendre\"\n",
"\n",
"# Maximum order of the Legendre polynomials\n",
"HD_order = 5\n",
"\n",
"# Whether to add some gaussian prior for the HD coefficients to the Fisher matrix\n",
"add_HD_prior = True\n",
"\n",
"# Specify the type of noise to be used in the analysis\n",
"which_experiment = eu.EPTAlike\n",
"\n",
"# Number of points to generate for the Fisher\n",
"len_fisher_data = int(1e4)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Set the inputs for the Fisher"
]
},
{
"cell_type": "code",
"execution_count": 78,
"metadata": {},
"outputs": [],
"source": [
"# Assemble the vector with two signal parameters\n",
"signal_parameters = np.array([log_amplitude, tilt])\n",
"\n",
"# Length of the parameter vector\n",
"parameter_len = (\n",
" len(signal_parameters) + HD_order + 1\n",
" if HD_order\n",
" else len(signal_parameters)\n",
")\n",
"\n",
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"generate_catalog_kwargs = {\n",
" \"n_pulsars\": n_pulsars,\n",
" \"save_catalog\": False,\n",
" **which_experiment,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors\n",
"get_tensors_kwargs = {\n",
" \"add_curn\": False,\n",
" \"HD_order\": HD_order,\n",
" \"HD_basis\": HD_basis,\n",
" \"regenerate_catalog\": True,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate the fisher matrix\n",
"fisher_kwargs = {\n",
" \"T_obs_yrs\": T_obs_yrs,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_parameters\": signal_parameters,\n",
"}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Compute the Fisher"
]
},
{
"cell_type": "code",
"execution_count": 79,
"metadata": {},
"outputs": [],
"source": [
"(\n",
" frequency,\n",
" signal,\n",
" HD_functions_IJ,\n",
" HD_coeffs,\n",
" effective_noise,\n",
" SNR,\n",
" fisher,\n",
") = compute_fisher(\n",
" **fisher_kwargs,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs,\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Generate some data from the Fisher (and reject with prior) "
]
},
{
"cell_type": "code",
"execution_count": 80,
"metadata": {},
"outputs": [],
"source": [
"# Get the length of the signal parameters\n",
"len_signal = len(signal_parameters)\n",
"\n",
"# Compute the covariance matrix\n",
"covariance = ut.compute_inverse(fisher)\n",
"\n",
"# Generate the Fisher data\n",
"fisher_data = np.random.multivariate_normal(\n",
" np.append(signal_parameters, HD_coeffs),\n",
" covariance,\n",
" size=len_fisher_data,\n",
")\n",
"\n",
"# If add_HD_prior, drop points that are outside the prior\n",
"if add_HD_prior:\n",
" fisher_data = fisher_data[\n",
" np.any(fisher_data[:, len_signal:] > -1, axis=1)\n",
" & np.any(fisher_data[:, len_signal:] < 1, axis=1)\n",
" ]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Plot the results"
]
},
{
"cell_type": "code",
"execution_count": 81,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 600x400 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Plot the results\n",
"if get_tensors_kwargs[\"HD_basis\"].lower() == \"legendre\":\n",
" pf.plot_HD_Legendre(\n",
" 1000,\n",
" fisher_data[:, len_signal:],\n",
" r\"$\\rm HD \\ reconstruction \\ Legendre$\",\n",
" )\n",
" plt.ylim(-0.5, 1.5)\n",
"\n",
"elif get_tensors_kwargs[\"HD_basis\"].lower() == \"binned\":\n",
" pf.plot_HD_binned(\n",
" fisher_data[:, len_signal:],\n",
" HD_coeffs,\n",
" r\"$\\rm HD \\ reconstruction \\ binned$\",\n",
" )\n",
" plt.ylim(-1.0, 1.0)\n",
"\n",
"else:\n",
" raise ValueError(\"Cannot use that\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
| 47,823 | Python | .py | 245 | 190.771429 | 41,670 | 0.908971 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,327 | strong_signal_limit.ipynb | Mauropieroni_fastPTA/examples/strong_signal_limit.ipynb | {
"cells": [
{
"cell_type": "markdown",
"id": "f5da63da",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "e5b92ec3",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Import global libs\n",
"import tqdm\n",
"import numpy as np\n",
"import jax.numpy as jnp\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Import local libs\n",
"import fastPTA.utils as ut\n",
"import fastPTA.get_tensors as gt"
]
},
{
"cell_type": "markdown",
"id": "d623d2e9",
"metadata": {},
"source": [
"### Then set some constants to be used below"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "f40c03dc",
"metadata": {},
"outputs": [],
"source": [
"# Set the number of pulsars\n",
"n_pulsars = 100\n",
"\n",
"# Number of realizations\n",
"n_realizations = 30\n",
"\n",
"# Spherical harmonics order\n",
"l_max = 6\n",
"\n",
"# Set nside (which sets the number of pixels) for healpy\n",
"nside = 12"
]
},
{
"cell_type": "markdown",
"id": "11d95f0e",
"metadata": {},
"source": [
"### A function to generate random positions for the pulsars on the sky, and compute the ORF_lm"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "18573af4",
"metadata": {},
"outputs": [],
"source": [
"def get_gamma_and_correlations(n_pulsars):\n",
" \n",
" # Theta and phi\n",
" theta = np.arccos(np.random.uniform(-1, 1, n_pulsars))\n",
" phi = np.random.uniform(0.0, 2 * np.pi, n_pulsars)\n",
"\n",
" # Generate the unit vectors\n",
" sky_directions = gt.unit_vector(theta, phi)\n",
"\n",
" # Compute the overlap reduction function for all pulsar pairs\n",
" correlations = gt.get_correlations_lm_IJ(sky_directions, l_max, nside)\n",
"\n",
" # Compute the inverse of the overlap reduction function for (l,m) = (0,0)\n",
" gamma_0_inverse = np.linalg.inv(correlations[0])\n",
"\n",
" return gamma_0_inverse, correlations"
]
},
{
"cell_type": "markdown",
"id": "42c73a18",
"metadata": {},
"source": [
"### A function to compute the uncertainties"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "e70645b1",
"metadata": {},
"outputs": [],
"source": [
"def get_uncertainties(\n",
" n_pulsars, gamma_0_inverse, correlations, autocorrelations\n",
"):\n",
"\n",
" # Use the autocorrelations if needed\n",
" if autocorrelations:\n",
" to_use = correlations\n",
"\n",
" else:\n",
" to_use = jnp.where(\n",
" (1.0 - jnp.eye(n_pulsars))[None, :, :], correlations, 0.0\n",
" )\n",
"\n",
" # Compute the Fisher matrix\n",
" fisher_matrix = (4 * jnp.pi) * jnp.einsum(\n",
" \"ij,kl,ajk,bli->ab\",\n",
" gamma_0_inverse,\n",
" gamma_0_inverse,\n",
" to_use,\n",
" to_use,\n",
" )\n",
"\n",
" # Compute the covariance matrix\n",
" covariance = np.linalg.inv(fisher_matrix)\n",
"\n",
" # Return the error bars with/without correlations\n",
" return np.sqrt(np.diag(covariance))"
]
},
{
"cell_type": "markdown",
"id": "e251f922",
"metadata": {},
"source": [
"### Get the errors for many realizations of the pulsar catalog"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "78274b22",
"metadata": {},
"outputs": [],
"source": [
"try:\n",
" data = np.load('generated_data/strong_signal.npz')\n",
" auto = data['auto']\n",
" no_auto = data['no_auto']\n",
"\n",
"except FileNotFoundError:\n",
" # Initialize the arrays to store the uncertainties\n",
" auto = np.zeros(shape=(n_realizations, (l_max + 1) ** 2))\n",
" no_auto = np.zeros(shape=(n_realizations, (l_max + 1) ** 2))\n",
"\n",
" # Loop over the realizations and compute the uncertainties\n",
" for i in tqdm.tqdm(range(n_realizations)):\n",
" gamma, corr = get_gamma_and_correlations(n_pulsars)\n",
" auto[i] = get_uncertainties(n_pulsars, gamma, corr, True)\n",
" no_auto[i] = get_uncertainties(n_pulsars, gamma, corr, False)\n",
"\n",
" np.savez('generated_data/strong_signal.npz', auto=auto, no_auto=no_auto)"
]
},
{
"cell_type": "markdown",
"id": "e2be5333",
"metadata": {},
"source": [
"### Plot the results"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "a061b9e8",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/plain": [
"<Figure size 640x480 with 0 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# Compute the quantiles to be plotted\n",
"to_plot = np.quantile(auto, [(1 - 0.95) / 2, 0.5, (1 + 0.95) / 2], axis=0)\n",
"to_plot_no_auto = np.quantile(\n",
" no_auto, [(1 - 0.95) / 2, 0.5, (1 + 0.95) / 2], axis=0\n",
")\n",
"\n",
"plt.figure(figsize=(8, 5))\n",
"\n",
"plt.fill_between(\n",
" np.arange(49), to_plot[0, :], to_plot[2, :], color=\"Blue\", alpha=0.2\n",
")\n",
"\n",
"plt.plot(to_plot[0], color=\"Blue\", label=r\"${\\rm With\\ auto-correlations}$\")\n",
"plt.plot(to_plot[1], color=\"Blue\")\n",
"plt.plot(to_plot[2], color=\"Blue\")\n",
"\n",
"plt.fill_between(\n",
" np.arange(49),\n",
" to_plot_no_auto[0, :],\n",
" to_plot_no_auto[2, :],\n",
" color=\"Red\",\n",
" alpha=0.2,\n",
")\n",
"\n",
"plt.plot(\n",
" to_plot_no_auto[0], color=\"Red\", label=r\"${\\rm Without auto-correlations}$\"\n",
")\n",
"plt.plot(to_plot_no_auto[1], color=\"Red\")\n",
"plt.plot(to_plot_no_auto[2], color=\"Red\")\n",
"\n",
"custom_xticks = [0] +[ut.get_n_coefficients_real(i) for i in range(0, l_max )]\n",
"custom_xtick_labels = [r\"$0$\", r\"$1$\", r\"$2$\", r\"$3$\", r\"$4$\", r\"$5$\", r\"$6$\"]\n",
"plt.xticks(custom_xticks, custom_xtick_labels)\n",
"\n",
"plt.xlabel(r\"$\\ell - {\\rm order}$\")\n",
"plt.ylabel(r\"$\\Delta{c_{\\ell m}}$\")\n",
"plt.yscale(\"log\")\n",
"plt.legend(loc=\"lower right\")\n",
"plt.show()\n",
"plt.tight_layout()\n",
"plt.savefig(\"plots/strong_signal.pdf\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.9.12 ('env_GWFast')",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
},
"vscode": {
"interpreter": {
"hash": "37bcb7777d3229dfb79a16124a08c500a91162130c755ccd3fb5278a92e948b4"
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}
| 54,651 | Python | .py | 281 | 189.996441 | 47,150 | 0.90561 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,328 | scan_parameter.ipynb | Mauropieroni_fastPTA/examples/scan_parameter.ipynb | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Global\n",
"import tqdm\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Local\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"import fastPTA.plotting_functions as pf\n",
"from fastPTA.Fisher_code import compute_fisher"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Constants to be used in the analysis"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# Total observation time in years\n",
"T_obs_yrs = 16.03\n",
"\n",
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# Number of pulsars in the analysis\n",
"n_pulsars = 25\n",
"\n",
"# Specify the type of noise to be used in the analysis\n",
"which_experiment = eu.EPTAlike"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Set the inputs for the Fisher"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"generate_catalog_kwargs = {\n",
" \"n_pulsars\": n_pulsars,\n",
" \"save_catalog\": False,\n",
" **which_experiment,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors\n",
"get_tensors_kwargs = {\n",
" \"add_curn\": False,\n",
" \"regenerate_catalog\": True,\n",
"}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Prepare the model to scan over"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"# Set the label to specify the signal model\n",
"signal_label = \"power_law\"\n",
"\n",
"# Labels for the signal parameters\n",
"parameter_labels = [\n",
" r\"$\\alpha_{\\rm PL}$\",\n",
" r\"$n_{\\rm T}$\",\n",
"]\n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Index to scan over\n",
"parameter_index = 0\n",
"\n",
"# Range of the parameter to scan over\n",
"parameter_range = np.linspace(-8, -6, 11)\n",
"\n",
"# Number of realizations to average over\n",
"n_realizations = 10"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Do the scan"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|██████████| 11/11 [00:03<00:00, 3.21it/s]\n"
]
}
],
"source": [
"# Assemble the vector with two signal parameters\n",
"s_params = np.array([log_amplitude, tilt])\n",
"\n",
"# Define the array to store the results\n",
"results = np.zeros(shape=(len(parameter_range), n_realizations, len(s_params)))\n",
"\n",
"for i in tqdm.tqdm(range(len(parameter_range))):\n",
" for n in range(n_realizations):\n",
" # Update the parameter of interest\n",
" s_params[parameter_index] = parameter_range[i]\n",
"\n",
" # Dictionary with the kwargs to generate the fisher matrix\n",
" fisher_kwargs = {\n",
" \"T_obs_yrs\": T_obs_yrs,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_label\": signal_label,\n",
" \"signal_parameters\": s_params,\n",
" }\n",
"\n",
" # Compute the fisher matrix\n",
" (\n",
" frequency,\n",
" signal,\n",
" HD_functions_IJ,\n",
" HD_coeffs,\n",
" effective_noise,\n",
" SNR,\n",
" fisher,\n",
" ) = compute_fisher(\n",
" **fisher_kwargs,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs,\n",
" )\n",
"\n",
" # Compute the covariance matrix\n",
" covariance = ut.compute_inverse(fisher)\n",
"\n",
" # Store the results\n",
" results[i, n] = np.sqrt(np.diag(covariance))"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Plot the results"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 640x480 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"cols = list(pf.my_colormap.keys())\n",
"for i in range(len(s_params)):\n",
" plt.errorbar(\n",
" parameter_range,\n",
" np.mean(results, axis=1)[:, i],\n",
" yerr=np.std(results, axis=1)[:, i],\n",
" color=pf.my_colormap[cols[i]],\n",
" label=parameter_labels[i],\n",
" )\n",
"\n",
"plt.yscale(\"log\")\n",
"\n",
"plt.ylabel(r\"$\\rm Uncertainty \\ (1 \\sigma)$\")\n",
"plt.xlabel(parameter_labels[parameter_index])\n",
"plt.legend(loc=1, ncols=2)\n",
"plt.tight_layout()"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
| 45,455 | Python | .py | 251 | 176.537849 | 39,174 | 0.907002 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,329 | N_scaling_figure.ipynb | Mauropieroni_fastPTA/examples/N_scaling_figure.ipynb | {
"cells": [
{
"cell_type": "markdown",
"id": "3d2a5638",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "a061b9e8",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Global\n",
"import os\n",
"import tqdm\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Local\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"import fastPTA.plotting_functions as pf\n",
"from fastPTA.Fisher_code import compute_fisher"
]
},
{
"cell_type": "markdown",
"id": "8c55c5d3",
"metadata": {},
"source": [
"### Constants to be used in the analysis"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "9afe1b2e",
"metadata": {},
"outputs": [],
"source": [
"# Total observation time in years\n",
"T_obs_yrs = 16.03\n",
"\n",
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Parameters for the HD computations:\n",
"# Method to compute the HD, either \"Legendre\" or \"Binned\"\n",
"HD_basis = \"Legendre\"\n",
"\n",
"# Maximum order of the Legendre polynomials\n",
"HD_order = 5\n",
"\n",
"# Whether to add some gaussian prior for the HD coefficients to the Fisher matrix\n",
"add_HD_prior = False\n",
"\n",
"# Specify the type of noise to be used in the analysis\n",
"which_experiment = eu.EPTAlike\n",
"\n",
"# Whether you want to rerun the analysis\n",
"rerun = False\n",
"\n",
"# Minimum number of pulsars to be used in the analysis\n",
"N_min = 30\n",
"\n",
"# Maximum number of pulsars to be used in the analysis\n",
"N_max = 200\n",
"\n",
"# Number of points in N_pulsars to scan over\n",
"N_times = 15\n",
"\n",
"# Number of realizations to be generated for each number of pulsars\n",
"N_realizations = 10\n",
"\n",
"# Name of the outfile, no need for the extension\n",
"# (will be stored in generated_data/)\n",
"outname = \"Default\"\n",
"\n",
"# Labels for the signal parameters\n",
"signal_labels = [r\"$\\alpha_{*}$\", r\"$n_{\\rm T}$\"]"
]
},
{
"cell_type": "markdown",
"id": "9130a943",
"metadata": {},
"source": [
"### Builds the dictionaries with inputs for the code"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "e1c85ecb",
"metadata": {},
"outputs": [],
"source": [
"# Assemble the vector with two signal parameters\n",
"signal_parameters = np.array([log_amplitude, tilt])\n",
"\n",
"# Length of the parameter vector\n",
"parameter_len = (\n",
" len(signal_parameters) + HD_order + 1\n",
" if HD_order\n",
" else len(signal_parameters)\n",
")\n",
"\n",
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"default_pulsars = {\n",
" \"n_pulsars\": 30,\n",
" \"save_catalog\": False,\n",
" **which_experiment,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors\n",
"get_tensors_kwargs = {\n",
" \"add_curn\": False,\n",
" \"HD_order\": HD_order,\n",
" \"HD_basis\": HD_basis,\n",
" \"regenerate_catalog\": True,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate the fisher matrix\n",
"fisher_kwargs = {\n",
" \"T_obs_yrs\": T_obs_yrs,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_parameters\": signal_parameters,\n",
"}"
]
},
{
"cell_type": "markdown",
"id": "24800e4c",
"metadata": {},
"source": [
"### Check if all the folders are in place and define save_path"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "5ee5ac36",
"metadata": {},
"outputs": [],
"source": [
"# If it's not there, create the folder to store the generated data\n",
"if not os.path.exists(\"generated_data/\"):\n",
" os.mkdir(\"generated_data/\")\n",
"\n",
"# If it's not there, create the folder to store the plots\n",
"if not os.path.exists(\"plots/\"):\n",
" os.mkdir(\"plots/\")\n",
"\n",
"# Build the save path\n",
"if outname != \"Default\":\n",
" save_path = \"generated_data/\" + outname + \".npz\"\n",
"elif HD_order == 0:\n",
" save_path = \"generated_data/N_scaling.npz\"\n",
"else:\n",
" save_path = \"generated_data/N_scaling_HD.npz\""
]
},
{
"cell_type": "markdown",
"id": "9f847562",
"metadata": {},
"source": [
"### Run for all values of N_pulsars"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "6b9de97a",
"metadata": {},
"outputs": [],
"source": [
"try:\n",
" # Check if the file is there and load the data if not rerun\n",
" if rerun:\n",
" raise FileNotFoundError(\"Forcing regeneration\")\n",
"\n",
" data = np.load(save_path)\n",
"\n",
" N_pulsars = data[\"N_pulsars\"]\n",
" SNR_values = data[\"SNR_values\"]\n",
" parameters_uncertainties = data[\"parameters_uncertainties\"]\n",
"\n",
"except FileNotFoundError:\n",
"\n",
" # Define the vector with the values of N_pulsars to scan over\n",
" N_pulsars = np.unique(np.geomspace(N_min, N_max, N_times, dtype=int))\n",
"\n",
" # Redefine N_times in case repeaded values have been dropped\n",
" N_times = len(N_pulsars)\n",
"\n",
" # Initialize the arrays to store the results\n",
" SNR_values = np.zeros(shape=(N_times, N_realizations))\n",
" parameters_uncertainties = np.zeros(\n",
" shape=(N_times, N_realizations, parameter_len)\n",
" )\n",
"\n",
" # Loop over the values of N_pulsars\n",
" for i in range(N_times):\n",
" # Set the value of N_pulsars\n",
" generate_catalog_kwargs = default_pulsars.copy()\n",
" generate_catalog_kwargs[\"n_pulsars\"] = N_pulsars[i]\n",
"\n",
" # Compute the fisher for all the realizations at a given N_pulsars\n",
" print(\"Here starts N = %d\" % (N_pulsars[i]))\n",
" for j in tqdm.tqdm(range(N_realizations)):\n",
" (\n",
" frequency,\n",
" signal,\n",
" HD_functions_IJ,\n",
" HD_coeffs,\n",
" effective_noise,\n",
" SNR,\n",
" fisher,\n",
" ) = compute_fisher(\n",
" n_frequencies=n_frequencies,\n",
" signal_label=\"power_law\",\n",
" signal_parameters=signal_parameters,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs,\n",
" )\n",
"\n",
" if HD_order and add_HD_prior:\n",
" fisher += np.diag(\n",
" np.append(\n",
" np.zeros(len(signal_parameters)), np.ones(HD_order + 1)\n",
" )\n",
" )\n",
"\n",
" # Get covariance matrix and errors\n",
" c_inverse = ut.compute_inverse(fisher)\n",
" errors = np.sqrt(np.diag(c_inverse))\n",
"\n",
" # Store the results\n",
" SNR_values[i, j] = SNR\n",
" parameters_uncertainties[i, j] = errors\n",
"\n",
" SNR_mean = np.mean(SNR_values[i], axis=-1)\n",
" SNR_std = np.std(SNR_values[i], axis=-1)\n",
" print(\"SNR=%.2e +-%.2e \\n\" % (SNR_mean, SNR_std))\n",
"\n",
" to_save = {\n",
" \"N_pulsars\": N_pulsars,\n",
" \"SNR_values\": SNR_values,\n",
" \"parameters_uncertainties\": parameters_uncertainties,\n",
" }\n",
"\n",
" np.savez(save_path, **to_save)"
]
},
{
"cell_type": "markdown",
"id": "7542705d",
"metadata": {},
"source": [
"### Compute means and stds"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "cc04418c",
"metadata": {},
"outputs": [],
"source": [
"# Mean and std of the SNR over the realizations\n",
"SNR_mean = np.mean(SNR_values, axis=-1)\n",
"SNR_std = np.std(SNR_values, axis=-1)\n",
"\n",
"# Mean and std of the uncertainties over the realizations\n",
"uncertainties_means = np.mean(parameters_uncertainties, axis=1)\n",
"uncertainties_stds = np.std(parameters_uncertainties, axis=1)"
]
},
{
"cell_type": "markdown",
"id": "65963783",
"metadata": {},
"source": [
"### Plot the scaling of SNR with N_pulsars"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "c67b6dfe",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.figure(figsize=(8, 5))\n",
"\n",
"plt.errorbar(\n",
" N_pulsars,\n",
" SNR_mean,\n",
" yerr=SNR_std,\n",
" color=pf.my_colormap[\"cyan\"],\n",
" fmt=\"o\",\n",
" markersize=4,\n",
" linestyle=\"dashed\",\n",
" capsize=7,\n",
")\n",
"\n",
"plt.loglog(\n",
" N_pulsars,\n",
" np.sqrt(N_pulsars),\n",
" linestyle=\"--\",\n",
" color=\"black\",\n",
")\n",
"\n",
"plt.text(1e2, 7, s=r\"$\\propto \\sqrt{N_{\\rm pulsars}}$\", fontsize=15)\n",
"plt.xlabel(r\"$N_{\\rm pulsars}$\", fontsize=20)\n",
"plt.ylabel(r\"$\\rm SNR$\", fontsize=20)\n",
"plt.ylim(4, 2e1)\n",
"plt.tight_layout()\n",
"plt.savefig(\"plots/SNR_N_scaling.pdf\")"
]
},
{
"cell_type": "markdown",
"id": "3f74d4ce",
"metadata": {},
"source": [
"### Plot the scaling of uncertainties on the SGWB shape parameters with N_pulsars"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "25846831",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.figure(figsize=(8, 5))\n",
"for i in range(len(signal_parameters)):\n",
" colors = list(pf.my_colormap.keys())\n",
" col = colors[np.mod(i, len(colors))]\n",
" plt.errorbar(\n",
" N_pulsars,\n",
" uncertainties_means[:, i],\n",
" yerr=uncertainties_stds[:, i],\n",
" color=col,\n",
" fmt=\"o\",\n",
" markersize=4,\n",
" linestyle=\"dashed\",\n",
" capsize=7,\n",
" label=signal_labels[i],\n",
" )\n",
"\n",
"plt.loglog(\n",
" N_pulsars,\n",
" 1.4 / N_pulsars**0.5,\n",
" linestyle=\"--\",\n",
" color=\"black\",\n",
")\n",
"\n",
"plt.text(1e2, 0.2, s=r\"$\\propto 1/\\sqrt{N_{\\rm pulsars}}$\", fontsize=15)\n",
"\n",
"plt.xticks([], fontsize=20)\n",
"plt.xlabel(r\"$N_{\\rm pulsars}$\", fontsize=20)\n",
"plt.ylabel(r\"$\\rm Uncertainties$\", fontsize=20)\n",
"plt.legend(fontsize=15)\n",
"plt.tight_layout()\n",
"plt.savefig(\"plots/Error_N_scaling.pdf\")"
]
},
{
"cell_type": "markdown",
"id": "26ca214d",
"metadata": {},
"source": [
"### Plot the scaling of uncertainties on the Legendre polynomials coefficients with N_pulsars"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "3ddf17e9",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"if HD_order:\n",
" if HD_basis == \"Legendre\":\n",
" label = r\"$\\rm Polynomial \\ \\ell = \\ $\"\n",
" else:\n",
" label = r\"$\\rm Bin \\ = \\ $\"\n",
" # here plot scaling of SNR with T_obs\n",
" plt.figure(figsize=(8, 5))\n",
" for i in range(HD_order + 1):\n",
" plt.errorbar(\n",
" N_pulsars,\n",
" uncertainties_means[:, len(signal_parameters) + i],\n",
" yerr=uncertainties_stds[:, len(signal_parameters) + i],\n",
" label=label + str(i),\n",
" color=pf.cmap_HD(0.1 + i / 1.1 / (HD_order + 1)),\n",
" )\n",
"\n",
" plt.loglog(\n",
" N_pulsars,\n",
" 2 / N_pulsars,\n",
" linestyle=\"--\",\n",
" color=\"black\",\n",
" )\n",
"\n",
" plt.xlabel(r\"$N_{\\rm pulsars}$\")\n",
" plt.ylabel(r\"$\\rm Uncertainties$\")\n",
" plt.ylim(1e-3, 4e-1)\n",
" plt.legend(fontsize=12, ncols=2)\n",
" plt.tight_layout()\n",
" plt.savefig(\"plots/SNR_N_scaling_HD.pdf\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "49f83d20",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.9.12 ('env_GWFast')",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
},
"vscode": {
"interpreter": {
"hash": "37bcb7777d3229dfb79a16124a08c500a91162130c755ccd3fb5278a92e948b4"
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}
| 181,178 | Python | .py | 502 | 356.310757 | 70,810 | 0.93009 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,330 | sqrt_basis_figure.ipynb | Mauropieroni_fastPTA/examples/sqrt_basis_figure.ipynb | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Import global libs\n",
"import numpy as np\n",
"from functools import partial\n",
"import matplotlib.pyplot as plt\n",
"from multiprocessing import Pool, cpu_count\n",
"\n",
"# Import local libs\n",
"import fastPTA.utils as ut"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Then set some constants to be used below"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# Confidence level for the Cls\n",
"limit_cl = 0.95 \n",
"\n",
"# Spherical harmonics order sqrt basis\n",
"l_max_sqrt = 3 \n",
"\n",
"# Spherical harmonics order linear basis\n",
"l_max_lin = 6 \n",
"\n",
"# Number of points to be generated\n",
"npoints = 10000 \n",
"\n",
"# Absolute priors for the coefficients\n",
"abs_prior = [0, 50] \n",
"\n",
"# Number of cores to be used for parallel computations\n",
"num_cores = cpu_count() "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Load the NANOgrav constraints in the sqrt basis"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"limits_NG_sqrt = np.loadtxt(\"data_paper_2/limits_Cl_powerlaw_sqrt_ng15.dat\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Generate the complex gLM in the sqrt basis from the flat prior"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"# Compute the number of complex coefficients to generate\n",
"n_coefficients = ut.get_n_coefficients_complex(l_max_sqrt)\n",
"\n",
"# Generate the gLMs from the priors\n",
"LL, MM, _, _, sort_indexes = ut.get_sort_indexes(l_max_sqrt)\n",
"\n",
"# Generate the absolute value for the gLM prior samples\n",
"abs_gLM_grid = np.random.uniform(\n",
" low=abs_prior[0],\n",
" high=abs_prior[1],\n",
" size=(npoints, n_coefficients),\n",
")\n",
"\n",
"phi_LM_grid = np.random.uniform(\n",
" low=0.0,\n",
" high=2.0 * np.pi,\n",
" size=(npoints, n_coefficients),\n",
")\n",
"\n",
"# Generate the phases for the gLM prior samples\n",
"exp_j_phi_LM_grid = np.exp(1j * phi_LM_grid)\n",
"\n",
"# Assemble to get the gLM\n",
"gLM_complex = exp_j_phi_LM_grid * abs_gLM_grid\n",
"\n",
"# For M = 0 the coefficients are real\n",
"gLM_complex[:, MM == 0.0] = (\n",
" gLM_complex[:, MM == 0.0].real\n",
" * 2.0\n",
" * (np.heaviside(np.pi - phi_LM_grid[:, MM == 0.0], 0.0) - 0.5)\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Convert from the sqrt to the linear basis"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"# Compute the linear coefficients from the sqrt coefficients\n",
"with Pool(num_cores) as p:\n",
" clm_linear_prior = np.array(\n",
" p.map(\n",
" partial(\n",
" ut.sqrt_to_lin_conversion,\n",
" l_max_lin=l_max_lin,\n",
" real_basis_input=False,\n",
" ),\n",
" gLM_complex,\n",
" chunksize=int(npoints / num_cores),\n",
" )\n",
" )"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Compute the Cls in the linear basis"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
"# Normalize such that c00 = sqrt(4.*pi), see NG15 paper\n",
"clm_linear_prior = (np.sqrt(4.0 * np.pi) / clm_linear_prior[:, 0])[\n",
" :, None\n",
"] * clm_linear_prior\n",
"\n",
"# Compute the Cls from the linear coefficients\n",
"CL_linear_prior = ut.get_CL_from_real_clm(clm_linear_prior.T)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Plot the results"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 720x660 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"fig = plt.figure(figsize=(0.4 * 12.0, 0.4 * 11.0), dpi=150, edgecolor=\"white\")\n",
"\n",
"ax = fig.add_subplot(1, 1, 1)\n",
"ax.tick_params(\n",
" axis=\"both\", which=\"both\", labelsize=11, direction=\"in\", width=0.5\n",
")\n",
"ax.xaxis.set_ticks_position(\"both\")\n",
"ax.yaxis.set_ticks_position(\"both\")\n",
"for axis in [\"top\", \"bottom\", \"left\", \"right\"]:\n",
" ax.spines[axis].set_linewidth(0.5)\n",
"\n",
"ax.semilogy(\n",
" *limits_NG_sqrt.T,\n",
" label=r\"NANOGrav limits\",\n",
" color=\"green\",\n",
" marker=\"+\",\n",
" lw=0,\n",
")\n",
"\n",
"ax.semilogy(\n",
" np.arange(1, l_max_lin + 1),\n",
" np.quantile(\n",
" CL_linear_prior[1:] / CL_linear_prior[0, :],\n",
" limit_cl,\n",
" axis=-1,\n",
" ),\n",
" \"blue\",\n",
" ls=\"--\",\n",
" label=r\"Only prior\",\n",
")\n",
"\n",
"# ax.set_title(r\"NANOGrav positions\")\n",
"plt.legend(loc=\"lower left\", fontsize=10, handlelength=1.5)\n",
"\n",
"ax.text(\n",
" 4,\n",
" 2.6e-1,\n",
" r\"Square root basis\",\n",
" horizontalalignment=\"left\",\n",
" fontsize=10,\n",
" verticalalignment=\"top\",\n",
" bbox=dict(\n",
" boxstyle=\"round\",\n",
" facecolor=\"white\",\n",
" alpha=1,\n",
" linewidth=1,\n",
" edgecolor=\"0.8\",\n",
" ),\n",
" linespacing=1.4,\n",
")\n",
"\n",
"# set x axis label\n",
"ax.set_xlabel(r\"$\\ell$\")\n",
"ax.set_ylabel(r\"$C_\\ell / C_0$\")\n",
"plt.ylim(2e-2, 3e-1)\n",
"\n",
"plt.tight_layout()\n",
"plt.savefig(\"sqrt_prior_limits.pdf\")\n",
"plt.show(block=True)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
| 46,406 | Python | .py | 292 | 154.482877 | 39,162 | 0.896322 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,331 | MCMC_Fisher.ipynb | Mauropieroni_fastPTA/examples/MCMC_Fisher.ipynb | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [],
"source": [
"# Global\n",
"import numpy as np\n",
"\n",
"# Local\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"import fastPTA.plotting_functions as pf\n",
"from fastPTA.Fisher_code import compute_fisher\n",
"from fastPTA.MCMC_code import run_MCMC"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Constants to be used in the analysis"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
"# Total observation time in years\n",
"T_obs_yrs = 16.03\n",
"\n",
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# Number of pulsars in the analysis\n",
"n_pulsars = 25\n",
"\n",
"# Specify the type of noise to be used in the analysis\n",
"which_experiment = eu.EPTAlike\n",
"\n",
"# Set the label to specify the signal model\n",
"signal_label = \"power_law\"\n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Specify the true signal parameters\n",
"signal_parameters = np.array([log_amplitude, tilt])\n",
"\n",
"# Specify the labels for the signal parameters\n",
"parameter_labels = [r\"$\\alpha_{\\rm PL}$\", r\"$n_{\\rm T}$\"]\n",
"\n",
"# Specify the priors for the signal parameters (should be a 2D array with the columns running over the parameters)\n",
"priors = []\n",
"\n",
"# Number of points to generate for the Fisher\n",
"len_fisher_data = int(1e4)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Some inputs for the MCMC"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [],
"source": [
"# Whether the MCMC should be rerun\n",
"rerun_MCMC = True\n",
"\n",
"# Whether the MCMC data should be regenerated\n",
"regenerate_MCMC_data = True\n",
"\n",
"# Whether to generate a data realization or take data at face value\n",
"realization = False\n",
"\n",
"# Number of burnin steps for the MCMC\n",
"burnin_steps_default = 300\n",
"\n",
"# Maximum number of iterations to get R close to 1\n",
"i_max_default = 100\n",
"\n",
"# Accepted value for the Gellman-Rubin convergence criterion\n",
"R_convergence_default = 1e-1\n",
"\n",
"# Criterion to be used for the convergence of the MCMC chains\n",
"R_criterion_default = \"mean_squared\"\n",
"\n",
"# Number of MCMC steps in each iteration\n",
"MCMC_iteration_steps_default = 500\n",
"\n",
"# Path to the MCCM_data\n",
"path_to_MCMC_data = \"generated_data/MCMC_data.npz\"\n",
"\n",
"# Path to the MCMC chains\n",
"path_to_MCMC_chains = \"generated_chains/MCMC_chains.npz\""
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Set the inputs for the Fisher and MCMC analysis"
]
},
{
"cell_type": "code",
"execution_count": 22,
"metadata": {},
"outputs": [],
"source": [
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"generate_catalog_kwargs = {\n",
" \"n_pulsars\": n_pulsars,\n",
" \"save_catalog\": False,\n",
" **which_experiment,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors\n",
"get_tensors_kwargs = {\n",
" \"add_curn\": False,\n",
" \"regenerate_catalog\": True,\n",
"}\n",
"\n",
"# Dictionary with the kwargs for the Fisher matrix\n",
"fisher_kwargs = {\n",
" \"T_obs_yrs\": T_obs_yrs,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_label\": signal_label,\n",
" \"signal_parameters\": signal_parameters,\n",
"}\n",
"\n",
"# Dictionary with the kwargs for the MCMC\n",
"MCMC_kwargs = {\n",
" \"regenerate_MCMC_data\": regenerate_MCMC_data,\n",
" \"realization\": realization,\n",
" \"path_to_MCMC_data\": path_to_MCMC_data,\n",
" \"path_to_MCMC_chains\": path_to_MCMC_chains,\n",
" \"i_max\": i_max_default,\n",
" \"R_convergence\": R_convergence_default,\n",
" \"R_criterion\": R_criterion_default,\n",
" \"burnin_steps\": burnin_steps_default,\n",
" \"MCMC_iteration_steps\": MCMC_iteration_steps_default,\n",
"}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Prepare the model to scan over"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fisher errors [0.23784615 0.31128867]\n"
]
}
],
"source": [
"if \"regenerate_catalog\" in get_tensors_kwargs.keys():\n",
" if get_tensors_kwargs[\"regenerate_catalog\"]:\n",
" rerun_MCMC = True\n",
"\n",
"(\n",
" frequency,\n",
" signal,\n",
" HD_functions_IJ,\n",
" HD_coeffs,\n",
" effective_noise,\n",
" SNR,\n",
" fisher,\n",
") = compute_fisher(\n",
" **fisher_kwargs,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs,\n",
")\n",
"\n",
"covariance = ut.compute_inverse(fisher)\n",
"fisher_data = np.random.multivariate_normal(\n",
" signal_parameters, covariance, size=len_fisher_data\n",
")\n",
"errors = np.sqrt(np.diag(covariance))\n",
"print(\"Fisher errors\", errors)\n",
"\n",
"fisher_ranges = np.vstack(\n",
" (signal_parameters - 5 * errors, signal_parameters + 5 * errors)\n",
")\n",
"\n",
"get_tensors_kwargs[\"regenerate_catalog\"] = False"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Entered\n",
"\n",
"Regenerating MCMC data\n",
"- Will generate a data realization\n",
"\n",
"Initial run\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
" 0%| | 0/300 [00:00<?, ?it/s]"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|██████████| 300/300 [00:02<00:00, 133.60it/s]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Burn-in dropped, here starts the proper run\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|██████████| 500/500 [00:03<00:00, 142.63it/s]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"At this step R = 1.1598\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|██████████| 500/500 [00:03<00:00, 140.62it/s]"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"At this step R = 1.0676\n",
"This took 9.3 seconds \n",
"\n",
"Storing as generated_chains/MCMC_chains.npz\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"\n"
]
}
],
"source": [
"try:\n",
" if rerun_MCMC:\n",
" raise FileNotFoundError(\"Flag forces MCMC chains regeneration\")\n",
"\n",
" MCMC_results = np.load(path_to_MCMC_chains)\n",
" MCMC_data = MCMC_results[\"samples\"]\n",
" pdfs = MCMC_results[\"pdfs\"]\n",
"\n",
"except FileNotFoundError:\n",
"\n",
" print(\"Entered\")\n",
"\n",
" if not np.any(priors):\n",
" priors = np.array(fisher_ranges)\n",
" print(priors)\n",
"\n",
" MCMC_data, pdfs = run_MCMC(\n",
" priors,\n",
" **fisher_kwargs,\n",
" **MCMC_kwargs,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": 28,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 1000x800 with 4 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"# This part should be moved out of this function\n",
"datasets = [fisher_data, MCMC_data]\n",
"weights = [\n",
" np.ones(len_fisher_data),\n",
" np.ones(MCMC_data.shape[0]), # type: ignore\n",
"]\n",
"smooth = [1.0, 1.0]\n",
"ranges = np.array(fisher_ranges.T)\n",
"\n",
"pf.plot_corner(\n",
" datasets,\n",
" colors=[pf.my_colormap[\"red\"], pf.my_colormap[\"green\"]],\n",
" truths=signal_parameters,\n",
" chain_labels=[\"Fisher future\", \"MCMC future\"],\n",
" weights=weights,\n",
" smooth=smooth,\n",
" labels=parameter_labels,\n",
" range=ranges,\n",
" truth_color=\"black\",\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
| 78,037 | Python | .py | 387 | 196.692506 | 68,078 | 0.912966 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,332 | linear_basis_figure.ipynb | Mauropieroni_fastPTA/examples/linear_basis_figure.ipynb | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Import global libs\n",
"import os\n",
"import time\n",
"import tqdm\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"from multiprocessing import Pool, cpu_count\n",
"\n",
"# Import local libs\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"from fastPTA.Fisher_code import compute_fisher"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Check if all the required folders are in place"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"# If it's not there, create the folder to store the generated data\n",
"if not os.path.exists(\"generated_data/\"):\n",
" os.mkdir(\"generated_data/\")\n",
"\n",
"# If it's not there, create the folder to store the plots\n",
"if not os.path.exists(\"plots/\"):\n",
" os.mkdir(\"plots/\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Constants to generate the Fisher results"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"# Total observation time in years\n",
"T_obs_yrs = 16.03\n",
"\n",
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# Number of pulsars used in the analysis\n",
"n_pulsars = 68\n",
"\n",
"# Maximum value of l used in the analysis\n",
"l_max = 6 \n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Number of catalog realizations to be generated\n",
"N_realizations = 30\n",
"\n",
"# Whether you want to regenerate the fishers\n",
"regenerate_fishers = False\n",
"\n",
"# Name of the outfile, no need for the extension\n",
"# (will be stored in generated_data/)\n",
"outname = \"NGlike\""
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Constants for the plots"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"# Confidence level for the Cls\n",
"limit_cl = 0.95\n",
"\n",
"# Number of samples used for each realization\n",
"n_points = int(1e4)\n",
"\n",
"# Number of catalog realizations used in the analysis.\n",
"# If larger than N_realizations taxes that as max\n",
"N_realizations_in_plot = 32\n",
"\n",
"# Number of cores to be used for parallel computations\n",
"num_cores = cpu_count()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Build the injection "
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"# All multipoles here are set to zero\n",
"signal_lm = (\n",
" 1e-30 / np.sqrt(4 * np.pi) * np.ones(ut.get_n_coefficients_real(l_max))\n",
")\n",
"\n",
"# The monopole is set to 1 (with the right normalization)\n",
"signal_lm[0] = 1.0 / np.sqrt(4 * np.pi)\n",
"\n",
"# The two parameters characterizing the signal in frequency\n",
"signal_parameters = np.array([log_amplitude, tilt])\n",
"\n",
"# Len of the parameters used in the analysis\n",
"shape_params = len(signal_parameters)\n",
"\n",
"# The monopole is not included in the fisher analysis since it's degenerate\n",
"means = np.concatenate((signal_parameters, signal_lm[1:]))\n",
"\n",
"# The number of parameters used in the fisher analysis\n",
"n_params = len(means)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Define the inputs for compute_fisher"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"generate_catalog_kwargs = {\n",
" \"n_pulsars\": n_pulsars,\n",
" \"save_catalog\": False,\n",
" \"use_ng_positions\" : True,\n",
" **eu.EPTAlike,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors \n",
"get_tensors_kwargs = {\n",
" \"path_to_pulsar_catalog\": \"pulsar_configurations/EPTA68.txt\",\n",
" \"add_curn\": False,\n",
" \"regenerate_catalog\": True,\n",
" \"anisotropies\": True,\n",
" \"l_max\": l_max,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate the fisher matrix\n",
"fisher_kwargs = {\n",
" \"T_obs_yrs\": T_obs_yrs,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_parameters\": signal_parameters,\n",
" \"signal_lm\": signal_lm,\n",
"}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Try to load the fisher data, if not there generate"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"try:\n",
" if regenerate_fishers:\n",
" raise FileNotFoundError(\"Flag forces regeneration\")\n",
"\n",
" fisher = np.load(\"generated_data/\" + outname + \".npz\")[\"all_fishers\"]\n",
"\n",
"except FileNotFoundError:\n",
" fisher = np.zeros((N_realizations, n_params, n_params))\n",
"\n",
" for index in tqdm.tqdm(range(N_realizations)):\n",
" res = compute_fisher(\n",
" **fisher_kwargs,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs\n",
" )[-1]\n",
"\n",
" # Delete the monopole since it's degenerate\n",
" fisher[index] = np.delete(np.delete(res, 2, axis=-1), 2, axis=-2)\n",
"\n",
" np.savez(\"generated_data/\" + outname + \".npz\", all_fishers=fisher)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Try to load the limits, if not there generate"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [],
"source": [
"try:\n",
" Cl_limits = np.loadtxt(\"generated_data/Cl_limits_\" + outname + \".dat\")\n",
" Cl_limits_prior = np.loadtxt(\n",
" \"generated_data/Cl_limits_prior_\" + outname + \".dat\"\n",
" )\n",
"\n",
" if len(Cl_limits) != N_realizations:\n",
" raise FileNotFoundError(\"Flag forces regeneration\")\n",
"\n",
"except FileNotFoundError:\n",
" cov = ut.compute_inverse(fisher)\n",
"\n",
" for j in range(min(N_realizations_in_plot, N_realizations)):\n",
" arg_grid = [\n",
" [means, cov[j]]\n",
" for j in range(min(N_realizations_in_plot, N_realizations))\n",
" ]\n",
"\n",
" def get_Cl_wrapper(parameters):\n",
" return ut.get_Cl_limits(\n",
" *parameters,\n",
" shape_params,\n",
" n_points=n_points,\n",
" limit_cl=limit_cl,\n",
" max_iter=100,\n",
" prior=5.0 / (4.0 * np.pi)\n",
" )\n",
"\n",
" start = time.time()\n",
" with Pool(num_cores) as p:\n",
" res = np.array(p.map(get_Cl_wrapper, arg_grid, chunksize=1))\n",
" end = time.time()\n",
" print(\"Time to compute Cl limts:\", end - start)\n",
"\n",
" Cl_limits = res[:, 0]\n",
" Cl_limits_prior = res[:, 1]\n",
"\n",
" Cl_limits_prior = Cl_limits_prior[np.isfinite(Cl_limits_prior[:, 0])]\n",
" np.savetxt(\"generated_data/Cl_limits_\" + outname + \".dat\", Cl_limits)\n",
" np.savetxt(\n",
" \"generated_data/Cl_limits_prior_\" + outname + \".dat\", Cl_limits_prior\n",
" )\n",
"\n",
"# Normalize with C0\n",
"Cl_limits *= 4 * np.pi \n",
"Cl_limits_prior *= 4 * np.pi "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Compute mean and quantiles among the realizations"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [],
"source": [
"# Compute the mean and the errors on the Cls without prior \n",
"mean_Cl_limits = np.mean(Cl_limits, axis=0)\n",
"err_low_Cl_limits = mean_Cl_limits - np.quantile(Cl_limits, 0.025, axis=0)\n",
"err_high_Cl_limits = np.quantile(Cl_limits, 0.975, axis=0) - mean_Cl_limits\n",
"\n",
"# Compute the mean and the errors on the Cls with the prior \n",
"mean_Cl_limits_prior = np.mean(Cl_limits_prior, axis=0)\n",
"err_low_Cl_limits_prior = mean_Cl_limits_prior - np.quantile(\n",
" Cl_limits_prior, 0.025, axis=0\n",
")\n",
"err_high_Cl_limits_prior = (\n",
" np.quantile(Cl_limits_prior, 0.975, axis=0) - mean_Cl_limits_prior\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Compute the limits due to the prior only"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Just prior:\n",
"C_l / C_0 = [1.27683451 1.11983646 1.0424372 1.00530637 0.9666254 0.94391062]\n"
]
}
],
"source": [
"data_prior = np.random.uniform(\n",
" -5.0 / (4.0 * np.pi), 5.0 / (4.0 * np.pi), (n_points, len(means[1:]))\n",
")\n",
"\n",
"Cl_prior_values = ut.get_CL_from_real_clm(data_prior.T)[1:]\n",
"\n",
"Cl_prior = np.quantile(Cl_prior_values, limit_cl, axis=-1)\n",
"\n",
"Cl_prior *= 4 * np.pi\n",
"\n",
"print(\"Just prior:\")\n",
"print(\"C_l / C_0 =\", Cl_prior)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Load the NANOgrav constraints"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"limits_NG = np.loadtxt(\"data_paper_2/limits_Cl_powerlaw_lin_ng15.dat\")[:, 1]"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 720x660 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.rc(\"text\", usetex=True)\n",
"plt.rc(\"font\", family=\"serif\", size=11)\n",
"plt.rcParams[\"text.latex.preamble\"] = (\n",
" r\"\\usepackage{amsmath} \\usepackage{amssymb}\"\n",
")\n",
"fig = plt.figure(figsize=(0.4 * 12.0, 0.4 * 11.0), dpi=150, edgecolor=\"white\")\n",
"ax = fig.add_subplot(1, 1, 1)\n",
"ax.tick_params(\n",
" axis=\"both\", which=\"both\", labelsize=11, direction=\"in\", width=0.5\n",
")\n",
"ax.xaxis.set_ticks_position(\"both\")\n",
"ax.yaxis.set_ticks_position(\"both\")\n",
"\n",
"\n",
"ax.semilogy(\n",
" np.arange(1, l_max + 1),\n",
" limits_NG,\n",
" label=r\"NANOGrav limits\",\n",
" color=\"green\",\n",
" marker=\"x\",\n",
" lw=0,\n",
")\n",
"\n",
"ax.semilogy(\n",
" np.arange(1, l_max + 1),\n",
" Cl_prior,\n",
" \"blue\",\n",
" label=r\"Only prior\",\n",
")\n",
"\n",
"ax.errorbar(\n",
" np.arange(1, l_max + 1),\n",
" mean_Cl_limits,\n",
" yerr=(err_low_Cl_limits, err_high_Cl_limits),\n",
" color=\"red\",\n",
" label=r\"Fisher\",\n",
" marker=\".\",\n",
")\n",
"\n",
"ax.errorbar(\n",
" np.arange(1, l_max + 1),\n",
" mean_Cl_limits_prior,\n",
" yerr=(err_low_Cl_limits_prior, err_high_Cl_limits_prior),\n",
" color=\"grey\",\n",
" label=r\"Fisher with prior\",\n",
" marker=\".\",\n",
")\n",
"\n",
"\n",
"plt.legend(loc=\"lower right\", fontsize=10, handlelength=1.5)\n",
"props = dict(\n",
" boxstyle=\"round\",\n",
" facecolor=\"white\",\n",
" alpha=1,\n",
" linewidth=1,\n",
" edgecolor=\"0.8\",\n",
")\n",
"\n",
"for axis in [\"top\", \"bottom\", \"left\", \"right\"]:\n",
" ax.spines[axis].set_linewidth(0.5)\n",
"\n",
" ax.text(\n",
" 1,\n",
" 4.8e0,\n",
" r\"NANOGrav positions, $T_\\mathrm{obs} \\sim 15 \\, \\mathrm{yr}$\",\n",
" horizontalalignment=\"left\",\n",
" fontsize=10,\n",
" verticalalignment=\"top\",\n",
" bbox=props,\n",
" linespacing=1.4,\n",
" )\n",
"\n",
" # set x axis label\n",
" ax.set_xlabel(r\"$\\ell$\")\n",
" ax.set_ylabel(r\"$C_\\ell / C_0$\")\n",
" plt.ylim(0.8e-1, 6e0)\n",
"\n",
" plt.tight_layout()\n",
" plt.savefig(\"Cl_sens_lim_ng_like.pdf\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
| 64,891 | Python | .py | 496 | 126.290323 | 51,046 | 0.873267 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,333 | examples_utils.py | Mauropieroni_fastPTA/examples/examples_utils.py | # Global
import os
# Local
import fastPTA.utils as ut
# Creates some folders you need to store data/plots
for k in [
"pulsar_configurations",
"generated_data",
"generated_chains",
"plots",
]:
if not os.path.exists(k):
os.makedirs(k)
path_to_file = os.path.dirname(__file__)
path_to_pulsar_parameters = os.path.join(path_to_file, "pulsar_configurations/")
# Default parameters for the pulsars
EPTAlike = ut.load_yaml(
path_to_pulsar_parameters + "EPTAlike_pulsar_parameters.yaml"
)
# Default parameters for the pulsars
mockSKA10 = ut.load_yaml(
path_to_pulsar_parameters + "mockSKA10_pulsar_parameters.yaml"
)
# Default parameters for the pulsars
EPTAlike_noiseless = ut.load_yaml(
path_to_pulsar_parameters + "EPTAlike_pulsar_parameters_noiseless.yaml"
)
| 804 | Python | .py | 27 | 26.962963 | 80 | 0.746094 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,334 | T_scaling_figure.ipynb | Mauropieroni_fastPTA/examples/T_scaling_figure.ipynb | {
"cells": [
{
"cell_type": "markdown",
"id": "3d2a5638",
"metadata": {},
"source": [
"### Start importing all libraries"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "a061b9e8",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
]
}
],
"source": [
"# Global\n",
"import os\n",
"import tqdm\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"\n",
"# Local\n",
"import examples_utils as eu\n",
"import fastPTA.utils as ut\n",
"import fastPTA.plotting_functions as pf\n",
"from fastPTA.Fisher_code import compute_fisher"
]
},
{
"cell_type": "markdown",
"id": "8c55c5d3",
"metadata": {},
"source": [
"### Constants to be used in the analysis"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "9afe1b2e",
"metadata": {},
"outputs": [],
"source": [
"# Number of frequencies used in the analysis\n",
"n_frequencies = 30\n",
"\n",
"# Number of pulsars in the analysis\n",
"n_pulsars = 50\n",
"\n",
"# The analysis assumes a power law template, specify here the input parameters\n",
"log_amplitude = -7.1995 # log amplitude\n",
"tilt = 2.0 # Tilt\n",
"\n",
"# Parameters for the HD computations:\n",
"# Method to compute the HD, either \"Legendre\" or \"Binned\"\n",
"HD_basis = \"Legendre\"\n",
"\n",
"# Maximum order of the Legendre polynomials\n",
"HD_order = 0\n",
"\n",
"# Whether to add some gaussian prior for the HD coefficients to the Fisher matrix\n",
"add_HD_prior = False\n",
"\n",
"# Specify the type of noise to be used in the analysis\n",
"which_experiment = eu.EPTAlike\n",
"\n",
"# Whether you want to rerun the analysis\n",
"rerun = False\n",
"\n",
"# Minimum number of years to be used in the analysis\n",
"T_min_yrs = .1\n",
"\n",
"# Maximum number of years to be used in the analysis\n",
"T_max_yrs = 200\n",
"\n",
"# Number of points in T to scan over\n",
"T_times = 15\n",
"\n",
"# Number of realizations to be generated for each value of T\n",
"N_realizations = 10\n",
"\n",
"# Name of the outfile, no need for the extension\n",
"# (will be stored in generated_data/)\n",
"outname = \"Default\"\n",
"\n",
"# Labels for the signal parameters\n",
"signal_labels = [r\"$\\alpha_{*}$\", r\"$n_{\\rm T}$\"]"
]
},
{
"cell_type": "markdown",
"id": "9130a943",
"metadata": {},
"source": [
"### Builds the dictionaries with inputs for the code"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "e1c85ecb",
"metadata": {},
"outputs": [],
"source": [
"# Assemble the vector with two signal parameters\n",
"signal_parameters = np.array([log_amplitude, tilt])\n",
"\n",
"# Length of the parameter vector\n",
"parameter_len = (\n",
" len(signal_parameters) + HD_order + 1\n",
" if HD_order\n",
" else len(signal_parameters)\n",
")\n",
"\n",
"# Dictionary with the kwargs to generete the pulsar catalogs\n",
"default_pulsars = {\n",
" \"n_pulsars\": n_pulsars,\n",
" \"save_catalog\": False,\n",
" **which_experiment,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate noise and orf tensors\n",
"get_tensors_kwargs = {\n",
" \"add_curn\": False,\n",
" \"HD_order\": HD_order,\n",
" \"HD_basis\": HD_basis,\n",
" \"regenerate_catalog\": True,\n",
"}\n",
"\n",
"# Dictionary with the kwargs to generate the fisher matrix\n",
"fisher_kwargs = {\n",
" \"T_obs_yrs\": 1.,\n",
" \"n_frequencies\": n_frequencies,\n",
" \"signal_parameters\": signal_parameters,\n",
"}"
]
},
{
"cell_type": "markdown",
"id": "24800e4c",
"metadata": {},
"source": [
"### Check if all the folders are in place and define save_path"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "5ee5ac36",
"metadata": {},
"outputs": [],
"source": [
"# If it's not there, create the folder to store the generated data\n",
"if not os.path.exists(\"generated_data/\"):\n",
" os.mkdir(\"generated_data/\")\n",
"\n",
"# If it's not there, create the folder to store the plots\n",
"if not os.path.exists(\"plots/\"):\n",
" os.mkdir(\"plots/\")\n",
"\n",
"# Build the save path\n",
"if outname != \"Default\":\n",
" save_path = \"generated_data/\" + outname + \".npz\"\n",
"elif HD_order == 0:\n",
" save_path = \"generated_data/T_scaling.npz\"\n",
"else:\n",
" save_path = \"generated_data/T_scaling_HD.npz\""
]
},
{
"cell_type": "markdown",
"id": "9f847562",
"metadata": {},
"source": [
"### Run for all values of T_obs"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "6b9de97a",
"metadata": {},
"outputs": [],
"source": [
"try:\n",
" # Check if the file is there and load the data if not rerun\n",
" if rerun:\n",
" raise FileNotFoundError(\"Forcing regeneration\")\n",
"\n",
" data = np.load(save_path)\n",
"\n",
" T_obs_values = data[\"T_obs_values\"]\n",
" SNR_values = data[\"SNR_values\"]\n",
" parameters_uncertainties = data[\"parameters_uncertainties\"]\n",
"\n",
"except FileNotFoundError:\n",
"\n",
" # Define the vector with the values of T_obs to scan over\n",
" T_obs_values = np.geomspace(T_min_yrs, T_max_yrs, T_times)\n",
"\n",
" # Initialize the arrays to store the results\n",
" SNR_values = np.zeros(shape=(T_times, N_realizations))\n",
" parameters_uncertainties = np.zeros(\n",
" shape=(T_times, N_realizations, parameter_len)\n",
" )\n",
"\n",
" # Loop over the values of T_obs\n",
" for i in range(T_times):\n",
" # Set the value of T_obs\n",
" generate_catalog_kwargs = default_pulsars.copy()\n",
" generate_catalog_kwargs[\"T_span_dict\"] = {\n",
" \"which_distribution\": \"gaussian\",\n",
" \"mean\": np.log10(T_obs_values[i]),\n",
" \"std\": 0.0125,\n",
" }\n",
"\n",
" # Compute the fisher for all the realizations at a given T_obs\n",
" print(\"Here starts T = %.2f yrs\" % (T_obs_values[i]))\n",
" for j in tqdm.tqdm(range(N_realizations)):\n",
" (\n",
" frequency,\n",
" signal,\n",
" HD_functions_IJ,\n",
" HD_coeffs,\n",
" effective_noise,\n",
" SNR,\n",
" fisher,\n",
" ) = compute_fisher(\n",
" T_obs_yrs=T_obs_values[i],\n",
" n_frequencies=n_frequencies + int(T_obs_values[i]),\n",
" signal_label=\"power_law\",\n",
" signal_parameters=signal_parameters,\n",
" get_tensors_kwargs=get_tensors_kwargs,\n",
" generate_catalog_kwargs=generate_catalog_kwargs,\n",
" )\n",
"\n",
" if HD_order and add_HD_prior:\n",
" fisher += np.diag(\n",
" np.append(\n",
" np.zeros(len(signal_parameters)), np.ones(HD_order + 1)\n",
" )\n",
" )\n",
"\n",
" # Get covariance matrix and errors\n",
" c_inverse = ut.compute_inverse(fisher)\n",
" errors = np.sqrt(np.diag(c_inverse))\n",
"\n",
" # Store the results\n",
" SNR_values[i, j] = SNR\n",
" parameters_uncertainties[i, j] = errors\n",
"\n",
" SNR_mean = np.mean(SNR_values[i], axis=-1)\n",
" SNR_std = np.std(SNR_values[i], axis=-1)\n",
" print(\"SNR=%.2e +-%.2e \\n\" % (SNR_mean, SNR_std))\n",
"\n",
" to_save = {\n",
" \"T_obs_values\": T_obs_values,\n",
" \"SNR_values\": SNR_values,\n",
" \"parameters_uncertainties\": parameters_uncertainties,\n",
" }\n",
"\n",
" np.savez(save_path, **to_save)"
]
},
{
"cell_type": "markdown",
"id": "7542705d",
"metadata": {},
"source": [
"### Compute means and stds"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "cc04418c",
"metadata": {},
"outputs": [],
"source": [
"# Mean and std of the SNR over the realizations\n",
"SNR_mean = np.mean(SNR_values, axis=-1)\n",
"SNR_std = np.std(SNR_values, axis=-1)\n",
"\n",
"# Mean and std of the uncertainties over the realizations\n",
"uncertainties_means = np.mean(parameters_uncertainties, axis=1)\n",
"uncertainties_stds = np.std(parameters_uncertainties, axis=1)"
]
},
{
"cell_type": "markdown",
"id": "65963783",
"metadata": {},
"source": [
"### Plot the scaling of SNR with T_obs"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "c67b6dfe",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.figure(figsize=(8, 5))\n",
"\n",
"plt.errorbar(\n",
" T_obs_values,\n",
" SNR_mean,\n",
" yerr=SNR_std,\n",
" color=pf.my_colormap[\"cyan\"],\n",
" fmt=\"o\",\n",
" markersize=4,\n",
" linestyle=\"dashed\",\n",
" capsize=7,\n",
")\n",
"\n",
"# plot x**3\n",
"plt.loglog(T_obs_values, 0.1 * T_obs_values**3, linestyle=\"--\", color=\"black\")\n",
"plt.text(1, 5e-3, s=r\"$\\propto T^3$\", fontsize=15)\n",
"\n",
"# plot x**(1/2)\n",
"plt.loglog(T_obs_values, 4 * T_obs_values**0.5, linestyle=\"--\", color=\"black\")\n",
"plt.text(30, 1.5, s=r\"$\\propto \\sqrt{T}$\", fontsize=15)\n",
"\n",
"plt.xlabel(r\"$T_{\\rm obs} \\rm [yr]$\")\n",
"plt.ylabel(r\"$\\rm SNR$\")\n",
"plt.ylim(5e-6, 1e2)\n",
"plt.tight_layout()\n",
"plt.savefig(\"plots/SNR_T_scaling.pdf\")"
]
},
{
"cell_type": "markdown",
"id": "3f74d4ce",
"metadata": {},
"source": [
"### Plot the scaling of uncertainties on the SGWB shape parameters with T_obs"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "25846831",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Figure size 800x500 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plt.figure(figsize=(8, 5))\n",
"for i in range(len(signal_parameters)):\n",
" colors = list(pf.my_colormap.keys())\n",
" col = colors[np.mod(i, len(colors))]\n",
" plt.errorbar(\n",
" T_obs_values,\n",
" uncertainties_means[:, i],\n",
" yerr=uncertainties_stds[:, i],\n",
" color=col,\n",
" fmt=\"o\",\n",
" markersize=4,\n",
" linestyle=\"dashed\",\n",
" capsize=7,\n",
" label=signal_labels[i],\n",
" )\n",
"\n",
"plt.loglog(\n",
" T_obs_values,\n",
" 100 / T_obs_values**3,\n",
" linestyle=\"--\",\n",
" color=\"black\",\n",
")\n",
"\n",
"plt.text(0.5, 1e4, s=r\"$\\propto 1/T^3$\", fontsize=15)\n",
"\n",
"plt.loglog(\n",
" T_obs_values,\n",
" 0.25 / T_obs_values ** (1 / 2),\n",
" linestyle=\"--\",\n",
" color=\"black\",\n",
")\n",
"\n",
"plt.text(30, 1, s=r\"$\\propto 1/\\sqrt{T}$\", fontsize=15)\n",
"\n",
"plt.xlabel(r\"$T_{\\rm obs} \\rm [yr]$\")\n",
"plt.ylabel(r\"$\\rm Uncertainties$\")\n",
"plt.ylim(5e-3, 1e6)\n",
"plt.legend(fontsize=15)\n",
"plt.tight_layout()\n",
"plt.savefig(\"plots/Error_T_scaling.pdf\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "49f83d20",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.9.12 ('env_GWFast')",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
},
"vscode": {
"interpreter": {
"hash": "37bcb7777d3229dfb79a16124a08c500a91162130c755ccd3fb5278a92e948b4"
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}
| 116,712 | Python | .py | 456 | 251.368421 | 60,294 | 0.917449 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,335 | signals.py | Mauropieroni_fastPTA/fastPTA/signals.py | # Global
import jax
import jax.numpy as jnp
# Local
import fastPTA.utils as ut
jax.config.update("jax_enable_x64", True)
# If you want to use your GPU change here
jax.config.update("jax_default_device", jax.devices("cpu")[0])
# Current SMBBH SGWB log_amplitude best-fit
SMBBH_log_amplitude = -7.1995
SMBBH_tilt = 2
# Current SMBBH SGWB parameters
SMBBH_parameters = jnp.array([SMBBH_log_amplitude, SMBBH_tilt])
# A value for a flat spectrum
CGW_flat_parameters = jnp.array([-7.0])
# Some values for a LN spectrum
LN_log_amplitude = -6.45167492
LN_log_width = -0.91240383
LN_log_pivot = -7.50455732
CGW_LN_parameters = jnp.array([LN_log_amplitude, LN_log_width, LN_log_pivot])
# Some values for a BPL spectrum
BPL_log_amplitude = -5.8
BPL_log_width = -7.3
BPL_tilt_1 = 3
BPL_tilt_2 = 1.5
CGW_BPL_parameters = jnp.array(
[BPL_log_amplitude, BPL_log_width, BPL_tilt_1, BPL_tilt_2]
)
def flat(frequency, parameters):
"""
Generate a flat spectrum.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the flat spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed flat spectrum.
"""
return 10 ** parameters[0] * frequency**0
def dflat(index, frequency, parameters):
"""
Derivative of the flat spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the flat spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the flat spectrum with
respect to the specified parameter.
"""
if index == 0:
return flat(frequency, parameters) * jnp.log(10)
else:
raise ValueError("Cannot use that for this signal")
def power_law(frequency, parameters, pivot=ut.f_yr):
"""
Generate a power law spectrum.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the power law spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed power law spectrum.
"""
# unpack parameters
log_amplitude, tilt = parameters
return 10**log_amplitude * (frequency / pivot) ** tilt
def dpower_law(index, frequency, parameters, pivot=ut.f_yr):
"""
Derivative of the power law spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the power law spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the power law spectrum with
respect to the specified parameter.
"""
# unpack parameters
log_amplitude, tilt = parameters
model = power_law(frequency, parameters)
if index == 0:
dlog_model = jnp.log(10)
elif index == 1:
dlog_model = jnp.log(frequency / pivot)
else:
raise ValueError("Cannot use that for this signal")
return model * dlog_model
def lognormal(frequency, parameters):
"""
Generate a lognormal spectrum.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the lognormal spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed lognormal spectrum.
"""
# unpack parameters
log_amplitude, log_width, log_pivot = parameters
return 10**log_amplitude * jnp.exp(
-0.5 * (jnp.log(frequency / (10**log_pivot)) / 10**log_width) ** 2
)
def dlognormal(index, frequency, parameters):
"""
Derivative of the lognormal spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the lognormal spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the lognormal spectrum with
respect to the specified parameter.
"""
# unpack parameters
log_amplitude, log_width, log_pivot = parameters
model = lognormal(frequency, parameters)
if index == 0:
dlog_model = jnp.log(10)
elif index == 1:
dlog_model = (
jnp.log(10)
/ 10 ** (2 * log_width)
* (jnp.log(frequency / 10**log_pivot)) ** 2
)
elif index == 2:
dlog_model = (
jnp.log(10)
/ 10 ** (2 * log_width)
* (jnp.log(frequency / 10**log_pivot))
)
else:
raise ValueError("Cannot use that for this signal")
return model * dlog_model
def SMBH_and_lognormal(frequency, parameters):
"""
Generate a spectrum combining the SMBH and lognormal models.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the SMBH and lognormal spectra.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed SMBH and lognormal spectra.
"""
return power_law(frequency, parameters[:2]) + lognormal(
frequency, parameters[2:]
)
def dSMBH_and_lognormal(index, frequency, parameters):
"""
Derivative of the SMBH + lognormal spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the SMBH + lognormal spectra.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the SMBH + lognormal
spectra with respect to the specified parameter.
"""
if index < 2:
return dpower_law(index, frequency, parameters[:2])
else:
return dlognormal(index - 2, frequency, parameters[2:])
def broken_power_law(frequency, parameters, smoothing=1.5):
"""
Generate a broken power law spectrum.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the broken power law spectrum.
smoothing: float, optional
Some parameter controlling how smooth the transition is in the BPL
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed broken power law spectrum.
"""
# unpack parameters
alpha, gamma, a, b = parameters
x = frequency / 10**gamma
return (
10**alpha
* (jnp.abs(a) + jnp.abs(b)) ** smoothing
/ (
jnp.abs(b) * x ** (-a / smoothing)
+ jnp.abs(a) * x ** (b / smoothing)
)
** smoothing
)
def dbroken_power_law(index, frequency, parameters, smoothing=1.5):
"""
Derivative of the BPL spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the BPL spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the BPL spectrum with
respect to the specified parameter.
"""
# unpack parameters
alpha, gamma, a, b = parameters
model = broken_power_law(frequency, parameters)
if index != 0:
x = frequency / 10**gamma
if index == 0:
dlog_model = jnp.log(10)
elif index == 1:
dlog_model = (
jnp.log(10)
* a
* b
* (-1 + x ** ((a + b) / smoothing))
/ (b + a * x ** ((a + b) / smoothing))
)
elif index == 2:
dlog_model = (
-((b * smoothing) / (a + b))
+ (b * (smoothing + a * jnp.log(x)))
/ (b + a * x ** ((a + b) / smoothing))
) / a
elif index == 3:
dlog_model = (
-a
* (
smoothing
- x ** ((a + b) / smoothing)
* (smoothing + (a + b) * jnp.log(x))
)
/ ((a + b) * (b + a * x ** ((a + b) / smoothing)))
)
elif index == 4:
dlog_model = (
jnp.log(a + b)
+ (a * b * (-1 + x ** ((a + b) / smoothing)) * jnp.log(x))
/ (smoothing * (b + a * x ** ((a + b) / smoothing)))
- jnp.log(b / (x) ** (a / smoothing) + a * (x) ** (b / smoothing))
)
else:
raise ValueError("Cannot use that for this signal")
return model * dlog_model
def SMBH_and_broken_power_law(frequency, parameters):
"""
Generate a spectrum combining the SMBH and BPL models.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the lognormal spectrum.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed lognormal spectrum.
"""
return power_law(frequency, parameters[:2]) + broken_power_law(
frequency, parameters[2:]
)
def dSMBH_and_broken_power_law(index, frequency, parameters):
"""
Derivative of the SMBH + BPL spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the SMBH + BPL spectra.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the SMBH + BPL
spectra with respect to the specified parameter.
"""
if index < 2:
return dpower_law(index, frequency, parameters[:2])
else:
return dbroken_power_law(index - 2, frequency, parameters[2:])
def SMBH_and_flat(frequency, parameters):
"""
Generate a spectrum combining the SMBH and flat models.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the SMBH and lognormal spectra.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed SMBH and lognormal spectra.
"""
return power_law(frequency, parameters[:2]) + flat(
frequency, parameters[2:]
)
def dSMBH_and_flat(index, frequency, parameters):
"""
Derivative of the SMBH + flat spectrum.
Parameters:
-----------
index : int
Index of the parameter to differentiate.
frequency : numpy.ndarray or jax.numpy.ndarray
Array containing frequency bins.
parameters : numpy.ndarray or jax.numpy.ndarray
Array containing parameters for the SMBH + flat spectra.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the computed derivative of the SMBH + flat
spectra with respect to the specified parameter.
"""
if index < 2:
return dpower_law(index, frequency, parameters[:2])
else:
return dflat(index - 2, frequency, parameters[2:])
def get_model(signal_label):
"""
Retrieve signal and derivative models based on the specified label.
Parameters:
-----------
signal_label : str
Label indicating the type of signal model.
Returns:
--------
dict
Dictionary containing the signal model and its derivative model.
Notes:
------
Supported signal labels are:
- "power_law":
Power law signal model.
- "lognormal":
Log-normal signal model.
- "power_law_flat":
Signal model combining SMBH and flat spectrum.
- "power_law_lognormal":
Signal model combining SMBH and log-normal spectrum.
- "power_law_broken_power_law":
Signal model combining SMBH and broken power law spectrum.
"""
if signal_label == "power_law":
signal = {"signal_model": power_law, "dsignal_model": dpower_law}
elif signal_label == "flat":
signal = {"signal_model": flat, "dsignal_model": dflat}
elif signal_label == "lognormal":
signal = {"signal_model": lognormal, "dsignal_model": dlognormal}
elif signal_label == "power_law_flat":
signal = {
"signal_model": SMBH_and_flat,
"dsignal_model": dSMBH_and_flat,
}
elif signal_label == "power_law_lognormal":
signal = {
"signal_model": SMBH_and_lognormal,
"dsignal_model": dSMBH_and_lognormal,
}
elif signal_label == "power_law_broken_power_law":
signal = {
"signal_model": SMBH_and_broken_power_law,
"dsignal_model": dSMBH_and_broken_power_law,
}
else:
raise ValueError("Cannot use", signal_label)
return signal
| 14,264 | Python | .py | 413 | 27.784504 | 79 | 0.631709 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,336 | MCMC_code.py | Mauropieroni_fastPTA/fastPTA/MCMC_code.py | # Global
import time
import numpy as np
import emcee
import jax
import jax.numpy as jnp
# Local
import fastPTA.utils as ut
from fastPTA.signals import SMBBH_parameters, get_model
from fastPTA.get_tensors import get_tensors
jax.config.update("jax_enable_x64", True)
jax.config.update("jax_default_device", jax.devices(ut.which_device)[0])
# Setting some constants
i_max_default = 100
R_convergence_default = 1e-1
R_criterion_default = "mean_squared"
burnin_steps_default = 300
MCMC_iteration_steps_default = 500
def generate_gaussian(mean, sigma):
"""
This function generates complex data by sampling real and imaginary parts
independently from a Gaussian distribution with the specified mean and
standard deviation.
Parameters:
-----------
mean : numpy.ndarray or jax.numpy.ndarray
Mean of the Gaussian distribution.
sigma : numpy.ndarray or jax.numpy.ndarray
Standard deviation of the Gaussian distribution.
Returns:
--------
data : numpy.ndarray or jax.numpy.ndarray
Complex data sampled from a Gaussian distribution.
"""
# Generate the real part
real = np.random.normal(loc=mean, scale=sigma)
# Then the imaginary part
imaginary = np.random.normal(loc=mean, scale=sigma)
# Return the sum
return (real + 1j * imaginary) / np.sqrt(2)
def generate_MCMC_data(
realization,
frequency,
signal_std,
strain_omega,
response_IJ,
HD_functions_IJ,
HD_coeffs,
save_MCMC_data=True,
path_to_MCMC_data="generated_data/MCMC_data.npz",
):
"""
Generates (and might save) data for Markov Chain Monte Carlo (MCMC)
analysis based on the provided parameters. If `realization` is True,
it generates a realization of the data. Otherwise, it uses the expectation
value. Data contain both signal and noise. If `save_MCMC_data` is True,
the generated data is saved to a specified path.
Parameters:
-----------
realization : bool
Whether to generate a realization of the data.
frequency : numpy.ndarray
Array containing the frequency bins.
signal_std : numpy.ndarray
Array containing the standard deviation of the signal.
strain_omega : numpy.ndarray
Array containing the strain noise.
response_IJ : numpy.ndarray
Array containing the response function.
HD_functions_IJ : numpy.ndarray
Array containing the HD functions.
HD_coeffs : numpy.ndarray
Array containing the HD coefficients.
save_MCMC_data : bool, optional
Whether to save the generated data
Default is True
path_to_MCMC_data : str, optional
Path to save the generated data
Default is "generated_data/MCMC_data.npz"
Returns:
--------
Tuple containing:
- frequency: numpy.ndarray
Array containing frequency bins.
- data: numpy.ndarray
Array containing the generated data.
- response_IJ: numpy.ndarray
Array containing response function.
- strain_omega: numpy.ndarray
Array containing strain noise.
"""
if realization:
# Use this part if you want data to be realized
print("- Will generate a data realization\n")
# Sqrt of pulsar-pulsar noise term
noise_std = np.diagonal(np.sqrt(strain_omega), axis1=-2, axis2=-1)
# Generate gaussian data with the right std for the noise
noise_data = generate_gaussian(0, noise_std)
# Combine to get nn take the real part only since the likelihood is
# real and is multiplied by a 2 to keep track of that
noise_part = np.real(
noise_data[:, :, None] * np.conj(noise_data[:, None, :])
)
# Since we work with variables of f only (already integrated over theta
# and phi), each signal component should be generated independently
signal_std = signal_std[:, None, None] * np.ones_like(
np.real(noise_part)
)
# Generate gaussian data with the right std for the signal
signal_data = generate_gaussian(0, signal_std)
# The response projects the signal in the different pulsar combinations
# as for the noise, take the real part only
signal_part = response_IJ * np.real(
signal_data * np.conjugate(signal_data)
)
else:
# Use this part if you don't want data to be realized
print("- Data will use the expectation value\n")
noise_part = strain_omega
signal_part = response_IJ * (signal_std**2)[:, None, None]
# Combine signal and noise
data = signal_part + noise_part
# Save the data
if save_MCMC_data:
np.savez(
path_to_MCMC_data,
frequency=frequency,
data=data,
response_IJ=response_IJ,
strain_omega=strain_omega,
)
return frequency, data, response_IJ, strain_omega
def get_MCMC_data(
regenerate_MCMC_data,
T_obs_yrs=10.33,
n_frequencies=30,
signal_label="power_law",
signal_parameters=SMBBH_parameters,
realization=True,
save_MCMC_data=True,
path_to_MCMC_data="generated_data/MCMC_data.npz",
get_tensors_kwargs={},
generate_catalog_kwargs={},
):
"""
Loads or regenerate data for Markov Chain Monte Carlo (MCMC) analysis. If
`regenerate_MCMC_data` is True or the MCMC data file is not found, the data
is regenerated. Otherwise, it is loaded from the specified path. Data
contain both signal and noise. If the data are generated, if `realization`
is True, it generates a realization of the data. If `save_MCMC_data` is
True, the generated data is saved to a specified path. Additional keyword
arguments for get_tensors and generate_pulsars_catalog can be provided via
get_tensors_kwargs and generate_catalog_kwargs.
Parameters:
-----------
regenerate_MCMC_data : bool
Whether to regenerate the MCMC data.
T_obs_yrs : float, optional
Observation time in years
Default is 10.33.
n_frequencies : int, optional
Number of frequency bins
Default is 30.
signal_label : str, optional
Label indicating the type of signal model to use
Default is "power_law".
signal_parameters : dict, optional
Dictionary containing parameters for the signal model
Default is SMBBH_parameters.
realization : bool, optional
Whether to generate a realization of the data
Default is True.
save_MCMC_data : bool, optional
Whether to save the generated data
Default is True.
path_to_MCMC_data : str, optional
Path to save or load the generated data
Default is "generated_data/MCMC_data.npz".
get_tensors_kwargs : dict, optional
Additional keyword arguments for get_tensors
Default is {}.
generate_catalog_kwargs : dict, optional
Additional keyword arguments for generate_catalog_kwargs
Default is {}.
Returns:
--------
Tuple containing:
- frequency: numpy.ndarray
Array containing frequency bins.
- data: numpy.ndarray
Array containing the generated data.
- response_IJ: numpy.ndarray
Array containing response function.
- strain_omega: numpy.ndarray
Array containing strain noise.
"""
try:
if regenerate_MCMC_data:
raise FileNotFoundError("Flag forces MCMC data regeneration")
data = np.load(path_to_MCMC_data)
frequency = data["frequency"]
MCMC_data = data["data"]
response_IJ = data["response_IJ"]
strain_omega = data["strain_omega"]
except FileNotFoundError:
print("\nRegenerating MCMC data")
# Setting the frequency vector from the observation time
T_tot = T_obs_yrs * ut.yr
fmin = 1 / T_tot
frequency = fmin * (1 + np.arange(n_frequencies))
# Get the functions for the signal and its derivatives
model = get_model(signal_label)
signal_model = model["signal_model"]
# Computing (sqrt of the) signal
signal_std = np.sqrt(signal_model(frequency, signal_parameters))
# Gets all the ingredients to compute the fisher
strain_omega, response_IJ, HD_functions_IJ, HD_coeffs = get_tensors(
frequency, **get_tensors_kwargs, **generate_catalog_kwargs
)
# Generate MCMC data
frequency, MCMC_data, response_IJ, strain_omega = generate_MCMC_data(
realization,
frequency,
signal_std,
strain_omega,
response_IJ,
HD_functions_IJ,
HD_coeffs,
save_MCMC_data=save_MCMC_data,
path_to_MCMC_data=path_to_MCMC_data,
)
return frequency, MCMC_data, response_IJ, strain_omega
def flat_prior(parameter, min_val, max_val):
"""
Calculate the logarithm of a flat prior probability density function.
Parameters:
-----------
parameter : float or jax.numpy.ndarray
Parameter value(s) for which the prior probability is calculated.
min_val : float
Minimum allowed value for the parameter.
max_val : float
Maximum allowed value for the parameter.
Returns:
--------
float or jax.numpy.ndarray
Logarithm of the prior probability density function.
"""
if not (parameter >= min_val and parameter <= max_val):
return -jnp.inf
else:
return 0.0
def gaussian_prior(parameter, mean_val, std_val):
"""
Calculate the logarithm of a Gaussian prior probability density function.
Parameters:
-----------
parameter : float or jax.numpy.ndarray
Parameter value(s) for which the prior probability is calculated.
mean_val : float
Mean value of the Gaussian distribution.
std_val : float
Standard deviation of the Gaussian distribution.
Returns:
--------
float or jax.numpy.ndarray
Logarithm of the prior probability density function.
"""
return -0.5 * jnp.sum(
jnp.log(2 * jnp.pi * std_val**2)
+ ((parameter - mean_val) / std_val) ** 2
)
def log_prior(parameters, prior_parameters, which_prior="flat"):
"""
Calculate the logarithm of the prior probability density function. Only
flat and gaussian priors are supported at the moment
Parameters:
-----------
parameters : jax.numpy.ndarray
Array of parameter values for which the prior probability is calculated.
prior_parameters : numpy.ndarray
Array containing prior parameters, where each row represents the
parameters for a single prior.
For flat prior: [min_val, max_val]
For Gaussian prior: [mean_val, std_val]
which_prior : str, optional
Type of prior distribution to use
Default is "flat"
Returns:
--------
float
Logarithm of the prior probability density function.
"""
log_p = 0.0
for i in range(len(parameters)):
if which_prior == "flat":
log_p += flat_prior(parameters[i], *prior_parameters[:, i])
elif which_prior == "gaussian":
log_p += gaussian_prior(parameters[i], *prior_parameters[:, i])
return log_p
@jax.jit
def log_likelihood(
frequency,
signal_value,
response_IJ,
strain_omega,
data,
):
"""
Compute the logarithm of the likelihood assujming a Whittle likelihood.
Parameters:
-----------
frequency : numpy.ndarray
Array containing frequency bins.
signal_value : numpy.ndarray
Array containing the signal evaluated in all frequency bins.
response_IJ : numpy.ndarray
Array containing response function.
strain_omega : numpy.ndarray
Array containing strain noise.
data : numpy.ndarray
Array containing the observed data.
Returns:
--------
float
Logarithm of the likelihood.
"""
# Covariance of the data
covariance = response_IJ * signal_value[:, None, None] + strain_omega
# Inverse of the covariance
c_inverse = ut.compute_inverse(covariance)
# Log determinant
sign, logdet = jnp.linalg.slogdet(covariance)
# data term
data_term = jnp.einsum("ijk,ikj->i", c_inverse, data)
# return the likelihood
return -jnp.sum(logdet + data_term)
def log_posterior(
signal_parameters,
frequency,
signal_model,
response_IJ,
strain_omega,
data,
prior_parameters,
which_prior="flat",
):
"""
Compute the logarithm of the posterior probability summing log likelihood
and prior.
Parameters:
-----------
signal_parameters : numpy.ndarray
Array containing parameters of the signal model.
frequency : numpy.ndarray
Array containing frequency bins.
signal_model : callable
Function representing the signal model.
response_IJ : numpy.ndarray
Array containing response function.
strain_omega : numpy.ndarray
Array containing strain noise.
data : numpy.ndarray
Array containing the observed data.
prior_parameters : numpy.ndarray
Array containing prior parameters.
which_prior : str, optional
Type of prior distribution to use (default is "flat").
Returns:
--------
float
Logarithm of the posterior probability.
"""
lp = log_prior(signal_parameters, prior_parameters)
if not jnp.isfinite(lp):
return -jnp.inf
signal_value = signal_model(frequency, signal_parameters)
log_lik = log_likelihood(
frequency, signal_value, response_IJ, strain_omega, data
)
return lp + log_lik
@jax.jit
def log_likelihood_lm(
frequency,
signal_lm,
signal_value,
response_IJ,
strain_omega,
data,
):
"""
Compute the logarithm of the likelihood assujming a Whittle likelihood.
Parameters:
-----------
frequency : numpy.ndarray
Array containing frequency bins.
signal_value : numpy.ndarray
Array containing the signal evaluated in all frequency bins.
response_IJ : numpy.ndarray
Array containing response function.
strain_omega : numpy.ndarray
Array containing strain noise.
data : numpy.ndarray
Array containing the observed data.
Returns:
--------
float
Logarithm of the likelihood.
"""
signal_lm_f = signal_lm[:, None] * signal_value[None, :]
# Assemble the signal tensor
signal_tensor = jnp.sum(response_IJ * signal_lm_f[..., None, None], axis=0)
# Covariance of the data
covariance = signal_tensor + strain_omega
# Inverse of the covariance
c_inverse = ut.compute_inverse(covariance)
# Log determinant
sign, logdet = jnp.linalg.slogdet(covariance)
# data term
data_term = jnp.abs(jnp.einsum("ijk,ikj->i", c_inverse, data))
# return the likelihood
return -jnp.sum(logdet + data_term)
def log_posterior_lm(
signal_parameters,
frequency,
signal_model,
response_IJ,
strain_omega,
data,
prior_parameters,
which_prior="flat",
):
"""
Compute the logarithm of the posterior probability summing log likelihood
and prior.
Parameters:
-----------
signal_parameters : numpy.ndarray
Array containing parameters of the signal model.
frequency : numpy.ndarray
Array containing frequency bins.
signal_model : callable
Function representing the signal model.
response_IJ : numpy.ndarray
Array containing response function.
strain_omega : numpy.ndarray
Array containing strain noise.
data : numpy.ndarray
Array containing the observed data.
prior_parameters : numpy.ndarray
Array containing prior parameters.
which_prior : str, optional
Type of prior distribution to use (default is "flat").
Returns:
--------
float
Logarithm of the posterior probability.
"""
lp = log_prior(signal_parameters, prior_parameters)
if not jnp.isfinite(lp):
return -jnp.inf
signal_value = signal_model(frequency, signal_parameters[:2])
log_lik = log_likelihood_lm(
frequency,
signal_parameters[2:],
signal_value,
response_IJ,
strain_omega,
data,
)
return lp + log_lik
def run_MCMC(
priors,
T_obs_yrs=10.33,
n_frequencies=30,
signal_label="power_law",
signal_parameters=SMBBH_parameters,
initial=[],
which_prior="flat",
regenerate_MCMC_data=False,
realization=True,
save_MCMC_data=True,
path_to_MCMC_data="generated_data/MCMC_data.npz",
i_max=i_max_default,
R_convergence=R_convergence_default,
R_criterion=R_criterion_default,
burnin_steps=burnin_steps_default,
MCMC_iteration_steps=MCMC_iteration_steps_default,
path_to_MCMC_chains="generated_chains/MCMC_chains.npz",
get_tensors_kwargs={},
generate_catalog_kwargs={},
):
"""
Run Markov Chain Monte Carlo (MCMC) to estimate the posterior distribution
of the parameters given the observed data. The initial points are generated
randomly within the priors. After a burn-in phase, the Gelman-Rubin
statistic is used as a convergence diagnostic. Several MCMC iterations are
run until the chains reach convergence. MCMC samples and log posterior
probabilities are stored in the specified path.
Parameters:
-----------
priors : numpy.ndarray
Array containing prior parameters. Each row represents the parameters
for a single prior distribution.
T_obs_yrs : float, optional
Total observation time in years
Default is 10.33
n_frequencies : int, optional
Number of frequency bins
Default is 30
signal_label : str, optional
Label indicating the type of signal model to use
Default is "power_law"
signal_parameters : numpy.ndarray, optional
Array containing signal model parameters
Default is SMBBH_parameters
initial : list or numpy.ndarray, optional
Initial parameter values for the MCMC walkers
Default is empty
which_prior : str, optional
Type of prior distribution to use
Default is "flat"
regenerate_MCMC_data : bool, optional
Flag indicating whether to regenerate MCMC data
Default is False
realization : bool, optional
Flag indicating whether to generate a data realization
Default is True
save_MCMC_data : bool, optional
Flag indicating whether to save MCMC data
Default is True
path_to_MCMC_data : str, optional
Path to save MCMC data
Default is "generated_data/MCMC_data.npz"
i_max : int, optional
Maximum number of iterations for convergence
Default is i_max_default
R_convergence : float, optional
Convergence threshold for the Gelman-Rubin statistic
Default is R_convergence_default
R_criterion : str, optional
Criterion to calculate the Gelman-Rubin statistic
Default is R_criterion_default
burnin_steps : int, optional
Number of burn-in steps for the MCMC sampler
Default is burnin_steps_default
MCMC_iteration_steps : int, optional
Number of MCMC iteration steps
Default is MCMC_iteration_steps_default
path_to_MCMC_chains : str, optional
Path to save MCMC chains
Default is "generated_chains/MCMC_chains.npz"
get_tensors_kwargs : dict, optional
Additional keyword arguments for getting tensors
Default is an empty dictionary
generate_catalog_kwargs : dict, optional
Additional keyword arguments for generating the catalog
Default is an empty dictionary
Returns:
--------
Tuple containing:
- samples: numpy.ndarray
Array containing MCMC samples.
- pdfs: numpy.ndarray
Array containing log posterior probabilities for MCMC samples.
"""
if "anisotropies" in get_tensors_kwargs.keys():
anisotropies = get_tensors_kwargs["anisotropies"]
else:
anisotropies = False
# This is not used now
# if "lm_order" in get_tensors_kwargs.keys():
# lm_order = get_tensors_kwargs["lm_order"]
# else:
# lm_order = False
# Get the data
frequency, MCMC_data, response_IJ, strain_omega = get_MCMC_data(
regenerate_MCMC_data,
T_obs_yrs=T_obs_yrs,
n_frequencies=n_frequencies,
signal_label=signal_label,
signal_parameters=signal_parameters[:2],
realization=realization,
save_MCMC_data=save_MCMC_data,
path_to_MCMC_data=path_to_MCMC_data,
get_tensors_kwargs=get_tensors_kwargs,
generate_catalog_kwargs=generate_catalog_kwargs,
)
nwalkers = max(2 * len(priors.T), 5)
# Generate the initial points if not provided
if not initial and which_prior.lower() == "flat":
initial = np.random.uniform(
priors[0, :], priors[1, :], size=(nwalkers, len(priors.T))
)
elif not initial and which_prior.lower() == "gaussian":
initial = np.random.uniform(
priors[0, :], priors[1, :], size=(nwalkers, len(priors.T))
)
elif initial:
pass
else:
raise ValueError("Cannot use that prior")
nwalkers, ndims = initial.shape
# Args for the posterior
args = [
jnp.array(frequency),
get_model(signal_label)["signal_model"],
jnp.array(response_IJ),
jnp.array(strain_omega),
jnp.array(MCMC_data),
priors,
]
# Kwargs for the posterior
kwargs = {"which_prior": which_prior}
posterior_to_use = log_posterior_lm if anisotropies else log_posterior
# Set the sampler
sampler = emcee.EnsembleSampler(
nwalkers, ndims, posterior_to_use, args=args, kwargs=kwargs
)
start = time.perf_counter()
print("Initial run")
state = sampler.run_mcmc(initial, burnin_steps, progress=True)
sampler.reset()
print("Burn-in dropped, here starts the proper run")
R = 1e100
i = 0
# Run until convergence or until reached maximum number of iterations
while np.abs(R - 1) > R_convergence and i < i_max:
# Run this iteration
state = sampler.run_mcmc(state, MCMC_iteration_steps, progress=True)
# Get Gelman-Rubin at this step
R_array = ut.get_R(sampler.get_chain())
if R_criterion.lower() == "mean_squared":
R = np.sqrt(np.mean(R_array**2))
elif R_criterion.lower() == "max":
R = np.max(R_array)
else:
raise ValueError("Cannot use R_criterion =", R_criterion)
# state = None
print("At this step R = %.4f" % (R))
i += 1
# Samples and pdfs
samples = np.array(sampler.get_chain(flat=True))
pdfs = sampler.lnprobability
print("This took {0:.1f} seconds \n".format(time.perf_counter() - start))
print("Storing as", path_to_MCMC_chains)
np.savez(path_to_MCMC_chains, samples=samples, pdfs=pdfs) # type: ignore
return samples, pdfs
| 23,238 | Python | .py | 648 | 29.182099 | 80 | 0.666221 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,337 | Fisher_code.py | Mauropieroni_fastPTA/fastPTA/Fisher_code.py | # Global
import jax
import jax.numpy as jnp
# Local
import fastPTA.utils as ut
from fastPTA.signals import SMBBH_parameters, get_model
from fastPTA.get_tensors import get_tensors
# Set some global parameters for jax
jax.config.update("jax_enable_x64", True)
jax.config.update("jax_default_device", jax.devices(ut.which_device)[0])
# Default value for signal_lm
default_signal_lm = jnp.array([1.0 / jnp.sqrt(4 * jnp.pi)])
@jax.jit
def get_SNR_integrand(signal_tensor, c_inverse):
"""
Compute the integrand for the Signal-to-Noise Ratio (SNR) (for some set of
frequency bins).
Parameters:
-----------
signal_tensor : numpy.ndarray or jax.numpy.ndarray
3D array containing signal data, assumed to have shape (F, N, N), where
F is the number of frequency bins and N is the number of pulsars.
c_inverse : numpy.ndarray or jax.numpy.ndarray
3D array representing the inverse covariance matrix, assumed to have
shape (F, N, N), where F is the number of frequency bins and N is the
number of pulsars.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
Array containing the integrand of Signal-to-Noise Ratio (SNR)
(for each frequency bin).
"""
# Builnding the C matrix
c_bar_SNR = jnp.einsum("ijk,ikl->ijl", c_inverse, signal_tensor)
# Contracting the 2 matrixes
return jnp.einsum("ijk,ikj->i", c_bar_SNR, c_bar_SNR)
@jax.jit
def get_fisher_integrand(dsignal_tensor, c_inverse):
"""
Compute the integrand used to compute the Fisher Information Matrix (for
some set of frequency bins).
Parameters:
-----------
dsignal_tensor : numpy.ndarray or jax.numpy.ndarray
4D array containing the derivative of the signal data with respect to
the model parameters. This function assumes dsignal_tensor to have
shape (P, F, N, N), where P is the number of parameters, F is the number
of frequency bins, and N is the number of pulsars.
c_inverse : numpy.ndarray or jax.numpy.ndarray
3D array representing the inverse covariance matrix, assumed to have
shape (F, N, N), where F is the number of frequency bins and N is the
number of pulsars.
Returns:
--------
numpy.ndarray or jax.numpy.ndarray
3D array containing the integrand to compute the Fisher Information
Matrix for each combination of parameters. It has shape (N, N, F),
where N is the number of pulsars and F is the number of frequency bins.
"""
# Building the C matrix for the fisher
c_bar = jnp.einsum("ijk,aikl->aijl", c_inverse, dsignal_tensor)
# Contracting the 2 matrixes
return jnp.einsum("aijk,bikj->abi", c_bar, c_bar)
@jax.jit
def get_integrands(
signal,
dsignal,
response_IJ,
noise_tensor,
HD_functions_IJ,
):
"""
Compute integrands for Signal-to-Noise Ratio (SNR) and Fisher Information
Matrix given some signal data, derivatives and other quantities that
characterize the pulsar configuration used for the analysis
Parameters:
-----------
signal : numpy.ndarray or jax.numpy.ndarray
Array containing signal data.
dsignal : numpy.ndarray or jax.numpy.ndarray
2D array containing derivative of the signal data with respect to the
signal parameters. This function assumes dsignal to have shape (P, F)
where P is the number of parameters, F is the number of frequency bins
response_IJ : numpy.ndarray or jax.numpy.ndarray
3D array containing the response for all frequencies and pulsar pairs.
Assumed to have shape (F, N, N), where F is the number of frequency bins
and N is the number of pulsars.
noise_tensor : numpy.ndarray or jax.numpy.ndarray
3D array containing the noise for all frequencies and pulsar pairs.
Assumed to have shape (F, N, N), where F is the number of frequency bins
and N is the number of pulsars.
HD_functions_IJ : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre or binned projection of the Hellings and
Downs correlations. The shape is (HD_order + 1, F, N, N), where HD_order
is the maximum order of Legendre polynomials / bins, F is the number of
frequencies,and N is the number of pulsars.
Returns:
--------
Tuple containing:
- SNR_integrand: numpy.ndarray or jax.numpy.ndarray
the integrand for the Signal-to-Noise Ratio (SNR) computation.
- effective_noise: numpy.ndarray or jax.numpy.ndarray
the effective noise as a function of frequency.
- fisher_integrand: numpy.ndarray or jax.numpy.ndarray
3D array containing the integrand to compute the Fisher Information
Matrix for each combination of parameters. It has shape (N, N, F),
where N is the number of pulsars and F is the number of frequency bins.
"""
# Assemble the signal tensor
signal_tensor = response_IJ * signal[:, None, None]
# Build the covariance
covariance = signal_tensor + noise_tensor
# Invert the covariance
c_inverse = ut.compute_inverse(covariance)
# This is the SNR integrand
SNR_integrand = get_SNR_integrand(signal_tensor, c_inverse)
# Build the effective noise
effective_noise = jnp.sqrt(signal**2 / SNR_integrand)
# Assemble the tensor with signal derivativesz
dsignal_tensor = dsignal[..., None, None] * response_IJ[None, ...]
# Append HD functions
dsignal_tensor = jnp.concatenate(
(
dsignal_tensor,
signal[..., None, None] * HD_functions_IJ,
),
axis=0,
)
# Get the fisher integrand
fisher_integrand = get_fisher_integrand(dsignal_tensor, c_inverse)
return SNR_integrand, effective_noise, fisher_integrand
@jax.jit
def get_signal_dsignal_tensors_lm_spherical_harmonics_basis(
signal_lm, signal, dsignal, response_IJ
):
"""
Compute the signal and its derivatives in the spherical harmonics basis
Parameters:
-----------
signal_lm: numpy.ndarray or jax.numpy.ndarray
Array containing the lm coefficients.
signal : numpy.ndarray or jax.numpy.ndarray
Array containing the signal evaluated at all frequencies.
dsignal : numpy.ndarray or jax.numpy.ndarray
2D array containing derivative of the signal data with respect to the
signal parameters. This function assumes dsignal to have shape (P, F)
where P is the number of parameters, F is the number of frequency bins
response_IJ : numpy.ndarray or jax.numpy.ndarray
4D array containing response tensor for all the pulsars pairs.
It has shape (lm, F, N, N), where lm is the number of coefficients for
the anisotropy decomposition (spherical harmonics or sqrt basis), F is
the number of frequencies, N is the number of pulsars.
Returns:
--------
Tuple containing:
- signal_tensor: numpy.ndarray or jax.numpy.ndarray
the signal tensor in the spherical harmonics basis.
- dsignal_tensor_frequency_shape: numpy.ndarray or jax.numpy.ndarray
the derivative of the signal tensor with respect to the signal
parameters in frequency space.
"""
# Assemble the signal tensor, signal_lm are the lm coefficients of the
# signal and signal is the signal in frequency space (with len = F)
signal_lm_f = signal_lm[:, None] * signal[None, :]
# Assemble the signal tensor, the response_IJ tensor has shape
# (lm, N, pulsars, pulsars)
signal_tensor = jnp.sum(response_IJ * signal_lm_f[..., None, None], axis=0)
# Derivatives of signal_lm_f in parameters dsignal is the derivative of the
# signal in frequency space shape (P, N)
dsignal_lm_f = signal_lm[None, :, None] * dsignal[:, None, :]
# Assemble the tensor with signal derivatives with respect to the signal
# parameters, we thus sum over the lm coefficients
dsignal_tensor_frequency_shape = jnp.sum(
response_IJ[None, ...] * dsignal_lm_f[..., None, None], axis=1
)
# Just build a kronecher delta in (P, lm) space
delta = jnp.eye(len(signal_lm_f))
# Derivatives of signal_lm_f with respect to the lm coefficients
dsignal_lm_f_anisotropies = delta[..., None] * signal[None, None, ...]
# Assemble the tensor with signal derivatives with respect to the lm
# coefficients by contracting dsignal_lm_f3 with the response
dsignal_tensor_anisotropies = jnp.einsum(
"ijkl,aij->ajkl", response_IJ, dsignal_lm_f_anisotropies
)
return (
signal_tensor,
dsignal_tensor_frequency_shape,
dsignal_tensor_anisotropies,
)
@jax.jit
def get_signal_dsignal_tensors_lm_sqrt_basis(
signal_lm, signal, dsignal, response_IJ
):
"""
Compute the signal and its derivatives in the sqrt basis
Parameters:
-----------
signal_lm: numpy.ndarray or jax.numpy.ndarray
Array containing the lm coefficients.
signal : numpy.ndarray or jax.numpy.ndarray
Array containing the signal evaluated at all frequencies.
dsignal : numpy.ndarray or jax.numpy.ndarray
2D array containing derivative of the signal data with respect to the
signal parameters. This function assumes dsignal to have shape (P, F)
where P is the number of parameters, F is the number of frequency bins
response_IJ : numpy.ndarray or jax.numpy.ndarray
4D array containing response tensor for all the pulsars pairs.
It has shape (lm, N, pulsars, pulsars), where lm is the number of
coefficients for the anisotropy decomposition (spherical harmonics or
sqrt basis), N is the number of frequencies, and pulsars is the number
of pulsars.
Returns:
--------
Tuple containing:
- signal_tensor: numpy.ndarray or jax.numpy.ndarray
the signal tensor in the sqrt basis.
- dsignal_tensor_frequency_shape: numpy.ndarray or jax.numpy.ndarray
the derivative of the signal tensor with respect to the signal
parameters in frequency space.
- dsignal_tensor_anisotropies: numpy.ndarray or jax.numpy.ndarray
the derivative of the signal tensor with respect to the lm coefficients
"""
# Assemble the signal tensor, signal_lm are the lm coefficients of the
# signal and signal is the signal in frequency space (with len = F)
# response_IJ tensor has shape (lm, lm, F, N, N)
signal_tensor = signal[:, None, None] * jnp.einsum(
"abcde,a,b->cde", response_IJ, signal_lm, signal_lm
)
# Assemble the tensor with signal derivatives with respect to the signal
# parameters, we thus sum over the lm coefficients and multiply by dsignal
# with axes introduced for pulsars
dsignal_tensor_frequency_shape = (
dsignal[..., None, None]
* jnp.einsum("abcde,a,b->cde", response_IJ, signal_lm, signal_lm)[
None, ...
]
)
# Assemble the tensor with signal derivatives with respect to the lm
# coefficients. Note that signal_lm enters quadratically in signal_tensor
dsignal_tensor_anisotropies = (
2.0
* signal[None, :, None, None]
* jnp.einsum("abcde,b->acde", response_IJ, signal_lm)
)
return (
signal_tensor,
dsignal_tensor_frequency_shape,
dsignal_tensor_anisotropies,
)
lm_basis_list = [
get_signal_dsignal_tensors_lm_spherical_harmonics_basis,
# get_signal_dsignal_tensors_lm_sqrt_basis,
]
lm_basis_map = {
"spherical_harmonics_basis": 0,
# "sqrt_basis": 1,
}
@jax.jit
def get_signal_dsignal_tensors_lm(
lm_basis_idx, signal_lm, signal, dsignal, response_IJ
):
"""
Compute the signal and its derivatives in the spherical harmonics basis
Parameters:
-----------
lm_basis_idx: int
Index indicating the basis to use for the anisotropy decomposition
signal_lm: numpy.ndarray or jax.numpy.ndarray
Array containing the lm coefficients.
signal : numpy.ndarray or jax.numpy.ndarray
Array containing the signal evaluated at all frequencies.
dsignal : numpy.ndarray or jax.numpy.ndarray
2D array containing derivative of the signal data with respect to the
signal parameters. This function assumes dsignal to have shape (P, F)
where P is the number of parameters, F is the number of frequency bins
response_IJ : numpy.ndarray or jax.numpy.ndarray
4D array containing response tensor for all the pulsars pairs.
It has shape (lm, F, N, N), where lm is the number of coefficients for
the anisotropy decomposition (spherical harmonics or sqrt basis), F is
the number of frequencies, N is the number of pulsars.
Returns:
--------
Tuple containing:
- signal_tensor: numpy.ndarray or jax.numpy.ndarray
the signal tensor in the chosen basis.
- dsignal_tensor_frequency_shape: numpy.ndarray or jax.numpy.ndarray
the derivative of the signal tensor with respect to the signal
parameters in frequency space.
- dsignal_tensor_anisotropies: numpy.ndarray or jax.numpy.ndarray
the derivative of the signal tensor with respect to the lm coefficients
"""
return jax.lax.switch(
lm_basis_idx, lm_basis_list, signal_lm, signal, dsignal, response_IJ
)
@jax.jit
def get_integrands_lm(
signal_lm,
signal,
dsignal,
response_IJ,
noise_tensor,
HD_functions_IJ,
lm_basis_idx,
):
"""
Compute integrands for Signal-to-Noise Ratio (SNR) and Fisher Information
Matrix given some signal data, derivatives and other quantities that
characterize the pulsar configuration used for the analysis
Parameters:
-----------
signal_lm: numpy.ndarray or jax.numpy.ndarray
Array containing the lm coefficients.
signal : numpy.ndarray or jax.numpy.ndarray
Array containing signal data.
dsignal : numpy.ndarray or jax.numpy.ndarray
2D array containing derivative of the signal data with respect to the
signal parameters. This function assumes dsignal to have shape (P, N)
where P is the number of parameters, N is the number of frequency bins
response_IJ : numpy.ndarray or jax.numpy.ndarray
4D array containing the response for all frequencies and pulsar pairs.
Assumed to have shape (lm, N, M, M), where lm are the spherical
harmonics coefficients N is the number of frequency bins and M is the
number of pulsars.
noise_tensor : numpy.ndarray or jax.numpy.ndarray
3D array containing the noise for all frequencies and pulsar pairs.
Assumed to have shape (N, M, M), where N is the number of frequency
bins and M is the number of pulsars.
HD_functions_IJ : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre or binned projection of the Hellings and
Downs correlations. The shape is (HD_order + 1, F, N, N), where HD_order
is the maximum order of Legendre polynomials / bins, F is the number of
frequencies,and N is the number of pulsars.
lm_basis_idx: int
Index indicating the basis to use for the anisotropy decomposition
Returns:
--------
Tuple containing:
- SNR_integrand: numpy.ndarray or jax.numpy.ndarray
the integrand for the Signal-to-Noise Ratio (SNR) computation.
- effective_noise: numpy.ndarray or jax.numpy.ndarray
the effective noise as a function of frequency.
- fisher_integrand: numpy.ndarray or jax.numpy.ndarray
3D array wite the integrand for the Fisher Information Matrix
computation.
"""
(
signal_tensor,
dsignal_tensor_frequency_shape,
dsignal_tensor_anisotropies,
) = get_signal_dsignal_tensors_lm(
lm_basis_idx, signal_lm, signal, dsignal, response_IJ
)
# Build the covariance
covariance = signal_tensor + noise_tensor
# Invert the covariance
c_inverse = ut.compute_inverse(covariance)
# This is the SNR integrand
SNR_integrand = get_SNR_integrand(signal_tensor, c_inverse)
# Build the effective noise
effective_noise = jnp.sqrt(signal**2 / SNR_integrand)
# Assemble the HD coefficients part, we multiply the monopole, which is
# given by signal and the HD functions
dsignal_tensor_HD = signal[:, None, None] * HD_functions_IJ
# Concatenate the three tensors along the parameter axis
dsignal_tensor = jnp.concatenate(
(
dsignal_tensor_frequency_shape,
dsignal_tensor_HD,
dsignal_tensor_anisotropies,
),
axis=0,
)
# Get the fisher integrand
fisher_integrand = get_fisher_integrand(dsignal_tensor, c_inverse)
# Return all the relevant quantities
return SNR_integrand, effective_noise, fisher_integrand
def compute_fisher(
T_obs_yrs=10.33,
n_frequencies=30,
signal_label="power_law",
signal_parameters=SMBBH_parameters,
signal_lm=default_signal_lm,
get_tensors_kwargs={},
generate_catalog_kwargs={},
):
"""
Compute Fisher Information and related quantities. Keyword arguments for
get_tensors and generate_pulsars_catalog can be provided via
get_tensors_kwargs and generate_catalog_kwargs.
Parameters:
-----------
T_obs_yrs : float, optional
Observation time in years
default is 10.33
n_frequencies : int, optional
Number of frequency bins
default is 30
signal_label : str, optional
Label indicating the type of signal model to use
default is "power_law".
signal_parameters : dict, optional
Dictionary containing parameters for the signal model
default is SMBBH_parameters.
get_tensors_kwargs : dict
Additional keyword arguments for the get_tensors function.
generate_catalog_kwargs : dict
Additional keyword arguments for the generate_catalog function.
Returns:
--------
Tuple containing:
- frequency: numpy.ndarray or jax.numpy.ndarray
frequency bins.
- signal: numpy.ndarray or jax.numpy.ndarray
the computed signal.
- HD_functions_IJ : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre or binned projection of the Hellings and
Downs correlations. The shape is (HD_order + 1, F, N, N), where HD_order
is the maximum order of Legendre polynomials / bins, F is the number of
frequencies,and N is the number of pulsars.
- HD_coefficients : numpy.ndarray or jax.numpy.ndarray
Legendre coefficients for Hellings and Downs correlations values up to
the given HD_order.
- effective_noise: numpy.ndarray or jax.numpy.ndarray
effective noise.
- SNR: float
Signal-to-Noise Ratio (SNR) value.
- fisher: numpy.ndarray or jax.numpy.ndarray
2D array with the Fisher Information Matrix.
"""
if "anisotropies" in get_tensors_kwargs.keys():
anisotropies = get_tensors_kwargs["anisotropies"]
if "lm_basis" in get_tensors_kwargs.keys():
lm_basis = get_tensors_kwargs["lm_basis"]
else:
lm_basis = "spherical_harmonics_basis"
else:
anisotropies = False
lm_basis = "spherical_harmonics_basis"
lm_basis_idx = lm_basis_map[lm_basis]
# Setting the frequency vector from the observation time
frequency = (1.0 + jnp.arange(n_frequencies)) / (T_obs_yrs * ut.yr)
# Get the functions for the signal and its derivatives
model = get_model(signal_label)
signal_model = model["signal_model"]
dsignal_model = model["dsignal_model"]
# Computing the signal
signal = signal_model(frequency, signal_parameters)
# Building the signal derivatives
dsignal = jnp.array(
[
dsignal_model(i, frequency, signal_parameters)
for i in range(0, len(signal_parameters))
]
)
# Gets all the ingredients to compute the fisher
strain_omega, response_IJ, HD_functions_IJ, HD_coefficients = get_tensors(
frequency, **get_tensors_kwargs, **generate_catalog_kwargs
)
if anisotropies:
# Computes the fisher
SNR_integrand, effective_noise, fisher_integrand = get_integrands_lm(
signal_lm,
signal,
dsignal,
response_IJ,
strain_omega,
HD_functions_IJ,
lm_basis_idx,
)
else:
# Computes the fisher
SNR_integrand, effective_noise, fisher_integrand = get_integrands(
signal,
dsignal,
response_IJ,
strain_omega,
HD_functions_IJ,
)
# Compute SNR and Fisher integrals
SNR = jnp.sqrt(jnp.sum(SNR_integrand, axis=-1))
fisher = jnp.sum(fisher_integrand, axis=-1)
return (
frequency,
signal,
HD_functions_IJ,
HD_coefficients,
effective_noise,
SNR,
fisher,
)
| 21,058 | Python | .py | 490 | 36.381633 | 80 | 0.687814 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,338 | get_tensors.py | Mauropieroni_fastPTA/fastPTA/get_tensors.py | # Global
import numpy as np
import healpy as hp
import pandas as pd
import jax
import jax.numpy as jnp
from scipy.special import legendre
from scipy.integrate import simpson
# Local
import fastPTA.utils as ut
from fastPTA.generate_new_pulsar_configuration import generate_pulsars_catalog
jax.config.update("jax_enable_x64", True)
jax.config.update("jax_default_device", jax.devices(ut.which_device)[0])
# Just some constants
log_A_curn_default = -13.94
log_gamma_curn_default = 2.71
integration_points = 10000
@jax.jit
def unit_vector(theta, phi):
"""
Compute the unit vector in 3D Cartesian coordinates given spherical
coordinates. Theta is the polar angle (co-latitude) and phi is the
azimuthal angle (longitude). The input angles theta and phi should be given
in radians.
Parameters:
-----------
theta : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the polar angle (co-latitude).
phi : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the azimuthal angle (longitude).
Returns:
--------
unit_vec : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors in 3D Cartesian coordinates corresponding to
the given spherical coordinates. The shape will be (N, 3), where N is
the number of unit vectors.
"""
# Compute the x component of the unit vector
x_term = jnp.sin(theta) * jnp.cos(phi)
# Compute the y component of the unit vector
y_term = jnp.sin(theta) * jnp.sin(phi)
# Compute the z component of the unit vector
z_term = jnp.cos(theta)
# Assemble the unit vector and return it with the right shape
return jnp.array([x_term, y_term, z_term]).T
@jax.jit
def dot_product(theta_1, phi_1, theta_2, phi_2):
"""
Compute the dot product of two unit vectors given their spherical
coordinates. Theta is the polar angle (co-latitude) and phi is the
azimuthal angle (longitude). The input angles theta and phi should be given
in radians.
Parameters:
-----------
theta_1 : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the polar angle (co-latitude)
of the first vector.
phi_1 : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the azimuthal angle (longitude)
of the first vector.
theta_2 : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the polar angle (co-latitude)
of the second vector.
phi_2 : numpy.ndarray or jax.numpy.ndarray
Array of angles in radians representing the azimuthal angle (longitude)
of the second vector.
Returns:
--------
dot_product : numpy.ndarray or jax.numpy.ndarray
Array of dot products computed for the given unit vectors.
"""
# Sum of the product of x and y components
term1 = jnp.sin(theta_1) * jnp.sin(theta_2) * jnp.cos(phi_1 - phi_2)
# The product of the z components
term2 = jnp.cos(theta_1) * jnp.cos(theta_2)
return term1 + term2
@jax.jit
def HD_correlations(zeta_IJ):
"""
Compute the Hellings and Downs correlations for two line of sights with
angular separations zeta_IJ (in radiants).
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
Returns:
--------
HD correlations : numpy.ndarray or jax.numpy.ndarray
2D array of correlations computed for the angular separations zeta_IJ.
The array has shape (N, N), where N is the number of pulsars.
"""
# Compute the difference between 1 and p_I dot p_j and divide by 2
diff_IJ = 0.5 * (1.0 - zeta_IJ)
# This function does not include the Kronecker-Delta term for pulsar
# self-correlations
return jnp.where(
diff_IJ > 0.0, 0.5 + 1.5 * diff_IJ * (jnp.log(diff_IJ) - 1 / 6), 0.5
)
# Some default values to compute the HD curve
x = jnp.linspace(-1.0, 1.0, integration_points)
HD_value = HD_correlations(x)
@jax.jit
def get_WN(WN_parameters, dt):
"""
Compute the white noise amplitude for a catalog of pulsars given the
the white noise amplitudes and sampling rates (see Eq. 5 of 2404.02864).
The time step dt should be provided in seconds.
Parameters:
-----------
WN_parameters : numpy.ndarray or jax.numpy.ndarray
White noise parameters for the pulsar.
dt : numpy.ndarray or jax.numpy.ndarray
Time steps for the pulsar.
Returns:
--------
WN_amplitude : numpy.ndarray or jax.numpy.ndarray
Array of white noise amplitudes.
"""
return 1e-100 + jnp.array(1e-12 * 2 * WN_parameters**2 * dt)
@jax.jit
def get_pl_colored_noise(frequencies, log10_ampl, gamma):
"""
Compute power-law colored noise for given frequencies and parameters.
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies (in Hz) at which to compute the colored noise.
log10_ampl : numpy.ndarray or jax.numpy.ndarray
Array of base-10 logarithm of amplitudes.
gamma : numpy.ndarray or jax.numpy.ndarray
Array of power-law indices.
Returns:
--------
colored_noise : numpy.ndarray or jax.numpy.ndarray
Array of power-law colored noise computed for the given frequencies.
"""
amplitude_prefactor = (10**log10_ampl) ** 2 / 12.0 / jnp.pi**2 / ut.f_yr**3
frequency_dependent_term = (ut.f_yr / frequencies)[None, :] ** gamma[
:, None
]
return amplitude_prefactor[:, None] * frequency_dependent_term
@jax.jit
def get_noise_omega(frequencies, noise):
"""
Takes pulsar noises and convert them in Omega units.
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies.
noise : numpy.ndarray or jax.numpy.ndarray
Array representing the noise.
Returns:
--------
noise_omega : numpy.ndarray or jax.numpy.ndarray
Noise converted to Omega units. The shape will be (F, N, N), where F
is the number of frequencies and N is the number of pulsars.
"""
# Factor to convert to strain
convert_to_strain_factor = 12 * jnp.pi**2 * frequencies**2
# Factor to convert to omega units
convert_to_omega_factor = (
ut.strain_to_Omega(frequencies) * convert_to_strain_factor
)
# Convert the noise to omega units
return convert_to_omega_factor[:, None, None] * (
(noise.T)[..., None] * jnp.eye(len(noise))[None, ...]
)
@jax.jit
def get_pulsar_noises(
frequencies,
WN_par,
log10_A_red,
gamma_red,
log10_A_dm,
gamma_dm,
log10_A_sv,
gamma_sv,
dt,
):
"""
Compute noise components for given parameters and frequencies.
The components included are (see Eq. 4 of 2404.02864):
- White noise (WN) with parameter WN_par.
- Red noise (RN) with (log10)amplitudes log10_A_red and power-law indices
gamma_red.
- Dispersion measure noise (DM) with amplitudes log10_A_dm and power-law
indices gamma_dm.
- Scattering variation noise (SV) with amplitudes log10_A_sv and power-law
indices gamma_sv.
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies (in Hz).
WN_par : float
Array of white noise parameter.
log10_A_red : numpy.ndarray or jax.numpy.ndarray
Array of base-10 logarithm of amplitudes for red noise.
gamma_red : numpy.ndarray or jax.numpy.ndarray
Array of power-law indices for red noise.
log10_A_dm : numpy.ndarray or jax.numpy.ndarray
Array of base-10 logarithm of amplitudes for dispersion measure noise.
gamma_dm : numpy.ndarray or jax.numpy.ndarray
Array of power-law indices for dispersion measure noise.
log10_A_sv : numpy.ndarray or jax.numpy.ndarray
Array of base-10 logarithm of amplitudes for scattering variation noise.
gamma_sv : numpy.ndarray or jax.numpy.ndarray
Array of power-law indices for scattering variation noise.
dt : float
Array of time steps in seconds.
Returns:
--------
noise : numpy.ndarray or jax.numpy.ndarray
3D array of noise components computed for the given parameters and
frequencies. The shape will be (F, N, N), where F is the number of
frequencies and N is the number of pulsars.
"""
# white noise coverted from microsecond to second
WN = get_WN(WN_par, dt)
# red noise powerlaw for A expressed in strain amplitude
RN = get_pl_colored_noise(frequencies, log10_A_red, gamma_red)
# dispersion measure noise powerlaw for A
# expressed in strain amplitude (evaluated at 1.4 GHz channel)
DM = get_pl_colored_noise(frequencies, log10_A_dm, gamma_dm)
# scattering variation noise powerlaw for A
# expressed in strain amplitude (evaluated at 1.4 GHz channel)
SV = get_pl_colored_noise(frequencies, log10_A_sv, gamma_sv)
return WN[:, None] + RN + DM + SV
@jax.jit
def transmission_function(frequencies, T_obs):
"""
Compute the transmission function (see Eq. 3 of 2404.02864), which
represents the attenuation of signals, for some frequencies given the
observation time.
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies (in Hz).
T_obs : float
Observation time (in seconds).
Returns:
--------
transmission : numpy.ndarray or jax.numpy.ndarray
Array of transmission values computed for the given frequencies and
observation time.
"""
return 1 / (1 + 1 / (frequencies * T_obs) ** 6)
@jax.jit
def get_time_tensor(frequencies, pta_span_yrs, Tspan_yr):
"""
Computes the time tensor (i.e., the part of the response depending on the
observation times) for given frequencies and observation times.
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies.
pta_span_yrs : float
Average span of the PTA data in years.
Tspan_yr : float
Time span for individual pulsars in years.
Returns:
--------
time_tensor : numpy.ndarray or jax.numpy.ndarray
3D array representing the time tensor computed for the given
frequencies, PTA span, and individual pulsar spans. It has shape
(F, N, N), where F is the number of frequencies and N is the number
of pulsars.
"""
# Do a mesh of the observation times
time_1, time_2 = jnp.meshgrid(Tspan_yr, Tspan_yr)
# pick the minimium time in each pair
time_IJ = jnp.min(jnp.array([time_1, time_2]), axis=0)
# Compute the transmission function for all times and frequencies
transmission = transmission_function(
frequencies[:, None], (Tspan_yr * ut.yr)[None, :]
)
# Build the tensor product of the two transimission function
transmission_tensor = transmission[:, :, None] * transmission[:, None, :]
# Return the tensor product weighted by the total observation time
return jnp.sqrt((time_IJ / pta_span_yrs)[None, ...] * transmission_tensor)
@jax.jit
def gamma_pulsar_pair_analytical(
theta_1, phi_1, theta_2, phi_2, theta_k, phi_k
):
"""
Compute the analytical expression for the gamma function (see Eq. 13 of
2407.14460).
Parameters:
-----------
theta_1 : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the first pulsar.
phi_1 : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the first pulsar.
theta_2 : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the second pulsar.
phi_2 : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the second pulsar.
theta_k : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the pixel vectors.
phi_k : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the pixel vectors.
Returns:
--------
gamma : numpy.ndarray or jax.numpy.ndarray
Array of gamma values computed for the given pulsar pairs and pixel
vectors.
"""
# Compute p_dot_k
p_dot_k = dot_product(theta_k, phi_k, theta_1, phi_1)
# Compute q_dot_k
q_dot_k = dot_product(theta_k, phi_k, theta_2, phi_2)
# Compute p_dot_q
p_dot_q = dot_product(theta_1, phi_1, theta_2, phi_2)
# This is the second term
second_term = -(1.0 - p_dot_k) * (1.0 - q_dot_k)
# This is the numerator of the first term
numerator = 2.0 * (p_dot_q - p_dot_k * q_dot_k) ** 2.0
# This is the denominator of the first term
denominator = (1.0 + p_dot_k) * (1.0 + q_dot_k)
# Where the denominator is non zero, just numerator / denominator,
# where it's zero a bit more care is needed, if pI != pJ is zero
first_term = jnp.where(denominator != 0.0, numerator / denominator, 0.0)
conditions = (
(denominator == 0.0)
& (phi_1 - phi_2 == 0.0)
& (theta_1 - theta_2 == 0.0)
)
# Correct first term where the denominator is zero and pI = pJ
first_term = jnp.where(
conditions, -2.0 * second_term, first_term # type: ignore
)
# Sum all the terms up
return first_term + second_term
@jax.jit
def gamma_analytical(theta, phi, theta_k, phi_k):
"""
Compute the analytical expression for the gamma function (see Eq. 13 of
2407.14460).
Parameters:
-----------
theta : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the pulsars.
phi : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the pulsars.
theta_k : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the pixel vectors.
phi_k : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the pixel vectors.
Returns:
--------
gamma : numpy.ndarray or jax.numpy.ndarray
3D array of gamma values computed for all pulsar pairs and for all
pixels. The shape will be (N, N, pp), where N is the number of pulsars
and pp is the number of pixels.
"""
return gamma_pulsar_pair_analytical(
theta[:, None, None],
phi[:, None, None],
theta[None, :, None],
phi[None, :, None],
theta_k[None, None, :],
phi_k[None, None, :],
)
@jax.jit
def gamma(p_I, hat_k):
"""
Compute the gamma function (see Eq. 13 of 2407.14460).
Parameters:
-----------
p_I : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors representing the pulsar directions.
Assumed to have shape (N, 3), N is the number of pulsars.
hat_k : numpy.ndarray or jax.numpy.ndarray
Array of unit vectors representing the pixel directions.
Assumed to have shape (pp, 3), pp is the number of pixels.
Returns:
--------
gamma : numpy.ndarray or jax.numpy.ndarray
3D array of gamma values computed for all pulsar pairs and pixels
The shape will be (N, N, pp) where N is the number of pulsars and pp is
the number of pixels.
"""
# This is the dot product of the unit vectors pointing towards the pulsars
pIpJ = jnp.einsum("iv,jv->ij", p_I, p_I)
# This is the dot product of p_I and hat_k
pIdotk = jnp.einsum("iv,jv->ij", p_I, hat_k)
# Create the sum and difference vectors to use later
sum = 1 + pIdotk
diff = 1 - pIdotk
# Compute the second term
second_term = -diff[:, None, :] * diff[None, ...]
# This is the pIdotk * pJdotk term in the numerator of the first term
pk_qk = pIdotk[:, None, :] * pIdotk[None, ...]
# Get the numerator of the first term
numerator = 2 * (pIpJ[..., None] - pk_qk) ** 2
# Get the denominator of the first term
denominator = sum[:, None, :] * sum[None, ...]
# Where the denominator is non zero, just numerator / denominator
first_term = jnp.where(denominator != 0.0, numerator / denominator, 0.0)
# Correct first term where the denominator is zero and pI = pJ
first_term = jnp.where(
((denominator == 0.0) & (jnp.bool_(jnp.floor(pk_qk)))),
-2.0 * second_term,
first_term,
)
# Sum the two terms and return
return first_term + second_term
def projection_spherial_harmonics_basis(quantity, l_max):
"""
Compute the spherical harmonics projection of a the correlation matrix
given quantity.
TBD: This function should use the fact that quantity is symmetric in the
pulsar pulsar indexes to reduce computations and increase efficiency.
Parameters:
-----------
quantity : numpy.ndarray
3D array to be projected on spherical harmonics. Should have shape
(N, N, P), where N is the number of pulsars and P is the number of
pixels, which should be compatible with healpy
(see https://healpy.readthedocs.io/en/latest/).
l_max : int
Maximum ell value.
Returns:
--------
real_alm : numpy.ndarray
3D array of real spherical harmonics coefficients.
It has shape (lm, N, N), where N is the number of pulsars and
lm = (l_max + 1)**2 is the number of spherical harmonics coefficients.
"""
# Get the shape of the quantity to project on spherical harmonics
shape = list(quantity.shape)
# Reshape quantity so that it can be passed to hp.map2alm
qquantity = np.reshape(quantity, (int(shape[0] ** 2), shape[-1]))
# Get all the alms
real_alm = np.apply_along_axis(
ut.spherical_harmonics_projection, 1, qquantity, l_max
)
# Reshape to get the same shape as before
return np.reshape(real_alm, (shape[0], shape[0], real_alm.shape[-1])).T
def get_correlations_lm_IJ_spherical_harmonics_basis(p_I, l_max, gamma_pq):
"""
Compute the correlations in spherical harmonics basis for a given pulsar
catalog. The correlations are computed up to a maximum ell value l_max and
for a given nside.
Parameters:
-----------
p_I : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors representing the pulsar directions.
Assumed to have shape (N, 3), N is the number of pulsars.
l_max : int
Maximum ell value.
gamma_pq : numpy.ndarray or jax.numpy.ndarray
3D array of gamma values computed for all pulsar pairs and pixels.
The shape should be (N, N, pp), where N is the number of pulsars and
pp is the number of pixels
Returns:
--------
correlations_lm : numpy.ndarray
3D array of correlations computed in spherical harmonics basis.
It has shape (lm, N, N), where N is the number of pulsars and
lm = (l_max + 1)**2 is the number of spherical harmonics coefficients.
"""
# Project gamma onto spherical harmonics
correlations_lm = projection_spherial_harmonics_basis(gamma_pq, l_max)
# Multiply by 1 + delta_{IJ} and return
return correlations_lm * (1 + np.eye(len(p_I)))[None, ...]
def get_correlations_lm_IJ_sqrt_basis(p_I, l_max, theta_k, phi_k, gamma_pq):
"""
Compute the correlations in sqrt basis for a given pulsar catalog. The
correlations are computed up to a maximum ell value l_max and for a given
nside.
Parameters:
-----------
p_I : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors representing the pulsar directions.
Assumed to have shape (N, 3), N is the number of pulsars.
l_max : int
Maximum ell value.
theta_k : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes) for the pixel vectors.
phi_k : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes) for the pixel vectors.
gamma_pq : numpy.ndarray or jax.numpy.ndarray
3D array of gamma values computed for all pulsar pairs and pixels.
The shape should be (N, N, pp), where N is the number of pulsars and
pp is the number of pixels
Returns:
--------
correlations_lm : numpy.ndarray
4D array of correlations computed in sqrt basis.
It has shape (lm, lm, N, N), where N is the number of pulsars and
lm = (l_max + 1)**2 is the number of spherical harmonics coefficients.
"""
# Get the number of pixels
npix = hp.nside2npix(theta_k)
# spherical harmonis with shape (lm, pp)
spherical_harmonics = ut.get_spherical_harmonics(l_max, theta_k, phi_k)
# Quadratic spherical_harmonics basis with shape (lm, lm, pp)
quadratic = spherical_harmonics[:, None] * spherical_harmonics[None, :]
# Project gamma onto the sqrt basis
correlations_lm = (
np.einsum("ijp,nmp->ijnm", quadratic, gamma_pq) * (4 * jnp.pi) / npix
)
return correlations_lm * (1 + np.eye(len(p_I)))[None, ...]
def get_correlations_lm_IJ(
p_I, l_max, nside, lm_basis="spherical_harmonics_basis"
):
"""
Compute the response tensor for given angular separations and time tensors.
Parameters:
-----------
p_I : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors representing the pulsar directions.
Assumed to have shape (N, 3), N is the number of pulsars.
l_max : int
Maximum ell value.
nside : int
Resolution parameter for the HEALPix grid.
lm_basis : str
Basis to compute the correlations.
Can be either "spherical_harmonics_basis" or "sqrt_basis".
Returns:
--------
correlations_lm_IJ : numpy.ndarray
3D array of correlations computed in the given basis.
It has shape (lm, N, N), where lm is the number of coefficients for
the anisotropy decomposition (spherical harmonics or sqrt basis) and N
is the number of pulsars.
"""
# Given nside get a pixelization of the sky
npix = hp.nside2npix(nside)
theta_k, phi_k = hp.pix2ang(nside, jnp.arange(npix))
theta_k = jnp.array(theta_k)
phi_k = jnp.array(phi_k)
# Get the k vector (i.e., the sky direction) for all the pixels
hat_k = unit_vector(theta_k, phi_k)
# Compute gamma in all the pixels, the shape is (N, N, pp)
gamma_pq = 3.0 / 8.0 * gamma(p_I, hat_k)
# Compute the correlations on lm basis
if lm_basis.lower() == "spherical_harmonics_basis":
correlations_lm_IJ = get_correlations_lm_IJ_spherical_harmonics_basis(
p_I, l_max, gamma_pq
)
elif lm_basis.lower() == "sqrt_basis":
correlations_lm_IJ = get_correlations_lm_IJ_sqrt_basis(
p_I, l_max, theta_k, phi_k, gamma_pq
)
# return the correlations
return correlations_lm_IJ
def get_response_IJ_lm(
p_I, time_tensor_IJ, l_max, nside, lm_basis="spherical_harmonics_basis"
):
"""
Compute the response tensor for given angular separations and time tensors.
Parameters:
-----------
p_I : numpy.ndarray or jax.numpy.ndarray
2D array of unit vectors representing the pulsar directions.
Assumed to have shape (N, 3), N is the number of pulsars.
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
3D array containing the attenuations due to the observation time for
all the pulsars and for all frequencies. Should have shape (F, N, N),
where F is the number of frequencies and N is the number of pulsars.
l_max : int
Maximum ell value.
nside : int
Resolution parameter for the HEALPix grid.
lm_basis : str
Basis to compute the correlations.
Can be either "spherical_harmonics_basis" or "sqrt_basis".
Returns:
--------
response_IJ : numpy.ndarray or jax.numpy.ndarray
4D array containing response tensor for all the pulsars pairs.
It has shape (lm, F, N, N), where lm is the number of coefficients for
the anisotropy decomposition (spherical harmonics or sqrt basis), F is
the number of frequencies, N is the number of pulsars.
"""
# Compute the correlations on lm basis
correlations_lm_IJ = get_correlations_lm_IJ(
p_I, l_max, nside, lm_basis=lm_basis
)
# combine the Hellings and Downs part and the time part
return time_tensor_IJ[None, ...] * correlations_lm_IJ[:, None, ...]
@jax.jit
def get_chi_tensor_IJ(zeta_IJ):
"""
Computes the chi_IJ tensor as expressed in eq. 15 of
https://arxiv.org/pdf/2404.02864.pdf for given angular separations.
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
Returns:
--------
chi_IJ : numpy.ndarray or jax.numpy.ndarray
2D array representing the chi_IJ tensor for all the pulsars pairs.
It has shape (N, N), where N is the number of pulsars.
"""
# Compute HD and add the self correlation term
return HD_correlations(zeta_IJ) + 0.5 * jnp.eye(len(zeta_IJ))
@jax.jit
def get_response_IJ(zeta_IJ, time_tensor_IJ):
"""
Compute the response tensor for given angular separations and time tensors.
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
3D array representing the time tensor containing the attenuations due
to the observation time for all the pulsars and for all frequencies.
It should have shape (F, N, N), where F is the number of frequencies
and N is the number of pulsars.
Returns:
--------
response_IJ : numpy.ndarray or jax.numpy.ndarray
3D array containing the response tensor for all the pulsars pairs. It
has shape (F, N, N), where F is the number of frequencies and N is the
number of pulsars.
"""
# Compute the chi_IJ tensor for the angular separations
chi_tensor_IJ = get_chi_tensor_IJ(zeta_IJ)
# combine the Hellings and Downs part and the time part
return time_tensor_IJ * chi_tensor_IJ[None, ...]
def get_HD_Legendre_coefficients(HD_order):
"""
Compute Legendre coefficients for Hellings and Downs correlations for
polynomials up to some HD_order
Parameters:
-----------
HD_order : int
Maximum order of Legendre coefficients to compute.
Returns:
--------
coefficients : numpy.ndarray or jax.numpy.ndarray
Array of Legendre coefficients computed up to the given HD_order.
"""
# Some l dependent normalization factor
l_coeffs = (2 * jnp.arange(HD_order + 1) + 1) / 2
return jnp.array(
[
# Project onto Legendre polynomials
simpson(legendre(i)(x) * HD_value, x=x) * l_coeffs[i]
for i in range(HD_order + 1)
]
)
def get_polynomials_IJ(zeta_IJ, HD_order):
"""
Compute Legendre polynomials for given angular separations and HD_order.
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
HD_order : int
Maximum HD_order of Legendre polynomials.
Returns:
--------
polynomials_IJ : numpy.ndarray or jax.numpy.ndarray
Array of Legendre polynomials computed for the given angular separations
and HD_order.
"""
# Create an array to store the Legendre polynomials
polynomials_IJ = []
# Compute the Legendre polynomials for all the angular separations
for i in range(HD_order + 1):
polynomials_IJ.append(legendre(i)(zeta_IJ))
return jnp.array(polynomials_IJ)
@jax.jit
def Legendre_projection(time_tensor_IJ, polynomials_IJ):
"""
Projects the pulsar angular information onto Legendre polynomials
Parameters:
-----------
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
Time tensor containing the attenuations due to the observation time for
all the pulsars and for all frequencies. The shape is (F, N, N), where
F is the number of frequencies and N is the number of pulsars.
polynomials_IJ : numpy.ndarray or jax.numpy.ndarray
Array of Legendre polynomials. The shape is (HD_order + 1, N, N), where
HD_order is the maximum order of Legendre polynomials and N is the
number of pulsars.
Returns:
--------
projection : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre projection of the Hellings and Downs
correlations. The shape is (HD_order + 1, F, N, N), where HD_order is
the maximum order of Legendre polynomials, F is the number of
frequencies, and N is the number of pulsars.
"""
return time_tensor_IJ[None, ...] * polynomials_IJ[:, None, ...]
def HD_projection_Legendre(zeta_IJ, time_tensor_IJ, HD_order):
"""
Projects Hellings and Downs correlations onto Legendre polynomials.
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
Time tensor containing the attenuations due to the observation time for
all the pulsars and for all frequencies. The shape is (F, N, N), where
F is the number of frequencies and N is the number of pulsars.
HD_order : int
Maximum HD_order of Legendre polynomials.
Returns:
--------
Tuple containing:
- HD_functions : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre projection of the Hellings and Downs
correlations. The shape is (HD_order + 1, F, N, N), where HD_order is
the maximum order of Legendre polynomials, F is the number of
frequencies, and N is the number of pulsars.
- HD_coefficients : numpy.ndarray or jax.numpy.ndarray
Legendre coefficients for Hellings and Downs correlations up to the
given HD_order.
"""
# Gets the Legendre coefficients for HD
HD_coefficients = get_HD_Legendre_coefficients(HD_order)
# Gets the values of the HD polynomials for all angular separations
polynomials_IJ = get_polynomials_IJ(zeta_IJ, HD_order)
# Projects the pulsar catalog onto Legendre polynomials
HD_functions = Legendre_projection(time_tensor_IJ, polynomials_IJ)
return HD_functions, HD_coefficients
@jax.jit
def binned_projection(zeta_IJ, time_tensor_IJ, masks):
"""
Compute binned projection of the Hellings and Downs correlations.
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
Time tensor containing the attenuations due to the observation time for
all the pulsars and for all frequencies. The shape is (F, N, N), where
F is the number of frequencies and N is the number of pulsars.
masks : numpy.ndarray or jax.numpy.ndarray
Array of masks representing binned intervals for Hellings and Downs
correlations.
Returns:
--------
binned_projection : numpy.ndarray or jax.numpy.ndarray
4D array with the binned projection of the Hellings and Downs
correlations. The shape is (HD_order + 1, F, N, N), where HD_order is number of bins, F is the number of frequencies, and N is the number of
pulsars.
"""
return (
time_tensor_IJ - jnp.eye(len(zeta_IJ))[None, ...] * time_tensor_IJ
) * masks[:, None, ...]
def HD_projection_binned(zeta_IJ, time_tensor_IJ, HD_order):
"""
Projects Hellings and Downs correlations onto binned intervals.
NB!! For consistency with the Legendre version it uses HD_order +1 bins!
Parameters:
-----------
zeta_IJ : numpy.ndarray or jax.numpy.ndarray
2D array of angular separations zeta_IJ. The angular separations should
be given in radians, and the array should have shape (N, N), where N
is the number of pulsars.
time_tensor_IJ : numpy.ndarray or jax.numpy.ndarray
Time tensor containing the attenuations due to the observation time for
all the pulsars and for all frequencies. The shape is (F, N, N), where
F is the number of frequencies and N is the number of pulsars.
HD_order : int
Number of bins used in the analysis.
Returns:
--------
Tuple containing:
- HD_functions : numpy.ndarray or jax.numpy.ndarray
4D array with the binned projection of the Hellings and Downs
correlations. The shape is (HD_order + 1, F, N, N), where HD_order is
number of bins, F is the number of frequencies, and N is the number of
pulsars.
- HD_coefficients : numpy.ndarray or jax.numpy.ndarray
Coefficients of the binned Hellings and Downs correlations
"""
# Ensure Hellings and Downs correlations values are within bounds
xi_vals = jnp.arccos(jnp.clip(zeta_IJ, -1.0, 1.0))
# Compute bin edges for binning the Hellings and Downs correlations values
bin_edges = jnp.linspace(0.0, jnp.pi, HD_order + 2)
# Compute mean values for each bin
mean_x = 0.5 * (jnp.cos(bin_edges[:-1]) + np.cos(bin_edges[1:]))
# Compute Hellings and Downs correlations coefficients for binned intervals
HD_coefficients = HD_correlations(mean_x)
# Create masks for each bin
masks = np.zeros(shape=(HD_order + 1, len(zeta_IJ), len(zeta_IJ)))
for i in range(len(masks)):
masks[i][(xi_vals > bin_edges[i]) & (xi_vals < bin_edges[i + 1])] = 1.0
# And converts to jax array
masks = jnp.array(masks)
# Project Hellings and Downs correlations functions onto binned intervals
HD_functions_IJ = binned_projection(zeta_IJ, time_tensor_IJ, masks)
return HD_functions_IJ, HD_coefficients
def get_tensors(
frequencies,
path_to_pulsar_catalog=ut.path_to_default_pulsar_catalog,
pta_span_yrs=10.33,
add_curn=False,
HD_order=0,
HD_basis="legendre",
anisotropies=False,
lm_basis="spherical_harmonics_basis",
l_max=0,
nside=16,
regenerate_catalog=False,
**generate_catalog_kwargs
):
"""
Generate all tensors (noise, response and Hellings and Downs) needed for
gravitational wave data analysis. The noise tensors is returned in omega
units. The Hellings and Downs correlations are projected onto Legendre
polynomials or binned intervals based on the chosen HD_basis
Parameters:
-----------
frequencies : numpy.ndarray or jax.numpy.ndarray
Array of frequencies.
path_to_pulsar_catalog : str, optional
Path to the pulsars data file.
Default is path_to_default_pulsar_catalog.
pta_span_yrs : float, optional
Average span of the PTA data in years.
Default is 10.33 years.
add_curn : bool, optional
Whether to add common (spatially) uncorrelated red noise (CURN).
Default is False.
HD_order : int, optional
Maximum order of Legendre polynomials/ number of bins for the Hellings
and Downs correlations projection.
Default is 0.
HD_basis : str, optional
Basis for Hellings and Downs correlations projection.
Options are "legendre" or "binned".
Default is "legendre".
anisotropies : bool, optional
Whether to include anisotropies in the response tensor.
Default is False.
lm_basis : str, optional
Basis for the anisotropy decomposition.
Options are "spherical_harmonics_basis" or "sqrt_basis".
Default is "spherical_harmonics_basis".
l_max : int, optional
Maximum ell value for the anisotropy decomposition.
Default is 0.
nside : int, optional
Resolution parameter for the HEALPix grid.
Default is 16.
regenerate_catalog : bool, optional
Whether to regenerate the pulsars data file.
Default is False.
**generate_catalog_kwargs : dict, optional
Additional keyword arguments for the generate_pulsars_catalog function.
Returns:
--------
Tuple containing:
- strain_omega : numpy.ndarray or jax.numpy.ndarray
Noise converted to Omega units. The shape will be (F, N, N), where F
is the number of frequencies and N is the number of pulsars.
- response_IJ : numpy.ndarray or jax.numpy.ndarray
Array containing the response tensor for all the pulsars pairs.
If anisotropies is False, it has shape (F, N, N), where F is the number
of frequencies and N is the number of pulsars. If anisotropies is True,
it has shape (lm, F, N, N), where lm is the number of coefficients for
the anisotropy decomposition (spherical harmonics or sqrt basis).
- HD_functions_IJ : numpy.ndarray or jax.numpy.ndarray
4D array with the Legendre or binned projection of the Hellings and
Downs correlations. The shape is (HD_order + 1, F, N, N), where HD_order is the maximum order of Legendre polynomials / bins, F is the
number of frequencies,and N is the number of pulsars.
- HD_coefficients : numpy.ndarray or jax.numpy.ndarray
Legendre coefficients for Hellings and Downs correlations values up to
the given HD_order.
"""
# Load or regenerate pulsar catalog
try:
if regenerate_catalog:
raise FileNotFoundError
pulsars_DF = pd.read_csv(path_to_pulsar_catalog, sep=" ")
except FileNotFoundError:
generate_catalog_kwargs["outname"] = path_to_pulsar_catalog
pulsars_DF = generate_pulsars_catalog(**generate_catalog_kwargs)
# unpack all parameters
WN_par = jnp.array(pulsars_DF["wn"].values)
log10_A_red = jnp.array(pulsars_DF["log10_A_red"].values)
gamma_red = jnp.array(pulsars_DF["g_red"].values)
log10_A_dm = jnp.array(pulsars_DF["log10_A_dm"].values)
gamma_dm = jnp.array(pulsars_DF["g_dm"].values)
log10_A_sv = jnp.array(pulsars_DF["log10_A_sv"].values)
gamma_sv = jnp.array(pulsars_DF["g_sv"].values)
Tspan_yr = jnp.array(pulsars_DF["Tspan"].values)
dt = jnp.array(pulsars_DF["dt"].values)
theta = jnp.array(pulsars_DF["theta"].values)
phi = jnp.array(pulsars_DF["phi"].values)
pi_vec = unit_vector(theta, phi)
# get noises for all pulsars
noise = get_pulsar_noises(
frequencies,
WN_par,
log10_A_red,
gamma_red,
log10_A_dm,
gamma_dm,
log10_A_sv,
gamma_sv,
dt,
)
# add curn if present
if add_curn:
log10_A_curn = jnp.repeat(log_A_curn_default, len(WN_par))
gamma_curn = jnp.repeat(log_gamma_curn_default, len(WN_par))
curn = get_pl_colored_noise(frequencies, log10_A_curn, gamma_curn)
noise += curn
# convert the noise in strain and then omega units
strain_omega = get_noise_omega(frequencies, noise)
# get the time tensor
time_tensor_IJ = get_time_tensor(frequencies, pta_span_yrs, Tspan_yr)
# compute angular separations
zeta_IJ = jnp.einsum("ik, jk->ij", pi_vec, pi_vec)
if not anisotropies:
# compute the response
response_IJ = get_response_IJ(zeta_IJ, time_tensor_IJ)
else:
response_IJ = get_response_IJ_lm(
pi_vec, time_tensor_IJ, l_max, nside, lm_basis=lm_basis
)
# and if needed the HD part
if HD_order > 0 and HD_basis.lower() == "legendre":
HD_functions_IJ, HD_coefficients = HD_projection_Legendre(
zeta_IJ, time_tensor_IJ, HD_order
)
elif HD_order > 0 and HD_basis.lower() == "binned":
HD_functions_IJ, HD_coefficients = HD_projection_binned(
zeta_IJ, time_tensor_IJ, HD_order
)
else:
HD_functions_IJ = jnp.zeros(
shape=(0, len(frequencies), len(WN_par), len(WN_par))
)
HD_coefficients = jnp.zeros(shape=(0,))
return strain_omega, response_IJ, HD_functions_IJ, HD_coefficients
| 40,944 | Python | .py | 948 | 36.905063 | 148 | 0.672704 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,339 | utils.py | Mauropieroni_fastPTA/fastPTA/utils.py | # Global
import os
import yaml
import numpy as np
import healpy as hp
from wigners import clebsch_gordan
import jax
import jax.numpy as jnp
from scipy.special import sph_harm
jax.config.update("jax_enable_x64", True)
# If you want to use your GPU change here
which_device = "cpu"
jax.config.update("jax_default_device", jax.devices(which_device)[0])
# H0/h = 100 km/s/Mpc expressed in meters
Hubble_over_h = 3.24e-18
# Hour
hour = 3600
# Day
day = 24 * hour
# Year in seconds
yr = 365.25 * day
# Frequency associated with 1yr
f_yr = 1 / yr
# Set the path to the default pulsar parameters
path_to_defaults = os.path.join(os.path.dirname(__file__), "defaults/")
# Set the path to the default pulsar parameters
path_to_default_pulsar_parameters = os.path.join(
path_to_defaults, "default_pulsar_parameters.yaml"
)
# Set the path to the default pulsar catalog
path_to_default_pulsar_catalog = os.path.join(
path_to_defaults, "default_catalog.txt"
)
# Set the path to the default pulsar catalog
path_to_default_NANOGrav_positions = os.path.join(
os.path.dirname(__file__), "defaults/NANOGrav_positions.txt"
)
# Set the path to the default pulsar catalog
path_to_default_NANOGrav_positions = os.path.join(
os.path.dirname(__file__), "defaults/NANOGrav_positions.txt"
)
def characteristic_strain_to_Omega(frequency):
"""
Computes the dimensionless gravitational wave energy density parameter
Omega_gw given the characteristic strain at certain frequencies.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Frequency (in Hz) at which the characteristic strain is measured.
Returns:
--------
Omega_gw : numpy.ndarray or jax.numpy.ndarray
Dimensionless gravitational wave energy density parameter
at the given frequencies.
Notes:
------
Hubble_over_h : float
Constant representing the Hubble parameter divided by the Hubble
constant.
"""
return 2 * jnp.pi**2 * frequency**2 / 3 / Hubble_over_h**2
def strain_to_Omega(frequency):
"""
Computes the dimensionless gravitational wave energy density parameter
Omega_gw given the strain at certain frequencies.
Parameters:
-----------
frequency : numpy.ndarray or jax.numpy.ndarray
Frequency (in Hz) at which the characteristic strain is measured.
Returns:
--------
Omega_gw : numpy.ndarray or jax.numpy.ndarray
Dimensionless gravitational wave energy density parameter
at the given frequencies.
Notes:
------
Hubble_over_h : float
Constant representing the Hubble parameter divided by the Hubble
constant.
"""
return 2 * jnp.pi**2 * frequency**3 / 3 / Hubble_over_h**2
# To go from CP delays to characteristic Strain
def hc_from_CP(CP, frequency, T_obs_s):
"""
Computes the characteristic strain given the Common Process (CP) delays.
Parameters:
-----------
CP : numpy.ndarray or jax.numpy.ndarray
Amplitude of the Common Process (in seconds^3)
frequency : numpy.ndarray or jax.numpy.ndarray
Frequency (in Hz) at which the characteristic strain is measured.
T_obs_s : float
Observation time (in seconds)
Returns:
--------
hc : numpy.ndarray or jax.numpy.ndarray
Characteristic strain at the given frequencies.
Notes:
------
Hubble_over_h : float
Constant representing the Hubble parameter divided by the Hubble
constant.
"""
return 2 * jnp.sqrt(3) * CP * frequency**1.5 * jnp.pi * jnp.sqrt(T_obs_s)
def load_yaml(path_to_file):
"""
Loads some inputs from a YAML file (located at the specified path) and
returns the parsed data as a dictionary.
Parameters:
-----------
path_to_file : str
Path to the YAML file to load.
Returns:
--------
data : dict
YAML data loaded from the file.
Notes:
------
"""
with open(path_to_file, "r+") as stream:
raw = "".join(stream.readlines())
return yaml.load(raw, Loader=yaml.SafeLoader)
default_pulsar_parameters = load_yaml(path_to_default_pulsar_parameters)
def get_l_max_real(real_spherical_harmonics):
"""
Given the real spherical harmonics coefficients, this function returns the
maximum ell value.
Parameters:
-----------
real_spherical_harmonics : numpy.ndarray
Array of real spherical harmonics coefficients. If dimension is > 1, lm
must be the first index
Returns:
--------
l_max : int
Maximum ell value.
"""
return int(np.sqrt(len(real_spherical_harmonics)) - 1)
def get_l_max_complex(complex_spherical_harmonics):
"""
Given the complex spherical harmonics coefficients, this function returns
the maximum ell value.
Parameters:
-----------
complex_spherical_harmonics : numpy.ndarray
Array of complex spherical harmonics coefficients. If dimension is > 1,
lm must be the first index
Returns:
--------
l_max : int
Maximum ell value.
"""
return int(np.sqrt(1.0 + 8.0 * len(complex_spherical_harmonics)) / 2 - 1.5)
def get_n_coefficients_complex(l_max):
"""
Given the maximum ell value, this function returns the number of spherical
harmonics coefficients for the complex representation.
Parameters:
-----------
l_max : int
Maximum ell value.
Returns:
--------
n_coefficients : int
Number of spherical harmonics coefficients.
"""
return int((l_max + 1) * (l_max + 2) / 2)
def get_n_coefficients_real(l_max):
"""
Given the maximum ell value, this function returns the number of spherical
harmonics coefficients for the real representation.
Parameters:
-----------
l_max : int
Maximum ell value.
Returns:
--------
n_coefficients : int
Number of spherical harmonics coefficients.
"""
return int((l_max + 1) ** 2)
def get_sort_indexes(l_max):
"""
Given the maximum ell value, this function returns the indexes to sort the
indexes of the spherical harmonics coefficients when going from real to
complex representation and viceversa.
The complex representation is assumed to be sorted as in map2alm of healpy
(see https://healpy.readthedocs.io/en/latest/) i.e. according to (m, l).
The ouput allows to sort the real representation according to (l, m).
Parameters:
-----------
l_max : int
Maximum ell value.
Returns:
--------
l_grid : numpy.ndarray
Array of l values.
m_grid : numpy.ndarray
Array of m values.
ll : numpy.ndarray
Array of l values corresponding to the sorted indexes.
mm : numpy.ndarray
Array of m values corresponding to the sorted indexes.
sort_indexes : numpy.ndarray
Array of indexes to sort the spherical harmonics coefficients.
"""
# Create arrays for l and m
l_values = np.arange(l_max + 1)
m_values = np.arange(l_max + 1)
# Create a grid of all possible (l, m) pairs
l_grid, m_grid = np.meshgrid(l_values, m_values, indexing="xy")
# Flatten the grid
l_flat = l_grid.flatten()
m_flat = m_grid.flatten()
# Select only the m values that are allowed for a given ell\
l_grid = l_flat[np.abs(m_flat) <= l_flat]
m_grid = m_flat[np.abs(m_flat) <= l_flat]
# Create a vector with all the m<0 and then all the m>=0
mm = np.append(-np.flip(m_grid[m_grid > 0]), m_grid)
# Create a vector with all the ls corresponding to mm
ll = np.append(np.flip(l_grid[m_grid > 0]), l_grid)
# Return the sorted indexes
return l_grid, m_grid, ll, mm, np.lexsort((mm, ll))
def spherical_harmonics_projection(quantity, l_max):
"""
Compute the spherical harmonics projection of a given quantity. Quantity
should be an array in pixel space, and compatible with healpy (see
https://healpy.readthedocs.io/en/latest/). The spherical harmonics
coefficients are sorted as described in the get_sort_indexes function in
utils.
Parameters:
-----------
quantity : numpy.ndarray
Array of quantities to project on spherical harmonics.
l_max : int
Maximum ell value.
Returns:
--------
real_alm : numpy.ndarray
Array of real spherical harmonics coefficients, with len
lm = (l_max + 1)**2 is the number of spherical harmonics coefficients.
"""
# Get the complex alm coefficients.
# These are sorted with the m values and are only for m>=0
alm = hp.map2alm(quantity, lmax=l_max)
# Create arrays the m_values and the indexes to sort
inds = get_sort_indexes(l_max)
# Unpack m_grid and sorted_indexes
m_grid = inds[1]
sorted_indexes = inds[-1]
# Compute the sign for the m > 0 values
sign = (-1.0) ** m_grid[m_grid > 0]
# The m != 0 values are multiplied by sqrt(2) and then take real/imag part
positive_alm = np.sqrt(2.0) * alm[m_grid > 0]
# Build the real alm selecting imaginary and real part and sort in the two
# blocks in ascending order
negative_m = np.flip(sign * positive_alm.imag)
positive_m = sign * positive_alm.real
# Concatenate the negative, zero and positive m values
real_alm = np.concatenate((negative_m, alm[m_grid == 0.0].real, positive_m))
# Sort with the indexes
return real_alm[sorted_indexes]
def complex_to_real_conversion(spherical_harmonics):
"""
Converts the complex spherical harmonics (or the coefficients) to real
spherical harmonics (or the coefficients).
Parameters:
-----------
spherical_harmonics : numpy.ndarray
2D (or 1D) array of complex spherical harmonics coefficients.
If 2D, the shape is (lm, pp), where lm runs over l,m (with m > 0), and
pp is the number of theta and phi values. If 1D, the shape is (lm,).
Returns:
--------
all_spherical_harmonics : numpy.ndarray
2D (or 1D) array of real spherical harmonics coefficients. If 2D, the
shape is (lm, pp), where lm runs over l,m (with -l <= m <= l), and pp
is the number of theta and phi values. If 1D, the shape is (lm,).
"""
# Get the right value of l_max from the input complex coefficients
l_max = get_l_max_complex(spherical_harmonics)
# Create arrays the m_values and the indexes to sort
_, m_grid, _, _, sort_indexes = get_sort_indexes(l_max)
# Pick only m = 0
zero_m = spherical_harmonics[m_grid == 0.0].real
# Compute the sign for the m > 0 values
sign = (-1.0) ** m_grid[m_grid > 0]
# The m != 0 values are multiplied by sqrt(2) and then take real/imag part
positive_spherical = np.sqrt(2.0) * spherical_harmonics[m_grid > 0.0]
# Build the m > 0 values
positive_m = np.einsum("i,i...->i...", sign, positive_spherical.real)
# Build the m < 0 values
negative_m = np.einsum("i,i...->i...", sign, positive_spherical.imag)
# Concatenate the negative, zero and positive m values
all_spherical_harmonics = np.concatenate(
(np.flip(negative_m, axis=0), zero_m, positive_m), axis=0
)
# Return spherical harmonics (coefficients) sorted by l and m
return all_spherical_harmonics[sort_indexes]
def real_to_complex_conversion(real_spherical_harmonics):
"""
Converts the real spherical harmonics (or the coefficients) back to complex
spherical harmonics (or the coefficients).
Parameters:
-----------
real_spherical_harmonics : numpy.ndarray
1D array of real spherical harmonics coefficients.
The shape is (lm,), where lm runs over l,m (with -l <= m <= l).
l_max : int
Maximum ell value.
Returns:
--------
complex_spherical_harmonics : numpy.ndarray
1D array of complex spherical harmonics coefficients.
The shape is (lm,), where lm runs over l,m (with m >= 0).
"""
# Get the right value of l_max from the input real coefficients
l_max = get_l_max_real(real_spherical_harmonics)
# Get sort indexes
_, _, _, mm, sort_indexes = get_sort_indexes(l_max)
# Reorder the input real coefficients to the original order
ordered_real_spherical_harmonics = np.zeros_like(real_spherical_harmonics)
ordered_real_spherical_harmonics[sort_indexes] = real_spherical_harmonics
# Split the ordered real coefficients into negative, zero, and positive m values
zero_m = ordered_real_spherical_harmonics[mm == 0]
positive_m = ordered_real_spherical_harmonics[mm > 0]
negative_m = ordered_real_spherical_harmonics[mm < 0]
# Compute the corresponding m values for positive and negative m
m_positive = mm[mm > 0]
# Reconstruct the complex coefficients
complex_positive_m = (positive_m + 1j * negative_m[::-1]) / (
np.sqrt(2.0) * (-1.0) ** m_positive
)
# Combine zero and positive m values to form the full complex coefficients
complex_spherical_harmonics = np.concatenate(
(zero_m, complex_positive_m), axis=0
)
return complex_spherical_harmonics
def get_real_spherical_harmonics(l_max, theta, phi):
"""
Compute the real spherical harmonics for a given maximum ell value and for
a given set of theta and phi values.
Parameters:
-----------
l_max : int
Maximum ell value.
theta : numpy.ndarray or jax.numpy.ndarray
Array of polar angles (co-latitudes).
phi : numpy.ndarray or jax.numpy.ndarray
Array of azimuthal angles (longitudes).
Returns:
--------
all_spherical_harmonics : numpy.ndarray
2D array of spherical harmonics computed for the given maximum ell
value, and theta and phi values. The shape will be (lm, pp), where pp
is the number of theta and phi values, and lm = (l_max + 1)**2 is the
number of spherical harmonics coefficients.
"""
# Create arrays the m_values and the indexes to sort
inds = get_sort_indexes(l_max)
# Unpack m_grid and sorted_indexes
l_grid = inds[0]
m_grid = inds[1]
# Compute all the spherical harmonics
spherical_harmonics = sph_harm(
m_grid[:, None], l_grid[:, None], phi[None, :], theta[None, :]
)
# Return sorted
return complex_to_real_conversion(spherical_harmonics)
def get_CL_from_real_clm(clm_real):
"""
Compute the angular power spectrum from the spherical harmonics coefficients.
Parameters:
-----------
clm_real : numpy.ndarray
Array of spherical harmonics coefficients, if dimension > 1 the first
axis must run over the coefficients.
Returns:
--------
CL : numpy.ndarray
Array of angular power spectrum.
"""
# Get the shape of the input coefficients
clm_shape = clm_real.shape
# Get the maximum ell value
l_max = get_l_max_real(clm_real)
# Compute the angular power spectrum
CL = np.zeros(tuple([l_max + 1] + list(clm_shape[1:])))
# A counter for the coefficients used up to that value of ell
i = 0
for ell in range(0, l_max + 1):
# Average the square of the coefficients for all ms at fixed ell
CL[ell] = np.mean(clm_real[i : i + 2 * ell + 1] ** 2, axis=0)
# Update the counter
i += 2 * ell + 1
return CL
def get_Cl_limits(
means,
cov,
shape_params,
n_points=int(1e4),
limit_cl=0.95,
max_iter=100,
prior=5.0 / (4.0 * np.pi),
):
"""
Compute the upper limit on the angular power spectrum from the means and
covariance matrix of the spherical harmonics coefficients.
Parameters:
-----------
means : numpy.ndarray
Array of means for the spherical harmonics coefficients.
cov : numpy.ndarray
Array of covariance matrix for the spherical harmonics coefficients.
shape_params : int
Number of parameters for the SGWB shape.
n_points : int, optional
Number of points to generate.
limit_cl : float, optional
Quantile to compute the upper limit.
max_iter : int, optional
Maximum number of iterations to generate points.
prior : float, optional
Prior value to restrict the points.
Returns:
--------
Cl_limits : numpy.ndarray
Array of upper limits on the angular power spectrum from the covariance
Cl_limits_prior : numpy.ndarray
Array of upper limits on the angular power spectrum including the prior
"""
# Generate gaussian data from the covariance matrix
data = np.random.multivariate_normal(
means, cov, n_points, check_valid="ignore", tol=1e-4
)
# Select only the points that are within the prior
data_prior = data[np.max(np.abs(data[:, shape_params:]), axis=-1) <= prior]
# Initialize the counter and the length of the data
i_add = 0
len_restricted = len(data_prior)
# Use a while loop to generate enough points
while len_restricted < n_points and i_add < max_iter:
# Generate more points
add_data = np.random.multivariate_normal(
means, cov, 10 * n_points, check_valid="ignore", tol=1e-4
)
# Select only the points that are within the prior and append
data_prior = np.append(
data_prior,
add_data[
np.max(np.abs(add_data[:, shape_params:]), axis=-1) <= prior
],
axis=0,
)
# Update the counter and the length of the data
len_restricted = len(data_prior)
i_add += 1
# Compute the angular power spectra without and with the prior
correlations_lm = get_CL_from_real_clm(data.T[shape_params - 1 :])[1:]
correlations_lm_prior = get_CL_from_real_clm(
data_prior.T[shape_params - 1 :]
)[1:]
# Compute the upper limits without the prior
Cl_limits = np.quantile(correlations_lm, limit_cl, axis=-1)
# And with the prior if there are enough points
if len_restricted == 0:
Cl_limits_prior = np.nan * Cl_limits
else:
Cl_limits_prior = np.quantile(correlations_lm_prior, limit_cl, axis=-1)
return Cl_limits, Cl_limits_prior
def sqrt_to_lin_conversion(gLM_grid, l_max_lin=-1, real_basis_input=False):
"""
Convert the sqrt basis to the linear basis.
Parameters:
-----------
gLM_grid : numpy.ndarray
Array of sqrt basis coefficients.
l_max_lin : int
Maximum ell value for the linear basis.
real_basis_input : bool
If True, the input is in the real basis. Default is False.
Returns:
--------
clm_real : numpy.ndarray
Array of real coefficients in the linear basis.
"""
# If gLM are in the real basis convert the complex basis
if real_basis_input:
gLM_complex = real_to_complex_conversion(gLM_grid)
else:
gLM_complex = gLM_grid
# Get the maximum ell value for the sqrt basis
l_max_sqrt = get_l_max_complex(gLM_complex)
# Get the grid of all possible L values for the sqrt basis
L_grid_all = np.arange(l_max_sqrt + 1)
# If the maximum ell value for the linear basis is not provided, set it
if l_max_lin < 0:
l_max_lin = 2 * l_max_sqrt
# Get the number of coefficients for the linear basis
n_coefficients = get_n_coefficients_complex(l_max_lin)
# Initialize the array for the linear basis
clm_complex = np.zeros(n_coefficients, dtype=np.cdouble)
# Get the indexes for the linear and sqrt basis
l_lin, m_lin, _, _, _ = get_sort_indexes(l_max_lin)
l_sqrt, m_sqrt, _, _, _ = get_sort_indexes(l_max_sqrt)
# Compute the sign for the gLM with m < 0
gLnegM_complex = (-1) ** np.abs(m_sqrt) * np.conj(gLM_complex)
for ind_linear in range(len(m_lin)):
# Get the values of ell and m
m = m_lin[ind_linear]
ell = l_lin[ind_linear]
for L1 in L_grid_all:
# Build a mask using the conditions from the selection rules
mask_L2 = (np.abs(L1 - L_grid_all) <= ell) * (
L_grid_all >= ell - L1
)
# Run over the L2 allowed by the mask
for L2 in L_grid_all[mask_L2]:
# Compute the Clebsch-Gordan coefficient for all ms = 0
cg0 = clebsch_gordan(L1, 0, L2, 0, ell, 0)
# If the coefficient is not zero compute the prefactor
if cg0 != 0.0:
prefac = np.sqrt(
(2.0 * L1 + 1.0)
* (2.0 * L2 + 1.0)
/ (4.0 * np.pi * (2.0 * ell + 1.0))
)
# These are all the values of M1 to use
M1_grid_all = np.arange(-L1, L1 + 1)
# Enforce m +M1 + M2 = 0
M2_grid_all = m - M1_grid_all
# Check that the values of M2 are consistent with L2
mask_M = np.abs(M2_grid_all) <= L2
# Apply the mask
M1_grid = M1_grid_all[mask_M]
M2_grid = M2_grid_all[mask_M]
for iM in range(len(M1_grid)):
# Get the values of M1 and M2
M1 = M1_grid[iM]
M2 = M2_grid[iM]
# Compute the Clebsch-Gordan coefficient for ms neq 0
cg1 = clebsch_gordan(L1, M1, L2, M2, ell, m)
# Mask to get the corresponding value of gLM_complex
b1_mask = (l_sqrt == L1) & (m_sqrt == np.abs(M1))
b2_mask = (l_sqrt == L2) & (m_sqrt == np.abs(M2))
# Get the values of gLM_complex for the given L and M
b1 = (
gLM_complex[b1_mask]
if M1 >= 0
else gLnegM_complex[b1_mask]
)
b2 = (
gLM_complex[b2_mask]
if M2 >= 0
else gLnegM_complex[b2_mask]
)
# Multiply everything and sum to the right index
clm_complex[ind_linear] += prefac * cg0 * cg1 * b1 * b2
return complex_to_real_conversion(clm_complex)
@jax.jit
def compute_inverse(matrix):
"""
A function to compute the inverse of a matrix. Applies rescaling to
maintain numerical stability, especially for near-singular matrices.
Parameters:
-----------
matrix : numpy.ndarray or jax.numpy.ndarray
Input matrix to compute the inverse of. The shape is assumed to be
(..., N, N) and the inverse is computed on the last 2 indexes
Returns:
--------
c_inverse : numpy.ndarray or jax.numpy.ndarray
Inverse of the input matrix.
Notes:
------
Assumes the input matrix is a square matrix on the last 2 axes.
"""
# Defines the matrix for the rescaling using the elements on the diagonal
rescaling_vec = jnp.sqrt(jnp.diagonal(matrix, axis1=-2, axis2=-1))
rescaling = rescaling_vec[..., :, None] * rescaling_vec[..., None, :]
return jnp.linalg.inv(matrix / rescaling) / rescaling
def get_R(samples):
"""
Computes the Gelman-Rubin (GR) statistic for convergence assessment. The
GR statistic is a convergence diagnostic used to assess whether multiple
Markov chains have converged to the same distribution. Values close to 1
indicate convergence. For details see
https://en.wikipedia.org/wiki/Gelman-Rubin_statistic
Parameters:
-----------
samples : numpy.ndarray
Array containing MCMC samples with dimensions
(N_steps, N_chains, N_parameters).
Returns:
--------
R : numpy.ndarray
Array containing the Gelman-Rubin statistics indicating convergence for
the different parameters. Values close to 1 indicate convergence.
"""
# Get the shapes
N_steps, N_chains, N_parameters = samples.shape
# Chain means
chain_mean = np.mean(samples, axis=0)
# Global mean
global_mean = np.mean(chain_mean, axis=0)
# Variance between the chain means
variance_of_means = (
N_steps
/ (N_chains - 1)
* np.sum((chain_mean - global_mean[None, :]) ** 2, axis=0)
)
# Variance of the individual chain across all chains
intra_chain_variance = np.std(samples, axis=0, ddof=1) ** 2
# And its averaged value over the chains
mean_intra_chain_variance = np.mean(intra_chain_variance, axis=0)
# First term
term_1 = (N_steps - 1) / N_steps
# Second term
term_2 = variance_of_means / mean_intra_chain_variance / N_steps
# This is the R (as a vector running on the paramters)
return term_1 + term_2
| 24,880 | Python | .py | 621 | 33.028986 | 84 | 0.639474 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,340 | __init__.py | Mauropieroni_fastPTA/fastPTA/__init__.py | __author__ = "Mauro Pieroni"
__version__ = "0.1"
__year__ = "2024"
__url__ = "https://github.com/Mauropieroni/fastPTA"
| 119 | Python | .py | 4 | 28.75 | 51 | 0.617391 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,341 | plotting_functions.py | Mauropieroni_fastPTA/fastPTA/plotting_functions.py | # Global
import corner
import matplotlib
import matplotlib.patches
import matplotlib.pyplot as plt
import numpy as np
from scipy.special import legendre
from fastPTA import get_tensors as gt
# Setting plotting parameters
matplotlib.rcParams["text.usetex"] = True
plt.rc("xtick", labelsize=20)
plt.rc("ytick", labelsize=20)
plt.rcParams.update(
{"axes.labelsize": 20, "legend.fontsize": 20, "axes.titlesize": 22}
)
# Some other useful things
cmap_HD = matplotlib.colormaps["coolwarm"] # type: ignore
cmap1_grid = matplotlib.colormaps["hot"] # type: ignore
cmap2_grid = matplotlib.colormaps["PiYG"] # type: ignore
my_colormap = {
"red": "#EE6677",
"green": "#228833",
"blue": "#4477AA",
"yellow": "#CCBB44",
"purple": "#AA3377",
"cyan": "#66CCEE",
"gray": "#BBBBBB",
}
def plot_HD_Legendre(x_points, data_HD_coeffs, plot_suplabel):
x = np.linspace(-1, 1, x_points)
p_sum = np.zeros(shape=(x_points, len(data_HD_coeffs)))
for i in range(data_HD_coeffs.shape[-1]):
pol_i = data_HD_coeffs[None, :, i] * (legendre(i)(x))[:, None]
p_sum += pol_i
res = np.quantile(p_sum, [0.025, 0.16, 0.5, 0.84, 0.975], axis=-1)
plt.figure(figsize=(6, 4))
plt.plot(np.arccos(x), res[2, :], color="dimgrey", label=r"$\rm Injection$")
plt.fill_between(
np.arccos(x),
y1=res[1, :],
y2=res[-2, :],
color="lightskyblue",
label=r"$1 \sigma$",
zorder=-1,
)
plt.fill_between(
np.arccos(x),
y1=res[0, :],
y2=res[-1, :],
color="dodgerblue",
label=r"$2 \sigma$",
zorder=-2,
)
plt.xlabel(r"$\zeta_{IJ} \equiv \arccos(\hat{p}_I \cdot \hat{p}_J)$")
plt.xticks(
np.pi / 4 * np.arange(5),
labels=[r"$0$", r"$\pi/4$", r"$\pi/2$", r"$3 \pi / 4$", r"$\pi$"],
)
plt.xlim(0, np.pi)
plt.ylim(-0.4, 0.75)
plt.legend(fontsize=20, loc=9)
plt.suptitle(plot_suplabel, fontsize=20)
plt.tight_layout()
def plot_HD_binned(data_HD_coeffs, HD_coeffs, plot_suplabel):
plt.figure(figsize=(6, 4))
bin_edges = np.linspace(0, np.pi, len(HD_coeffs) + 1)
bin_val = 0.5 * (bin_edges[1:] + bin_edges[:-1])
plt.violinplot(
data_HD_coeffs,
widths=0.2,
positions=bin_val,
showextrema=True,
quantiles=[[0.16, 0.5, 0.84] for xx in range(len(HD_coeffs))],
)
plt.scatter(bin_val, HD_coeffs, color="dimgrey", label=r"$\rm Injection$")
plt.plot(
np.arccos(gt.x),
gt.HD_value,
color="red",
linestyle="--",
label="HD curve",
)
plt.xlabel(r"$\zeta_{IJ} \equiv \arccos(\hat{p}_I \cdot \hat{p}_J)$")
plt.xticks(
np.pi / 4 * np.arange(5),
labels=[r"$0$", r"$\pi/4$", r"$\pi/2$", r"$3 \pi / 4$", r"$\pi$"],
)
plt.xlim(0, np.pi)
plt.legend(fontsize=20, loc=9)
plt.suptitle(plot_suplabel, fontsize=20)
plt.tight_layout()
def plot_corner(
samples,
colors=False,
smooth=False,
weights=False,
fig=False,
chain_labels=False,
show_titles=False,
plot_density=True,
plot_datapoints=False,
fill_contours=True,
bins=25,
title_kwargs={"fontsize": 20, "pad": 12},
hist_kwargs={"density": True, "linewidth": 2},
**kwargs,
):
if not colors:
vals = list(my_colormap.values())
colors = []
for i in range(len(samples)):
colors = [
vals[np.mod(i, len(samples))] for i in range(len(samples))
]
if not smooth:
smooth = np.zeros(len(samples))
if not weights:
weights = [None for i in range(len(samples))]
if not fig:
fig = plt.figure(figsize=(10, 8))
if not chain_labels:
chain_labels = [None for i in range(len(samples))]
for i in range(len(samples)):
hist_kwargs["color"] = colors[i] # type: ignore
if samples[i].shape[-1] > 1:
fig = corner.corner(
samples[i],
color=colors[i], # type: ignore
smooth=smooth[i], # type: ignore
weights=weights[i], # type: ignore
fig=fig,
show_titles=show_titles,
plot_density=plot_density,
plot_datapoints=plot_datapoints,
fill_contours=fill_contours,
bins=bins,
title_kwargs=title_kwargs,
hist_kwargs=hist_kwargs,
**kwargs,
)
else:
plt.hist(
samples[i],
color=colors[i], # type: ignore
weights=weights[i], # type: ignore
bins=bins,
histtype="step",
label=chain_labels[i], # type: ignore
density=True,
)
if "truths" in kwargs.keys():
truths = kwargs["truths"]
if "truth_color" in kwargs.keys():
truth_color = kwargs["truth_color"]
else:
truth_color = "black"
plt.axvline(truths, color=truth_color)
custom_lines = []
if samples[0].shape[-1] > 1:
for i in range(len(chain_labels)): # type: ignore
custom_lines.append(
matplotlib.patches.Patch(
facecolor=colors[i], label=chain_labels[i] # type: ignore
)
)
plt.legend(
handles=custom_lines, bbox_to_anchor=(0.0, 1.0, 1.0, 1.0), loc=0
)
else:
plt.legend(loc=0)
| 5,573 | Python | .py | 170 | 24.594118 | 80 | 0.548183 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,342 | generate_new_pulsar_configuration.py | Mauropieroni_fastPTA/fastPTA/generate_new_pulsar_configuration.py | # Global
import numpy as np
import pandas as pd
# Local
from fastPTA.utils import (
default_pulsar_parameters,
path_to_default_NANOGrav_positions,
)
def generate_parameter(n_pulsars, parameter_dict):
"""
Generate the parameters for a given number of pulsars based on specified
info in parameter_dict. Supported distribution types are 'gaussian' and
'uniform'.
Parameters:
-----------
n_pulsars : int
Number of pulsars for which parameters will be generated.
parameter_dict : dict
Dictionary containing parameters for generating data.
It should have the following keys:
- "which_distribution": str, the specific distribution to use.
- "mean": float, mean value for Gaussian distribution.
- "std": float, standard deviation for Gaussian distribution.
- "min": float, minimum value for uniform distribution.
- "max": float, maximum value for uniform distribution.
Returns:
--------
data : numpy.ndarray or jax.numpy.ndarray
Generated parameters for the catalog of pulsars.
"""
if parameter_dict["which_distribution"] == "gaussian":
data = np.random.normal(
parameter_dict["mean"], parameter_dict["std"], n_pulsars
)
elif parameter_dict["which_distribution"] == "uniform":
data = np.random.uniform(
parameter_dict["min"], parameter_dict["max"], n_pulsars
)
else:
raise ValueError(
"Cannot use that pdf", parameter_dict["which_distribution"]
)
return data
def generate_noise_parameter(
n_pulsars,
noise_parameters,
noise_probabilities,
do_filter=False,
):
"""
Generate noise parameters for a given number of pulsars based on specified
distributions. Each noise parameter is generated independently from a
uniform distribution defined by its min and max values. If do_filter is
True, noise parameters are filtered based on noise_probabilities.
Parameters:
-----------
n_pulsars : int
Number of pulsars for which noise parameters needs to be generated.
noise_parameters : list of dict
List of dictionaries containing parameters for generating noise data.
Each dictionary should have keys:
- "min": float, minimum value for noise parameter.
- "max": float, maximum value for noise parameter.
noise_probabilities : numpy.ndarray or jax.numpy.ndarray
Array of probabilities corresponding to each noise parameter.
do_filter : bool, optional
Flag indicating whether to apply filtering to noise parameters,
default is False.
Returns:
--------
noise_vals : numpy.ndarray or jax.numpy.ndarray
Generated noise parameters for the pulsars.
"""
# Unpack minimal and maximal values
mins = np.array([n["min"] for n in noise_parameters])
maxs = np.array([n["max"] for n in noise_parameters])
# Generate the noise parameters
noise_vals = np.random.uniform(mins, maxs, size=(n_pulsars, len(mins)))
if do_filter:
# Generate acceptance probabilities
acceptance_prob = np.random.uniform(0, 1, size=(n_pulsars, len(mins)))
noise_prob_filter = np.zeros(shape=(n_pulsars, len(mins)))
noise_prob_filter[acceptance_prob < noise_probabilities[None, :]] = 1
# Accepts/rejects depending on the acceptance probabilities
for i in range(len(noise_prob_filter)):
if np.all(
noise_prob_filter[i] == np.zeros(len(noise_prob_filter[i]))
):
val = np.random.uniform(0, 1)
for i in range(len(mins)):
if val < noise_probabilities[0]:
noise_prob_filter[i, 0] = 1.0
break
noise_prob_filter = np.log10(noise_prob_filter + 1e-30)
noise_vals += noise_prob_filter
return noise_vals
def generate_pulsars_catalog(
n_pulsars=30,
dt_dict=default_pulsar_parameters["dt_dict"],
T_span_dict=default_pulsar_parameters["T_span_dict"],
wn_dict=default_pulsar_parameters["wn_dict"],
dm_noise_log_10_A_dict=default_pulsar_parameters["dm_noise_log_10_A_dict"],
red_noise_log_10_A_dict=default_pulsar_parameters[
"red_noise_log_10_A_dict"
],
sv_noise_log_10_A_dict=default_pulsar_parameters["sv_noise_log_10_A_dict"],
dm_noise_g_dict=default_pulsar_parameters["dm_noise_g_dict"],
red_noise_g_dict=default_pulsar_parameters["red_noise_g_dict"],
sv_noise_g_dict=default_pulsar_parameters["sv_noise_g_dict"],
noise_probabilities_dict=default_pulsar_parameters[
"noise_probabilities_dict"
],
save_catalog=False,
outname="pulsar_configurations/new_pulsars_catalog.txt",
use_ng_positions=False,
):
"""
Generate a catalog of pulsars with specified parameters (mostly timing and
noise characteristics). The generated catalog is returned as a pandas
DataFrame. If save_catalog is True, the generated catalog is saved to the
specified output file.
Parameters:
-----------
n_pulsars : int, optional
Number of pulsars to generate in the catalog, default is 30.
dt_dict : dict, optional
Dictionary containing parameters for generating pulsar dt.
T_span_dict : dict, optional
Dictionary containing parameters for generating pulsar T_span.
wn_dict : dict, optional
Dictionary containing parameters for generating pulsar white noise.
dm_noise_log_10_A_dict : dict, optional
Dictionary containing parameters for generating log10_A.for DM noise
red_noise_log_10_A_dict : dict, optional
Dictionary containing parameters for generating log10_A.for red noise
sv_noise_log_10_A_dict : dict, optional
Dictionary containing parameters for generating log10_A for sv noise
dm_noise_g_dict : dict, optional
Dictionary containing parameters for generating the tilt for DM noise.
red_noise_g_dict : dict, optional
Dictionary containing parameters for generating the tilt for red noise.
sv_noise_g_dict : dict, optional
Dictionary containing parameters for generating the tilt for sv noise.
noise_probabilities_dict : dict, optional
Dictionary containing noise probabilities.
save_catalog : bool, optional
Flag indicating whether to save the generated catalog to a file,
default is False.
outname : str, optional
Name of the output file to save the generated catalog,
default is "pulsar_configurations/test_pulsars.txt".
use_ng_positions : bool, optional
Flag indicating whether to use NANOGrav positions for pulsars,
default is False.
Returns:
--------
DataFrame
Generated catalog of pulsars with specified parameters.
"""
noise_probabilities = np.array(list(noise_probabilities_dict.values()))
normed = noise_probabilities / np.sum(noise_probabilities)
catalog = {}
if use_ng_positions:
data_ng = np.loadtxt(
path_to_default_NANOGrav_positions, skiprows=1, dtype=np.str_
)
n_pulsars = len(data_ng)
catalog["names"] = data_ng[:, 0]
catalog["phi"] = data_ng[:, 1].astype(float)
catalog["theta"] = data_ng[:, 2].astype(float)
else:
catalog["names"] = [
"pulsar_" + str(ell) for ell in (1 + np.arange(n_pulsars))
]
catalog["phi"] = np.random.uniform(0.0, 2 * np.pi, n_pulsars)
catalog["theta"] = np.arccos(np.random.uniform(-1, 1, n_pulsars))
catalog["dt"] = 10 ** generate_parameter(n_pulsars, dt_dict)
catalog["Tspan"] = 10 ** generate_parameter(n_pulsars, T_span_dict)
catalog["wn"] = 10 ** generate_parameter(n_pulsars, wn_dict)
catalog["log10_A_dm"], catalog["log10_A_red"], catalog["log10_A_sv"] = (
generate_noise_parameter(
n_pulsars,
[
dm_noise_log_10_A_dict,
red_noise_log_10_A_dict,
sv_noise_log_10_A_dict,
],
normed,
do_filter=True,
).T
)
catalog["g_dm"], catalog["g_red"], catalog["g_sv"] = (
generate_noise_parameter(
n_pulsars,
[
dm_noise_g_dict,
red_noise_g_dict,
sv_noise_g_dict,
],
normed,
do_filter=False,
).T
)
DF = pd.DataFrame(catalog)
if save_catalog:
DF.to_csv(outname, sep=" ", index=False)
return DF
| 8,644 | Python | .py | 207 | 33.917874 | 79 | 0.650994 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,343 | test_catalog_generator.py | Mauropieroni_fastPTA/tests/test_catalog_generator.py | # Global
import unittest
import numpy as np
from scipy.stats import kstest
from scipy import stats as scipy_stats
# Local
import utils as tu
from fastPTA.generate_new_pulsar_configuration import generate_pulsars_catalog
@tu.not_a_test
def test_generation(parameter, n_pulsars, pulsar_dictionary):
test_dictionary = {**tu.test_distributions}
for k, v in pulsar_dictionary.items():
if k != "noise_probabilities_dict":
test_dictionary[k] = v
test_catalog = generate_pulsars_catalog(
n_pulsars=n_pulsars, save_catalog=False, **pulsar_dictionary
)
x = test_catalog[tu.parameters_to_test[parameter]]
x = x[x > -40]
if parameter in ["dt", "T_span", "wn"]:
x = np.log10(x)
elif parameter in ["theta"]:
x = np.cos(x)
dist = test_dictionary[parameter + "_dict"]
# This is a Kolmogorov–Smirnov test to check whether the pulsars in the catalog are distributed as expected
if dist["which_distribution"] == "uniform" and len(x) > 3:
_stats, p = kstest(
x,
scipy_stats.uniform(
dist["min"], scale=dist["max"] - dist["min"]
).cdf,
)
elif dist["which_distribution"] == "gaussian" and len(x) > 3:
_stats, p = kstest(
x, scipy_stats.norm(dist["mean"], scale=dist["std"]).cdf
)
elif len(x) <= 3:
p = 1.0
else:
raise ValueError("Cannot use distribution", dist["which_distribution"])
return p
class TestCatalogGenerator(unittest.TestCase):
def test_generation_EPTA(self):
"""
Test the generation of pulsar catalogs with EPTA-like noise parameters
"""
for p in tu.parameters_to_test.keys():
self.assertTrue(test_generation(p, 30, tu.EPTAlike_test) > 1e-4) # type: ignore
def test_generation_EPTA_noiseless(self):
"""
Test the generation of a noiseless pulsar catalogs
"""
for p in tu.parameters_to_test.keys():
self.assertTrue(
test_generation(p, 50, tu.EPTAlike_noiseless_test) > 1e-4 # type: ignore
)
def test_generation_SKAlike(self):
"""
Test the generation of pulsar catalogs with EPTA-like noise parameters
"""
for p in tu.parameters_to_test.keys():
self.assertTrue(test_generation(p, 500, tu.mockSKA10_test) > 1e-4) # type: ignore
def test_generation_ng_positions(self):
"""
Test the generation of pulsar catalogs with NANOgrav positions
"""
test_catalog = generate_pulsars_catalog(
n_pulsars=30, save_catalog=False, use_ng_positions=True
)
data_ng = np.loadtxt(tu.NANOGrav_positions, skiprows=1, dtype=np.str_)
self.assertEqual(
np.sum(np.abs(test_catalog["phi"] - data_ng[:, 1].astype(float))),
0.0,
)
self.assertEqual(
np.sum(np.abs(test_catalog["theta"] - data_ng[:, 2].astype(float))),
0.0,
)
if __name__ == "__main__":
unittest.main(verbosity=2)
| 3,110 | Python | .py | 80 | 30.7625 | 113 | 0.614948 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,344 | test_utils.py | Mauropieroni_fastPTA/tests/test_utils.py | # Global
import unittest
import numpy as np
import healpy as hp
from scipy.special import sph_harm
# Local
import utils as tu
from fastPTA import utils as ut
nside = 64
npix = hp.nside2npix(nside)
theta, phi = hp.pix2ang(nside, np.arange(npix))
class TestGetTensors(unittest.TestCase):
def test_get_sort_indexes(self):
"""
Test the function to get the (l,m) pairs sorted correctly
"""
data = np.loadtxt(tu.lm_indexes)
inds = ut.get_sort_indexes(5)
self.assertTrue(np.allclose(inds[2][inds[-1]], data[:, 0]))
self.assertTrue(np.allclose(inds[3][inds[-1]], data[:, 1]))
def test_complex_to_real(self):
"""
Test the function to go from complex to real spherical harmonics coefficients assuming complex are sorted according to the healpy scheme
"""
l_max = 2
n_coefficients = ut.get_n_coefficients_complex(l_max)
complex_vals = np.random.normal(
0.0, 1.0, n_coefficients
) + 1j * np.random.normal(0.0, 1.0, n_coefficients)
test_real_vals = np.array(
[
complex_vals[0].real,
-np.sqrt(2) * complex_vals[3].imag,
complex_vals[1].real,
-np.sqrt(2) * complex_vals[3].real,
np.sqrt(2) * complex_vals[5].imag,
-np.sqrt(2) * complex_vals[4].imag,
complex_vals[2].real,
-np.sqrt(2) * complex_vals[4].real,
np.sqrt(2) * complex_vals[5].real,
]
)
real_vals = ut.complex_to_real_conversion(complex_vals)
self.assertTrue(np.allclose(real_vals, test_real_vals))
def test_complex_to_real_to_complex(self):
"""
Check that starting from complex coefficients the operations commute
"""
l_max = 2
n_coefficients = ut.get_n_coefficients_complex(l_max)
m_grid = np.array([0, 0, 0, 1, 1, 2], dtype=int)
complex_vals = np.random.normal(
0.0, 1.0, n_coefficients
) + 1j * np.random.normal(0.0, 1.0, n_coefficients)
complex_vals[m_grid == 0] = complex_vals[m_grid == 0].real
real_vals = ut.complex_to_real_conversion(complex_vals)
test_complex_vals = ut.real_to_complex_conversion(real_vals)
self.assertTrue(np.allclose(complex_vals, test_complex_vals))
def test_get_real_spherical_harmonics(self, l_max=5):
"""
Test the function to get the spherical harmonics
"""
sp_harm = ut.get_real_spherical_harmonics(l_max, theta, phi)
c = 0
for ell in range(l_max + 1):
for m in range(-ell, ell + 1):
sp = sph_harm(np.abs(m), ell, phi, theta)
if m == 0:
sp = sp.real
elif m > 0:
sp = np.sqrt(2.0) * (-1.0) ** m * sp.real
elif m < 0:
sp = np.sqrt(2.0) * (-1.0) ** m * sp.imag
else:
raise ValueError("Nope")
# Checks that (with the correct normalization) the scalar
# product is 1 withing 3 decimal places
self.assertAlmostEqual(
np.abs(np.mean(4 * np.pi * sp_harm[c] * sp) - 1.0),
0.0,
places=3,
)
c += 1
def test_spherical_harmonics_multipoles(self, l_max=3):
"""
Test the spherical harmonics projection function
"""
inds = ut.get_sort_indexes(l_max)
ll = inds[2][inds[-1]]
mm = inds[3][inds[-1]]
for ind in range(len(ll)):
YY = np.array(sph_harm(np.abs(mm[ind]), ll[ind], phi, theta))
if mm[ind] < 0:
# The m < 0 is the complex conjugate of the m > 0 so need a -1
YY = -np.sqrt(2.0) * (-1.0) ** mm[ind] * YY.imag
elif mm[ind] > 0:
# For the m > 0 we just need the real part of Y
YY = np.sqrt(2.0) * (-1.0) ** mm[ind] * YY.real
else:
# For the m = 0 we just need the real part of Y
YY = YY.real
res_YY = ut.spherical_harmonics_projection(YY, l_max)
self.assertAlmostEqual(np.abs(res_YY[ind] - 1.0), 0.0, places=4)
if __name__ == "__main__":
unittest.main(verbosity=2)
| 4,410 | Python | .py | 105 | 30.714286 | 144 | 0.538101 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,345 | utils.py | Mauropieroni_fastPTA/tests/utils.py | import os
import numpy as np
from fastPTA.utils import load_yaml
test_data_path = os.path.join(os.path.dirname(__file__), "test_data/")
# Default parameters for the pulsars
EPTAlike_test = load_yaml(test_data_path + "/EPTAlike_pulsar_parameters.yaml")
EPTAlike_noiseless_test = load_yaml(
test_data_path + "/EPTAlike_pulsar_parameters_noiseless.yaml"
)
mockSKA10_test = load_yaml(test_data_path + "/mockSKA10_pulsar_parameters.yaml")
NANOGrav_positions = test_data_path + "/NANOGrav_positions.txt"
# Paths to the test data
lm_indexes = test_data_path + "lm_indexes.txt"
test_catalog_path = test_data_path + "test_catalog.txt"
test_catalog_path2 = test_data_path + "test_catalog2.txt"
test_catalog_path3 = test_data_path + "test_catalog3.txt"
get_tensors_data_path = test_data_path + "get_tensors.npz"
get_correlations_lm_IJ_data_path = test_data_path + "get_correlations_lm_IJ.npz"
get_tensors_Binned_data_path = test_data_path + "get_tensors_Binned.npz"
get_tensors_Legendre_data_path = test_data_path + "get_tensors_Legendre.npz"
Fisher_data_path = test_data_path + "Fisher_data.npz"
Fisher_data_path2 = test_data_path + "Fisher_data2.npz"
parameters_to_test = {
"phi": "phi",
"theta": "theta",
"dt": "dt",
"T_span": "Tspan",
"wn": "wn",
"dm_noise_log_10_A": "log10_A_dm",
"red_noise_log_10_A": "log10_A_red",
"sv_noise_log_10_A": "log10_A_sv",
"dm_noise_g": "g_dm",
"red_noise_g": "g_red",
"sv_noise_g": "g_sv",
}
test_distributions = {
"phi_dict": {"which_distribution": "uniform", "min": 0, "max": 2 * np.pi},
"theta_dict": {"which_distribution": "uniform", "min": -1, "max": 1},
}
get_tensor_labels = [
"strain_omega",
"response_IJ",
"HD_functions_IJ",
"HD_coefficients",
]
def not_a_test(object):
object.__test__ = False
return object
test_frequency = np.arange(1e-10, 1e-6, 1e-8)
| 1,881 | Python | .py | 49 | 35.489796 | 80 | 0.683892 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,346 | test_get_tensors.py | Mauropieroni_fastPTA/tests/test_get_tensors.py | # Global
import unittest
import numpy as np
import healpy as hp
from scipy.special import sph_harm
import jax
import jax.numpy as jnp
# Local
import utils as tu
from fastPTA import utils as ut
from fastPTA import get_tensors as gt
jax.config.update("jax_enable_x64", True)
jax.config.update("jax_default_device", jax.devices(ut.which_device)[0])
nside = 64
npix = hp.nside2npix(nside)
theta, phi = hp.pix2ang(nside, jnp.arange(npix))
theta = jnp.array(theta)
phi = jnp.array(phi)
@tu.not_a_test
def get_tensors_and_shapes(
path_to_pulsar_catalog, n_pulsars, HD_basis, HD_order
):
"""
A utility function that returns the results of the get_tensors function and the expected shapes of the results.
"""
res = gt.get_tensors(
tu.test_frequency,
path_to_pulsar_catalog=path_to_pulsar_catalog,
save_catalog=True,
n_pulsars=n_pulsars,
regenerate_catalog=True,
HD_basis=HD_basis,
HD_order=HD_order,
**tu.EPTAlike_test
)
HD_shape = HD_order + 1 if HD_order else HD_order
test_shapes = [
(len(tu.test_frequency), n_pulsars, n_pulsars),
(len(tu.test_frequency), n_pulsars, n_pulsars),
(HD_shape, len(tu.test_frequency), n_pulsars, n_pulsars),
(HD_shape,),
]
return res, test_shapes
@tu.not_a_test
def get_tensors_data(
path_to_pulsar_catalog, HD_basis, HD_order, data_path, anisotropies=False
):
"""
A utility function that returns the results of the get_tensors function and the expected results.
"""
get_tensor_results = gt.get_tensors(
tu.test_frequency,
path_to_pulsar_catalog=path_to_pulsar_catalog,
HD_basis=HD_basis,
HD_order=HD_order,
anisotropies=anisotropies,
)
return get_tensor_results, np.load(data_path)
class TestGetTensors(unittest.TestCase):
def test_gamma(self):
"""
Test the gamma function
"""
npoints = 10
pulsars = 15
theta, phi = hp.pix2ang(
nside, np.linspace(0, npix - 1, npoints, dtype=int)
)
theta_p, phi_p = hp.pix2ang(
nside, np.array(np.random.uniform(0, npix - 1, pulsars), dtype=int)
)
analytical = gt.gamma_analytical(
theta_p,
phi_p,
theta,
phi,
)
p_I = gt.unit_vector(theta_p, phi_p)
hat_k = gt.unit_vector(theta, phi)
gamma = gt.gamma(p_I, hat_k)
self.assertAlmostEqual(
float(jnp.sum(np.abs(gamma - analytical))),
0.0,
places=7,
)
def test_spherical_harmonics_multipoles_2(self):
"""
Test the spherical harmonics projection function for pulsars pulsar matrix
"""
Y00 = np.array([[sph_harm(0.0, 0.0, phi, theta).real]])
Y1m1 = np.array([[np.sqrt(2.0) * sph_harm(-1.0, 1.0, phi, theta).imag]])
Y10 = np.array([[sph_harm(0.0, 1.0, phi, theta).real]])
Y1p1 = np.array([[-np.sqrt(2.0) * sph_harm(1.0, 1.0, phi, theta).real]])
Y2m2 = np.array([[np.sqrt(2.0) * sph_harm(-2.0, 2.0, phi, theta).imag]])
Y2m1 = np.array([[np.sqrt(2.0) * sph_harm(-1.0, 2.0, phi, theta).imag]])
Y20 = np.array([[sph_harm(0.0, 2.0, phi, theta).real]])
Y2p1 = np.array([[-np.sqrt(2.0) * sph_harm(1.0, 2.0, phi, theta).real]])
Y2p2 = np.array([[np.sqrt(2.0) * sph_harm(2.0, 2.0, phi, theta).real]])
Y3m3 = np.array([[np.sqrt(2.0) * sph_harm(-3.0, 3.0, phi, theta).imag]])
Y3p3 = np.array([[-np.sqrt(2.0) * sph_harm(3.0, 3.0, phi, theta).real]])
res_Y00 = gt.projection_spherial_harmonics_basis(Y00, 1)
res_Y1m1 = gt.projection_spherial_harmonics_basis(Y1m1, 1)
res_Y10 = gt.projection_spherial_harmonics_basis(Y10, 1)
res_Y1p1 = gt.projection_spherial_harmonics_basis(Y1p1, 1)
res_Y2m2 = gt.projection_spherial_harmonics_basis(Y2m2, 2)
res_Y2m1 = gt.projection_spherial_harmonics_basis(Y2m1, 2)
res_Y20 = gt.projection_spherial_harmonics_basis(Y20, 2)
res_Y2p1 = gt.projection_spherial_harmonics_basis(Y2p1, 2)
res_Y2p2 = gt.projection_spherial_harmonics_basis(Y2p2, 2)
res_Y3m3 = gt.projection_spherial_harmonics_basis(Y3m3, 3)
res_Y3p3 = gt.projection_spherial_harmonics_basis(Y3p3, 3)
self.assertAlmostEqual(jnp.abs(res_Y00[0] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y1m1[1] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y10[2] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y1p1[3] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y2m2[4] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y2m1[5] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y20[6] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y2p1[7] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y2p2[8] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y3m3[9] - 1.0), 0.0, places=4)
self.assertAlmostEqual(jnp.abs(res_Y3p3[15] - 1.0), 0.0, places=4)
def test_get_correlations_lm_IJ(self):
"""
Test the get_correlations_lm_IJ function
"""
l_max = 6
nside = 32
npixels = 35
npix = hp.nside2npix(nside)
pixels = np.linspace(0, npix - 1, npixels, dtype=int)
p_I = jnp.array(hp.pix2vec(nside, pixels))
GammalmIJ = gt.get_correlations_lm_IJ(p_I.T, l_max, nside)
data = np.load(tu.get_correlations_lm_IJ_data_path)["data"]
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(data - GammalmIJ))), 0.0, places=5
)
def test_get_tensors_generation(self):
"""
Test the get_tensors function
"""
npulsars = 30
HD_order = 0
result, test_shapes = get_tensors_and_shapes(
tu.test_catalog_path2, npulsars, "legendre", HD_order
)
for i in range(len(tu.get_tensor_labels)):
self.assertTupleEqual(result[i].shape, test_shapes[i])
def test_get_tensors_generation_Legendre(self):
"""
Test the get_tensors function assuming you want the Legendre projection
"""
npulsars = 50
HD_basis = "legendre"
HD_order = 6
result, test_shapes = get_tensors_and_shapes(
tu.test_catalog_path2, npulsars, HD_basis, HD_order
)
for i in range(len(tu.get_tensor_labels)):
self.assertTupleEqual(result[i].shape, test_shapes[i])
def test_get_tensors_generation_Binned(self):
"""
Test the get_tensors function assuming you want the Binned projection
"""
npulsars = 30
HD_basis = "legendre"
HD_order = 10
result, test_shapes = get_tensors_and_shapes(
tu.test_catalog_path2, npulsars, HD_basis, HD_order
)
for i in range(len(tu.get_tensor_labels)):
self.assertTupleEqual(result[i].shape, test_shapes[i])
def test_get_tensors_results(self):
"""
Test the get_tensors function results
"""
HD_basis = "legendre"
HD_order = 0
results, loaded_data = get_tensors_data(
tu.test_catalog_path,
HD_basis,
HD_order,
tu.get_tensors_data_path,
)
for i in range(len(tu.get_tensor_labels)):
to_assert = jnp.sum(
jnp.abs(results[i] - loaded_data[tu.get_tensor_labels[i]])
)
self.assertAlmostEqual(float(to_assert), 0.0)
def test_Legendre_projection(self):
"""
Test the results of the Legendre projection of the get_tensors function (going to large HD_order the projection gives chi_IJ)
"""
HD_order = 30
n_pulsars = 70
vec = np.random.uniform(-1.0, 1.0, n_pulsars)
matrix = vec[:, None] * vec[None, :]
for i in range(n_pulsars):
matrix[i, i] = 1.0
chi_IJ = gt.get_chi_tensor_IJ(matrix) - 0.5 * np.eye(n_pulsars)
HD_coefficients = gt.get_HD_Legendre_coefficients(HD_order)
polynomials = gt.get_polynomials_IJ(matrix, HD_order)
HD_val = HD_coefficients[:, None, None] * polynomials
diff = 1.0 - np.sum(HD_val, axis=0) / chi_IJ
self.assertAlmostEqual(float(jnp.mean(jnp.abs(diff))), 0.0, places=3)
def test_get_tensors_Binned_results(self):
"""
Test the results of the Binned projection of the get_tensors function
"""
HD_basis = "binned"
HD_order = 6
results, loaded_data = get_tensors_data(
tu.test_catalog_path,
HD_basis,
HD_order,
tu.get_tensors_Binned_data_path,
)
for i in range(len(tu.get_tensor_labels)):
to_assert = jnp.sum(
jnp.abs(results[i] - loaded_data[tu.get_tensor_labels[i]])
)
self.assertAlmostEqual(float(to_assert), 0.0)
def test_get_tensors_Legendre_results(self):
"""
Another test of the Legendre projection of the get_tensors function
"""
HD_basis = "legendre"
HD_order = 6
results, loaded_data = get_tensors_data(
tu.test_catalog_path,
HD_basis,
HD_order,
tu.get_tensors_Legendre_data_path,
)
for i in range(len(tu.get_tensor_labels)):
to_assert = jnp.sum(
jnp.abs(results[i] - loaded_data[tu.get_tensor_labels[i]])
)
self.assertAlmostEqual(float(to_assert), 0.0)
def test_get_tensors_anisotropies(self):
"""
Test the results of the get_tensors function with anisotropies, check that the monopole is correct
"""
HD_basis = "legendre"
HD_order = 6
results, loaded_data = get_tensors_data(
tu.test_catalog_path,
HD_basis,
HD_order,
tu.get_tensors_Legendre_data_path,
anisotropies=True,
)
to_assert = jnp.mean(
jnp.abs(
results[1][0] / np.sqrt(4 * np.pi) - loaded_data["response_IJ"]
)
)
self.assertAlmostEqual(float(to_assert), 0.0, places=3)
if __name__ == "__main__":
unittest.main(verbosity=2)
| 10,455 | Python | .py | 260 | 31.311538 | 133 | 0.596894 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,347 | test_compute_Fisher.py | Mauropieroni_fastPTA/tests/test_compute_Fisher.py | # Global
import unittest
import numpy as np
import jax
import jax.numpy as jnp
# Local
import utils as tu
import fastPTA.utils as ut
from fastPTA import Fisher_code as Fc
jax.config.update("jax_enable_x64", True)
jax.config.update("jax_default_device", jax.devices(ut.which_device)[0])
n_frequencies = 35
data = np.load(tu.Fisher_data_path)
class TestFisher(unittest.TestCase):
def test_compute_SNR_integrand(self):
"""
Test the function that computes the integrand of the SNR
"""
# Generate 3 random matrixes 5 x 5
data = np.random.uniform(0.0, 1.0, size=(3, 5, 5))
# Symmetrize
data += np.moveaxis(data, -1, -2)
inverse = np.linalg.inv(data)
result = Fc.get_SNR_integrand(data, inverse)
to_assert = jnp.sum(jnp.abs(result - 5 * np.ones(3)))
self.assertAlmostEqual(float(to_assert), 0.0, places=10)
def test_get_fisher_integrand(self):
"""
Test the function that computes the integrand of the Fisher matrix
"""
# Generate 3 random matrixes 5 x 5
data = np.random.uniform(0.0, 1.0, size=(2, 5, 5))
# Symmetrize
data += np.moveaxis(data, -1, -2)
inverse = np.linalg.inv(data)
data = np.array([data, data, data])
result = Fc.get_fisher_integrand(data, inverse)
to_assert = jnp.sum(jnp.abs(result - 5 * np.ones(shape=(3, 3, 2))))
self.assertAlmostEqual(float(to_assert), 0.0, places=10)
def test_compute_fisher(self):
"""
Test the function that computes the Fisher matrix
"""
get_tensors_kwargs = {
"path_to_pulsar_catalog": tu.test_catalog_path,
"verbose": True,
}
res = Fc.compute_fisher(
n_frequencies=n_frequencies, get_tensors_kwargs=get_tensors_kwargs
)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(res[0] - data["frequency"]))), 0.0, places=7
)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(res[1] - data["signal"]))), 0.0, places=7
)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(res[4] - data["effective_noise"]))), 0.0
)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(res[5] - data["snr"]))), 0.0, places=7
)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(res[6] - data["fisher"]))), 0.0, places=7
)
def test_compute_fisher_legendre(self):
"""
Test the function that computes the Fisher matrix with HD projected onto Legendre polynomials
"""
get_tensors_kwargs = {
"path_to_pulsar_catalog": tu.test_catalog_path,
"HD_basis": "legendre",
"HD_order": 6,
"verbose": True,
}
HD_legendre = Fc.compute_fisher(get_tensors_kwargs=get_tensors_kwargs)
self.assertAlmostEqual(
float(
jnp.sum(jnp.abs(HD_legendre[6] - data["fisher_HD_legendre"]))
),
0.0,
places=8,
)
def test_compute_fisher_binned(self):
"""
Test the function that computes the Fisher matrix with HD projected onto the binned basis
"""
get_tensors_kwargs = {
"path_to_pulsar_catalog": tu.test_catalog_path,
"HD_basis": "binned",
"HD_order": 10,
"verbose": True,
}
HD_binned = Fc.compute_fisher(get_tensors_kwargs=get_tensors_kwargs)
self.assertAlmostEqual(
float(jnp.sum(jnp.abs(HD_binned[6] - data["fisher_HD_binned"]))),
0.0,
places=7,
)
def test_compute_fisher_anisotropic(self):
"""
TO ADD !!!
"""
if __name__ == "__main__":
unittest.main(verbosity=2)
| 3,851 | Python | .py | 105 | 27.733333 | 101 | 0.583423 | Mauropieroni/fastPTA | 8 | 0 | 3 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,348 | gen_widget_doc.py | zimolab_PyGUIAdapter/tools/gen_widget_doc.py | from __future__ import annotations
import dataclasses
import inspect
import os.path
from typing import Type, Tuple, List, Literal
import jinja2
from markdown_table_generator import generate_markdown, table_from_string_list
import pyguiadapter.utils.messagebox
from pyguiadapter import widgets, utils
from pyguiadapter.adapter import GUIAdapter, uoutput, udialog
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.extend_types import file_t
from pyguiadapter.paramwidget import BaseParameterWidget
from pyguiadapter.widgets.common import CommonParameterWidgetConfig
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
# warnings.filterwarnings("error")
CONFIG_CLASS_TABLE_HEADER_CN = ("配置项名称", "类型", "默认值", "说明")
def _all_widget_classes() -> List[str]:
all_imported = dir(widgets)
ret = []
for name in all_imported:
obj = getattr(widgets, name, None)
if not obj:
continue
if inspect.isclass(obj) and issubclass(obj, BaseParameterWidget):
ret.append(obj.__name__)
return ret
def _get_default_value(field: dataclasses.Field) -> str:
if field.default is not dataclasses.MISSING:
return str(field.default)
if field.default_factory is not dataclasses.MISSING:
return str(field.default_factory)
return ""
def dataclass2table(
clazz: Type,
header: Tuple[str, str, str, str] | None = None,
exclude_props: List[str] | None = None,
):
assert dataclasses.is_dataclass(clazz)
if exclude_props is None:
exclude_props = []
if not header:
header = CONFIG_CLASS_TABLE_HEADER_CN
header = header
assert isinstance(header, tuple) and len(header) >= 4
props = [list(header)]
fields = dataclasses.fields(clazz)
for field in fields:
if field.name in exclude_props:
continue
prop_name = f"`{field.name}`"
prop_type = "`{}`".format(str(field.type).replace("|", "\|"))
prop_default = _get_default_value(field)
prop_desc = ""
props.append([prop_name, prop_type, f"`{prop_default}`", prop_desc])
table = table_from_string_list(props)
return generate_markdown(table)
def generate_widget_class_doc(
widget_type: str,
widget_category: Literal["basic", "extend"] = "basic",
template_file: file_t = "",
output_file: file_t = "",
headers: Tuple[str, str, str, str] = CONFIG_CLASS_TABLE_HEADER_CN,
exclude_props: List = None,
):
"""
## Document Generator
This tool is mainly used for generating the document page of the selected widget class.
User must provide a document template file of jinja2.
"""
if exclude_props is None:
exclude_props = []
if not widget_type:
raise ParameterError("widget_type", "widget_type cannot be empty")
if not widget_category:
raise ParameterError("widget_category", "widget_category cannot be empty")
if not template_file:
raise ParameterError("template_file", "template_file cannot be empty")
if not output_file:
raise ParameterError("output_file", "output_file cannot be empty")
if os.path.isfile(output_file):
uoutput.warning(f"Output file {output_file} already exists! Overwrite?")
ret = udialog.show_question_messagebox(
f"Output file {output_file} already exists! Do you want to overwrite it?"
)
if ret != pyguiadapter.utils.messagebox.StandardButton.Yes:
uoutput.warning(" |")
uoutput.warning(" - User cancelled")
uoutput.info("==Generating Finished!==")
return
uoutput.info(" |")
uoutput.info(" - Overwrite Approved")
widget_class = getattr(widgets, widget_type, None)
if not (
inspect.isclass(widget_class) and issubclass(widget_class, BaseParameterWidget)
):
raise ValueError(f"cannot find widget class for {widget_type}")
widget_class_name = widget_class.__name__
uoutput.info(f"Generating config doc for {widget_class_name}...")
uoutput.info(f"Widget Class: {widget_class_name}")
widget_class_filename = utils.get_object_filename(widget_class)
if widget_class_filename:
widget_class_filename = os.path.basename(widget_class_filename)
uoutput.info(" |")
uoutput.info(f" -Filename: {widget_class_filename}")
else:
widget_class_filename = "<unknown.py>"
uoutput.warning(" |")
uoutput.warning(f" -Filename: {widget_class_filename}")
widget_config_class = widget_class.ConfigClass
widget_config_class_name = widget_config_class.__name__
uoutput.info(f"Config Class: {widget_config_class_name}")
widget_config_class_filename = utils.get_object_filename(widget_config_class)
if widget_config_class_filename:
widget_config_class_filename = os.path.basename(widget_config_class_filename)
uoutput.info(" |")
uoutput.info(f" -Filename: {widget_config_class_filename}")
else:
widget_config_class_filename = "<unknown.py>"
uoutput.warning(f" |")
uoutput.warning(f" -Filename: {widget_config_class_filename}")
widget_config_props_table = dataclass2table(
widget_config_class, headers, exclude_props
)
widget_config_class_source = utils.get_object_sourcecode(widget_config_class)
if widget_config_class_source:
widget_config_class_source = (
"@dataclasses.dataclass(frozen=True)\n" + widget_config_class_source
)
else:
widget_config_class_source = "UNKNOWN"
try:
template_content = utils.read_text_file(template_file)
except Exception as e:
uoutput.critical("Error:")
uoutput.critical(f" {e}")
uoutput.info("==Generating Finished!==")
raise e
else:
uoutput.info("Template File Loaded")
uoutput.info(" |")
uoutput.info(f" - {os.path.normpath(template_file)}")
template = jinja2.Template(template_content)
try:
output_content = template.render(
widget_class_name=widget_class_name,
widget_category=widget_category,
widget_class_filename=widget_class_filename,
widget_config_class_name=widget_config_class_name,
widget_config_class_filename=widget_config_class_filename,
widget_config_class_source=widget_config_class_source,
widget_config_props_table=widget_config_props_table,
)
utils.write_text_file(output_file, output_content)
except Exception as e:
uoutput.critical("Error:")
uoutput.critical(f" {e}")
uoutput.info("==Generating Finished!==")
raise e
else:
uoutput.info("Document File Generated")
uoutput.info(" |")
uoutput.info(f" - {os.path.normpath(output_file)}")
uoutput.info("==Generating Finished!==")
udialog.show_info_messagebox(
f"Document file for widget '{widget_class_name}' generated successfully!",
"Success",
)
if __name__ == "__main__":
########################## parameter configs ################################################
common_props = (
field.name for field in dataclasses.fields(CommonParameterWidgetConfig)
)
all_props = list(common_props)
default_exclude_props = [prop for prop in all_props if prop != "default_value"]
exclude_props_conf = widgets.MultiChoiceBoxConfig(
default_value=default_exclude_props, choices=all_props
)
headers_conf = widgets.TupleEditConfig(
default_value=CONFIG_CLASS_TABLE_HEADER_CN,
)
template_file_conf = widgets.FileSelectConfig(
filters="Jinja2 files(*.j2);;Markdown files(*.md);;Text files(*.txt);;All files(*.*)",
)
output_file_conf = widgets.FileSelectConfig(
filters="Markdown files(*.md);;Text files(*.txt);;All files(*.*)",
save_file=True,
)
widget_type_conf = widgets.ChoiceBoxConfig(choices=_all_widget_classes())
########################## parameter configs ################################################
########################## window configs ################################################
win_conf = FnExecuteWindowConfig(
show_function_result=False, print_function_result=False
)
########################## window configs ################################################
adapter = GUIAdapter()
adapter.add(
generate_widget_class_doc,
widget_configs={
"widget_type": widget_type_conf,
"template_file": template_file_conf,
"output_file": output_file_conf,
"headers": headers_conf,
"exclude_props": exclude_props_conf,
},
window_config=win_conf,
)
adapter.run()
| 8,854 | Python | .py | 208 | 35.413462 | 97 | 0.645304 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,349 | make_table.py | zimolab_PyGUIAdapter/tools/make_table.py | from tools.gen_widget_doc import dataclass2table
from pyguiadapter.windows.fnselect import FnSelectWindowConfig
from pyguiadapter.windows import DocumentBrowserConfig
if __name__ == "__main__":
# print(dataclass2table(FnSelectWindowConfig))
print(dataclass2table(DocumentBrowserConfig))
| 297 | Python | .py | 6 | 46.833333 | 62 | 0.83045 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,350 | multiple_function_example.py | zimolab_PyGUIAdapter/examples/multiple_function_example.py | from pyguiadapter.adapter import GUIAdapter
def function_1(arg: int):
"""
description of function_1
"""
pass
def function_2(arg: int):
"""
description of function_2
"""
pass
def function_3(arg: int):
"""
description of function_3
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(function_1)
adapter.add(function_2)
adapter.add(function_3)
adapter.run()
| 451 | Python | .py | 22 | 16.045455 | 43 | 0.629454 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,351 | anchors_in_document.py | zimolab_PyGUIAdapter/examples/anchors_in_document.py | import os
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.utils import read_text_file
from pyguiadapter.windows import DocumentBrowserConfig
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
DOCUMENT_PATH = os.path.join(os.path.dirname(__file__), "anchors_in_document.md")
def anchors_in_document(
a: str,
b: str,
c: str,
d: str,
e: str,
f: str,
g: str,
h: str,
i: str,
j: str,
k: str,
l: str,
m: str,
n: str,
o: str,
p: str,
):
"""
This is an example demonstrating how to use parameter anchors and group anchors in the document of function.
Args:
a: description of parameter a.
b: description of parameter b.
c: description of parameter c.
d: description of parameter d.
e: description of parameter e.
f: description of parameter f.
g: description of parameter g.
h: description of parameter h.
i: description of parameter i.
j: description of parameter j.
k: description of parameter k.
l: description of parameter l.
m: description of parameter m.
n: description of parameter n.
o: description of parameter o.
p: description of parameter p.
"""
pass
if __name__ == "__main__":
widget_configs = {
# parameters in Group-A
"a": {"group": "Group-A"},
"b": {"group": "Group-A"},
"c": {"group": "Group-A"},
"d": {"group": "Group-A"},
# parameters in Group-B
"e": {"group": "Group-B"},
"f": {"group": "Group-B"},
"g": {"group": "Group-B"},
"h": {"group": "Group-B"},
# parameters in Group-C
"i": {"group": "Group-C"},
"j": {"group": "Group-C"},
"k": {"group": "Group-C"},
"l": {"group": "Group-C"},
# parameters in default group
"m": {"group": None},
"n": {"group": None},
"o": {"group": None},
"p": {"group": None},
}
document = read_text_file(DOCUMENT_PATH)
adapter = GUIAdapter()
adapter.add(
anchors_in_document,
document=document,
document_format="markdown",
widget_configs=widget_configs,
window_config=FnExecuteWindowConfig(
document_browser_config=DocumentBrowserConfig(
parameter_anchor=True, group_anchor=True
)
),
)
adapter.run()
| 2,460 | Python | .py | 82 | 22.939024 | 112 | 0.574082 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,352 | qt_material_example.py | zimolab_PyGUIAdapter/examples/qt_material_example.py | from datetime import datetime
from qtpy.QtWidgets import QApplication
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t, json_obj_t
def app_style_example(
arg1: str,
arg2: int,
arg3: float,
arg4: bool,
arg5: text_t,
arg6: datetime,
arg7: json_obj_t,
):
"""
This example requires [Qt-Material](https://github.com/UN-GCPDS/qt-material).
Please install it before you run the example.
<br />
e.g. using `pip`:
> `pip install qt-material`
@param arg1: arg1 description
@param arg2: arg2 description
@param arg3: arg3 description
@param arg4: arg4 description
@param arg5: arg5 description
@param arg6: arg6 description
@param arg7: arg7 description
@return:
@params
[arg6]
calendar_popup = true
@end
"""
pass
if __name__ == "__main__":
import qt_material
def on_app_start(app: QApplication):
# this will be called after the instantiation of QApplication.
print("app started")
qt_material.apply_stylesheet(app, theme="light_teal.xml")
adapter = GUIAdapter(on_app_start=on_app_start)
adapter.add(app_style_example)
adapter.run()
| 1,228 | Python | .py | 42 | 24.261905 | 81 | 0.684885 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,353 | show_select_win_anyway.py | zimolab_PyGUIAdapter/examples/show_select_win_anyway.py | from pyguiadapter.adapter import GUIAdapter
def show_select_win_anyway():
"""
This example shows how to show function select window(FnSelectWindow) even if only one function was added.
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(show_select_win_anyway)
adapter.run(show_select_window=True)
| 349 | Python | .py | 10 | 30.7 | 110 | 0.713433 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,354 | tiny_window_example.py | zimolab_PyGUIAdapter/examples/tiny_window_example.py | from typing import Optional
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def equation_solver(a: float, b: float, c: float) -> Optional[tuple]:
"""
Solving Equations: ax^2 + bx + c = 0 (a,b,c ∈ R, a ≠ 0)
@param a: a ∈ R, a ≠ 0
@param b: b ∈ R
@param c: c ∈ R
@return:
"""
if a == 0:
raise ParameterError(parameter_name="a", message="a cannot be zero!")
delta = b**2 - 4 * a * c
if delta < 0:
return None
x1 = (-b + delta**0.5) / (2 * a)
if delta == 0:
return x1, x1
x2 = (-b - delta**0.5) / (2 * a)
return x1, x2
if __name__ == "__main__":
window_config = FnExecuteWindowConfig(
title="Equation Solver",
icon="mdi6.function-variant",
execute_button_text="Solve",
size=(350, 450),
document_dock_visible=False,
output_dock_visible=False,
clear_button_visible=False,
clear_checkbox_visible=False,
show_function_result=True,
function_result_message="real roots: {}",
default_parameter_group_name="Equation Parameters",
print_function_error=False,
print_function_result=False,
)
adapter = GUIAdapter()
adapter.add(equation_solver, window_config=window_config)
adapter.run()
| 1,402 | Python | .py | 41 | 27.780488 | 77 | 0.624721 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,355 | toast_example.py | zimolab_PyGUIAdapter/examples/toast_example.py | from pyguiadapter.action import Action
from pyguiadapter.adapter import GUIAdapter, utoast
from pyguiadapter.extend_types import text_t
from pyguiadapter.toast import ToastConfig
from pyguiadapter.toolbar import ToolBar
from pyguiadapter.windows.fnexec import FnExecuteWindow
def show_toast(win: FnExecuteWindow, action: Action):
win.show_toast("Paused!", duration=3000, clear=True)
def clear_toasts(win: FnExecuteWindow, action: Action):
win.clear_toasts()
action_toast = Action(
text="Toast",
icon="mdi.message-text-clock",
tooltip="Show toast message",
on_triggered=show_toast,
)
action_clear_toasts = Action(
text="Clear toasts",
icon="ei.remove-circle",
tooltip="Clear all toasts",
on_triggered=clear_toasts,
)
def toast_example(
message: text_t = "Hello world!",
duration: int = 3000,
clear: bool = False,
fade_out: int = 0,
):
if not message:
return
fade_out = max(fade_out, 0)
utoast.show_toast(message, duration, ToastConfig(fade_out=fade_out), clear)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(
toast_example,
window_toolbar=ToolBar(actions=[action_toast, action_clear_toasts]),
)
adapter.run()
| 1,243 | Python | .py | 39 | 27.820513 | 79 | 0.719195 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,356 | menu_example_1.py | zimolab_PyGUIAdapter/examples/menu_example_1.py | import json
from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.utils import filedialog, inputdialog, messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
def menu_example():
pass
###################Action Callbacks#########################
def on_action_open(window: FnExecuteWindow, _: Action):
print("on_action_open()")
ret = filedialog.get_open_file(
parent=window,
title="Open File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be opened: {ret}")
def on_action_save(window: FnExecuteWindow, _: Action):
print("on_action_save()")
ret = filedialog.get_save_file(
parent=window,
title="Save File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be saved to: {ret}")
def on_action_settings(window: FnExecuteWindow, _: Action):
default_settings = {
"opt1": 1,
"opt2": "2",
"opt3": True,
}
new_settings = inputdialog.input_json_object(
parent=window,
title="Settings",
icon="fa.cog",
size=(600, 400),
ok_button_text="Save",
cancel_button_text="Cancel",
initial_text=json.dumps(default_settings, indent=4, ensure_ascii=False),
auto_indent=True,
indent_size=4,
auto_parentheses=True,
line_wrap_mode=inputdialog.LineWrapMode.WidgetWidth,
line_wrap_width=88,
)
if isinstance(new_settings, dict):
messagebox.show_info_message(window, f"new settings: {new_settings}")
def on_action_confirm_quit(window: FnExecuteWindow, action: Action, checked: bool):
print("on_action_confirm_close(): ", checked)
def on_action_close(window: FnExecuteWindow, _: Action):
print("on_action_close()")
window.close()
def on_action_about(window: FnExecuteWindow, _: Action):
print("on_action_about()")
about_text = """
<h1>PyGUIAdapter V2</h1>
<p>PyGUIAdapter is a GUI lib for those who want make GUI application without writing GUI code!</p>
<p>
You can access the source code <a href="https://github.com/zimolab/PyGUIAdapter">here</a>!
</p>
"""
messagebox.show_text_content(
window,
text_content=about_text,
text_format="html",
title="About PyGUIAdapter",
icon="fa.info-circle",
)
def on_action_license(window: FnExecuteWindow, _: Action):
print("on_action_license()")
license_file = "../../LICENSE"
messagebox.show_text_file(
window,
text_file=license_file,
text_format="plaintext",
title="License",
icon="fa.copyright",
)
###################~Action Callbacks#########################
if __name__ == "__main__":
###################Actions#############################
action_open = Action(
text="Open",
icon="fa.folder-open",
on_triggered=on_action_open,
shortcut="Ctrl+O",
)
action_save = Action(
text="Save",
icon="fa.save",
on_triggered=on_action_save,
shortcut="Ctrl+S",
)
action_settings = Action(
text="Settings",
icon="fa.cog",
on_triggered=on_action_settings,
shortcut="Ctrl+,",
)
action_quit = Action(
text="Quit",
icon="fa.power-off",
on_triggered=on_action_close,
shortcut="Ctrl+Q",
)
action_confirm_quit = Action(
text="Confirm Quit",
icon="fa.question-circle",
checkable=True,
checked=False,
on_toggled=on_action_confirm_quit,
)
action_about = Action(
text="About",
icon="fa.info-circle",
on_triggered=on_action_about,
)
action_license = Action(
text="License",
icon="fa.copyright",
on_triggered=on_action_license,
)
###################~Actions#############################
####################Menus#############################
submenu_help = Menu(
title="Help",
actions=[action_about, action_license],
)
menu_file = Menu(
title="File",
actions=[
action_open,
action_save,
Separator(),
submenu_help,
Separator(),
action_quit,
],
)
menu_settings = Menu(
title="Settings",
actions=[action_settings, Separator(), action_confirm_quit],
)
menus = [menu_file, menu_settings]
###################~Menus#############################
################Window Event Listener###################
def on_window_create(window: FnExecuteWindow):
print("on_window_create()")
# make action_confirm_quit checked after the select window is created
window.set_action_state(action_confirm_quit, True)
def on_window_close(window: FnExecuteWindow) -> bool:
# get the state of action_confirm_quit
# if it is checked, show a question message box to ask if the user really wants to close the window
# if it is not checked, return True to close the window directly.
state = window.get_action_state(action_confirm_quit)
if state:
# access the
ret = messagebox.show_question_message(
window,
message="Do you really want to close the window?",
title="Quit",
buttons=messagebox.Yes | messagebox.No,
)
return ret == messagebox.Yes
return True
window_listener = SimpleWindowEventListener(
on_create=on_window_create, on_close=on_window_close
)
#################Window Event Listener##################
adapter = GUIAdapter()
adapter.add(menu_example, window_menus=menus, window_listener=window_listener)
adapter.run()
| 6,106 | Python | .py | 177 | 27.180791 | 107 | 0.585692 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,357 | global_style_example.py | zimolab_PyGUIAdapter/examples/global_style_example.py | import os.path
from datetime import datetime
from pyguiadapter import utils
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t
def app_style_example(
arg1: str, arg2: int, arg3: float, arg4: bool, arg5: text_t, arg6: datetime
):
"""
This example shows how to apply a global stylesheet to the application.
First, read the stylesheet from a qss file. In this example, [Ubuntu.qss](https://github.com/GTRONICK/QSS/blob/master/Ubuntu.qss)
will be used. This qss file is from GTRONICK's [QSS](https://github.com/GTRONICK/QSS) repo.
Then, pass the qss content to the GUIAdapter constructor's **global_style** argument.
"""
pass
if __name__ == "__main__":
QSS_FILE = os.path.join(os.path.dirname(__file__), "Ubuntu.qss")
global_stylesheet = utils.read_text_file(QSS_FILE)
adapter = GUIAdapter(global_stylesheet=global_stylesheet)
adapter.add(app_style_example)
adapter.run()
| 968 | Python | .py | 21 | 42.142857 | 133 | 0.733191 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,358 | menu_example.py | zimolab_PyGUIAdapter/examples/menu_example.py | import json
from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.utils import filedialog, inputdialog, messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnselect import FnSelectWindow
def menu_example():
pass
############Action Callbacks##########
def on_action_open(window: FnSelectWindow, _: Action):
print("on_action_open()")
ret = filedialog.get_open_file(
parent=window,
title="Open File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be opened: {ret}")
def on_action_save(window: FnSelectWindow, _: Action):
print("on_action_save()")
ret = filedialog.get_save_file(
parent=window,
title="Save File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be saved to: {ret}")
def on_action_settings(window: FnSelectWindow, _: Action):
default_settings = {
"opt1": 1,
"opt2": "2",
"opt3": True,
}
new_settings = inputdialog.input_json_object(
parent=window,
title="Settings",
icon="fa.cog",
size=(600, 400),
ok_button_text="Save",
cancel_button_text="Cancel",
initial_text=json.dumps(default_settings, indent=4, ensure_ascii=False),
auto_indent=True,
indent_size=4,
auto_parentheses=True,
line_wrap_mode=inputdialog.LineWrapMode.WidgetWidth,
line_wrap_width=88,
)
if isinstance(new_settings, dict):
messagebox.show_info_message(window, f"new settings: {new_settings}")
def on_action_confirm_quit(window: FnSelectWindow, action: Action, checked: bool):
print("on_action_confirm_close(): ", checked)
def on_action_close(window: FnSelectWindow, _: Action):
print("on_action_close()")
window.close()
def on_action_about(window: FnSelectWindow, _: Action):
print("on_action_about()")
###############################
if __name__ == "__main__":
action_open = Action(
text="Open",
icon="fa.folder-open",
on_triggered=on_action_open,
shortcut="Ctrl+O",
)
action_save = Action(
text="Save",
icon="fa.save",
on_triggered=on_action_save,
shortcut="Ctrl+S",
)
action_settings = Action(
text="Settings",
icon="fa.cog",
on_triggered=on_action_settings,
shortcut="Ctrl+,",
)
action_quit = Action(
text="Quit",
icon="fa.power-off",
on_triggered=on_action_close,
shortcut="Ctrl+Q",
)
action_confirm_quit = Action(
text="Confirm Quit",
checkable=True,
checked=False,
on_toggled=on_action_confirm_quit,
)
def on_window_create(window: FnSelectWindow):
print("on_window_create()")
# make action_confirm_quit checked after the select window is created
window.set_action_state(action_confirm_quit, True)
def on_window_close(window: FnSelectWindow) -> bool:
# get the state of action_confirm_quit
# if it is checked, show a question message box to ask if the user really wants to close the window
# if it is not checked, return True to close the window directly.
state = window.get_action_state(action_confirm_quit)
if state:
# access the
ret = messagebox.show_question_message(
window,
message="Do you really want to close the window?",
title="Quit",
buttons=messagebox.Yes | messagebox.No,
)
return ret == messagebox.Yes
return True
window_listener = SimpleWindowEventListener(
on_create=on_window_create, on_close=on_window_close
)
menus = [
Menu(
title="File",
actions=[
action_open,
action_save,
Separator(),
action_confirm_quit,
action_quit,
],
),
Menu(title="Help", actions=[action_settings]),
]
adapter = GUIAdapter()
adapter.add(menu_example, window_menus=menus, window_listener=window_listener)
adapter.run()
| 4,454 | Python | .py | 129 | 26.751938 | 107 | 0.612468 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,359 | hello_world.py | zimolab_PyGUIAdapter/examples/hello_world.py | from typing import Literal
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
def temperature_converter(
temperature: float,
from_unit: Literal["°C", "°F"] = "°F",
to_unit: Literal["°C", "°F"] = "°C",
) -> float:
"""
Convert temperature from one unit to another.
---
*This is a simple demo of [`PyGUIAdapter`](https://github.com/zimolab/PyGUIAdapter).*
Args:
temperature (flot): current value of temperature
from_unit (str): current unit of temperature
to_unit (str): target unit of temperature
Returns:
the converted value of temperature
"""
if from_unit == "°C" and to_unit == "°F":
converted = temperature * 9 / 5 + 32
elif from_unit == "°F" and to_unit == "°C":
converted = (temperature - 32) * 5 / 9
else:
converted = temperature
uprint(f"{temperature}{from_unit} = {converted}{to_unit}")
return converted
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(temperature_converter)
adapter.run()
| 1,097 | Python | .py | 31 | 29.709677 | 89 | 0.639657 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,360 | qdarktheme_example.py | zimolab_PyGUIAdapter/examples/qdarktheme_example.py | """
This example requires PyQtDarkTheme. Please install it before you run this example.
"""
from datetime import datetime
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t
def app_style_example(
arg1: str, arg2: int, arg3: float, arg4: bool, arg5: text_t, arg6: datetime
):
"""
This example requires [PyQtDarkTheme](https://github.com/5yutan5/PyQtDarkTheme).
Please install it before you run the example.
<br />
e.g. using `pip`:
> `pip install pyqtdarktheme`
@param arg1: arg1 description
@param arg2: arg2 description
@param arg3: arg3 description
@param arg4: arg4 description
@param arg5: arg5 description
@param arg6: arg6 description
@return:
"""
pass
if __name__ == "__main__":
import qdarktheme
def on_app_start(app):
# this will be called after the instantiation of QApplication.
print("app started")
qdarktheme.setup_theme("dark")
adapter = GUIAdapter(on_app_start=on_app_start)
adapter.add(app_style_example)
adapter.run()
| 1,093 | Python | .py | 33 | 28.484848 | 84 | 0.705153 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,361 | fn_select_window_menu_example.py | zimolab_PyGUIAdapter/examples/fn_select_window_menu_example.py | from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.utils import messagebox
from pyguiadapter.windows.fnselect import FnSelectWindow
def on_action_test(window: FnSelectWindow, action: Action):
messagebox.show_info_message(
window, message=f"Action Triggered!(Action: {action.text})"
)
def on_action_close(window: FnSelectWindow, _: Action):
ret = messagebox.show_question_message(
window,
message="Are you sure to close the application?",
buttons=messagebox.Yes | messagebox.No,
)
if ret == messagebox.Yes:
window.close()
action_test = Action(
text="Test", icon="fa.folder-open", on_triggered=on_action_test, shortcut="Ctrl+O"
)
action_close = Action(
text="Close", icon="fa.close", on_triggered=on_action_close, shortcut="Ctrl+Q"
)
menu_file = Menu(
title="File",
actions=[action_test, Separator(), action_close],
)
def foo():
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(foo)
adapter.run(show_select_window=True, select_window_menus=[menu_file])
| 1,169 | Python | .py | 33 | 31.272727 | 86 | 0.719751 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,362 | window_event_example.py | zimolab_PyGUIAdapter/examples/window_event_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.window import BaseWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
from pyguiadapter.utils import messagebox
def event_example_1():
pass
class ExampleEventListener(BaseWindowEventListener):
def on_create(self, window: FnExecuteWindow):
print("on_create")
super().on_create(window)
def on_close(self, window: FnExecuteWindow) -> bool:
print("on_close")
ret = messagebox.show_question_message(
window,
title="Confirm Quit",
message="Are you sure to quit?",
buttons=messagebox.Yes | messagebox.No,
)
if ret == messagebox.Yes:
return True
return False
def on_destroy(self, window: FnExecuteWindow):
print("on_destroy")
super().on_destroy(window)
def on_hide(self, window: FnExecuteWindow):
print("on_hide")
super().on_hide(window)
def on_show(self, window: FnExecuteWindow):
print("on_show")
super().on_show(window)
if __name__ == "__main__":
event_listener = ExampleEventListener()
adapter = GUIAdapter()
adapter.add(event_example_1, window_listener=event_listener)
adapter.run()
| 1,281 | Python | .py | 35 | 29.457143 | 64 | 0.669636 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,363 | toolbar_example.py | zimolab_PyGUIAdapter/examples/toolbar_example.py | import json
from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.toolbar import (
ToolBar,
RightToolBarArea,
LeftToolBarArea,
ToolButtonTextUnderIcon,
TopToolBarArea,
)
from pyguiadapter.utils import filedialog, inputdialog, messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
def toolbar_example():
pass
###################Action Callbacks#########################
def on_action_open(window: FnExecuteWindow, _: Action):
print("on_action_open()")
ret = filedialog.get_open_file(
parent=window,
title="Open File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be opened: {ret}")
def on_action_save(window: FnExecuteWindow, _: Action):
print("on_action_save()")
ret = filedialog.get_save_file(
parent=window,
title="Save File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be saved to: {ret}")
def on_action_settings(window: FnExecuteWindow, _: Action):
default_settings = {
"opt1": 1,
"opt2": "2",
"opt3": True,
}
new_settings = inputdialog.input_json_object(
parent=window,
title="Settings",
icon="fa.cog",
size=(600, 400),
ok_button_text="Save",
cancel_button_text="Cancel",
initial_text=json.dumps(default_settings, indent=4, ensure_ascii=False),
auto_indent=True,
indent_size=4,
auto_parentheses=True,
line_wrap_mode=inputdialog.LineWrapMode.WidgetWidth,
line_wrap_width=88,
)
if isinstance(new_settings, dict):
messagebox.show_info_message(window, f"new settings: {new_settings}")
def on_action_confirm_quit(window: FnExecuteWindow, _: Action, checked: bool):
print("on_action_confirm_close(): ", checked)
def on_action_close(window: FnExecuteWindow, _: Action):
print("on_action_close()")
window.close()
###################~Action Callbacks#########################
if __name__ == "__main__":
###################Actions#############################
action_open = Action(
text="Open",
icon="fa.folder-open",
on_triggered=on_action_open,
shortcut="Ctrl+O",
)
action_save = Action(
text="Save",
icon="fa.save",
on_triggered=on_action_save,
shortcut="Ctrl+S",
)
action_settings = Action(
text="Settings",
icon="fa.cog",
on_triggered=on_action_settings,
shortcut="Ctrl+,",
)
action_quit = Action(
text="Quit",
icon="fa.power-off",
on_triggered=on_action_close,
shortcut="Ctrl+Q",
)
action_confirm_quit = Action(
text="Confirm Quit",
icon="fa.question-circle",
checkable=True,
checked=False,
on_toggled=on_action_confirm_quit,
)
###################~Actions#############################
####################ToolBar#############################
toolbar = ToolBar(
actions=[
action_open,
action_save,
action_settings,
Separator(),
action_confirm_quit,
action_quit,
],
moveable=True,
floatable=True,
initial_area=RightToolBarArea,
allowed_areas=RightToolBarArea | LeftToolBarArea | TopToolBarArea,
button_style=ToolButtonTextUnderIcon,
)
###################~ToolBar#############################
################Window Event Listener###################
def on_window_create(window: FnExecuteWindow):
print("on_window_create()")
# make action_confirm_quit checked after the select window is created
window.set_action_state(action_confirm_quit, True)
def on_window_close(window: FnExecuteWindow) -> bool:
# get the state of action_confirm_quit
# if it is checked, show a question message box to ask if the user really wants to close the window
# if it is not checked, return True to close the window directly.
state = window.get_action_state(action_confirm_quit)
if state:
# access the
ret = messagebox.show_question_message(
window,
message="Do you really want to close the window?",
title="Quit",
buttons=messagebox.Yes | messagebox.No,
)
return ret == messagebox.Yes
return True
window_listener = SimpleWindowEventListener(
on_create=on_window_create, on_close=on_window_close
)
#################Window Event Listener##################
adapter = GUIAdapter()
adapter.add(
toolbar_example, window_toolbar=toolbar, window_listener=window_listener
)
adapter.run()
| 5,060 | Python | .py | 144 | 27.75 | 107 | 0.588103 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,364 | fn_execute_window_event_example.py | zimolab_PyGUIAdapter/examples/fn_execute_window_event_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
from pyguiadapter.utils import messagebox
def on_window_create(window: FnExecuteWindow):
print("on_create")
def on_window_show(window: FnExecuteWindow):
print("on_show")
def on_window_hide(window: FnExecuteWindow):
print("on_hide")
def on_window_close(window: FnExecuteWindow) -> bool:
print("on_close")
ret = messagebox.show_question_message(
window,
title="Confirm Quit",
message="Are you sure to quit?",
buttons=messagebox.Yes | messagebox.No,
)
if ret == messagebox.Yes:
return True
return False
def on_window_destroy(window: FnExecuteWindow):
print("on_destroy")
def event_example_2():
pass
if __name__ == "__main__":
event_listener = SimpleWindowEventListener(
on_create=on_window_create,
on_show=on_window_show,
on_hide=on_window_hide,
on_close=on_window_close,
on_destroy=on_window_destroy,
)
adapter = GUIAdapter()
adapter.add(event_example_2, window_listener=event_listener)
adapter.run()
| 1,211 | Python | .py | 36 | 28.361111 | 64 | 0.710594 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,365 | utoast_example.py | zimolab_PyGUIAdapter/examples/utoast_example.py | from typing import Literal
from pyguiadapter.adapter import GUIAdapter, utoast
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.extend_types import text_t, color_hex_t, int_slider_t
from pyguiadapter.toast import ToastConfig, AlignLeft, AlignRight, AlignCenter
def utoast_example(
message: text_t,
duration: int = 3000,
opacity: float = 0.9,
fade_out: int = 500,
text_align: Literal["left", "right", "center"] = "center",
text_padding: int = 50,
background_color: color_hex_t = "#323232",
text_color: color_hex_t = "#FFFFFF",
font_size: int = 26,
font_family: str = "Consolas",
position_x: int_slider_t = 50,
position_y: int_slider_t = 50,
):
"""
Show a toast message on the screen.
Args:
message: The message to be displayed.
duration: The duration of the toast message in milliseconds.
opacity: The opacity of the toast message.
fade_out: The duration of the fade out animation in milliseconds.
text_align: The alignment of the text inside the toast message.
text_padding: The padding of the text inside the toast message.
background_color: The background color of the toast message.
text_color: The text color of the toast message.
font_size: The font size of the text inside the toast message.
font_family: The font family of the text inside the toast message.
position_x: The x position (by percentage) of the toast message on the screen.
position_y: The y position (by percentage) of the toast message on the screen.
@params
[opacity]
min_value = 0.0
max_value = 1.0
step = 0.01
[position_x]
min_value = 0
max_value = 100
prefix = "x: "
suffix = "%"
[position_y]
min_value = 0
max_value = 100
prefix = "y: "
suffix = "%"
[background_color]
alpha_channel = false
[text_color]
alpha_channel = false
@end
"""
if not message or message.strip() == "":
raise ParameterError(
parameter_name="message", message="Message cannot be empty."
)
if text_align == "left":
align = AlignLeft
elif text_align == "right":
align = AlignRight
else:
align = AlignCenter
position_x = float(position_x) / 100.0
position_y = float(position_y) / 100.0
toast_config = ToastConfig(
opacity=opacity,
background_color=background_color,
text_color=text_color,
text_padding=text_padding,
text_alignment=align,
font_size=font_size,
font_family=font_family,
position=(position_x, position_y),
fade_out=fade_out,
)
utoast.show_toast(message, duration, toast_config)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(utoast_example)
adapter.run()
| 2,877 | Python | .py | 83 | 28.506024 | 86 | 0.654788 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,366 | equation_solver_5.py | zimolab_PyGUIAdapter/examples/equation_solver_5.py | from typing import Optional
from pyguiadapter.action import Action
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.menu import Menu
from pyguiadapter.utils import messagebox
from pyguiadapter.widgets import FloatSpinBoxConfig
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig, FnExecuteWindow
def equation_solver_5(
a: float = 1.0, b: float = 0.0, c: float = 0.0
) -> Optional[tuple]:
"""A simple equation solver for equations like:
**ax^2 + bx + c = 0** (a, b, c ∈ **R** and a ≠ 0)
@param a: a ∈ R and a ≠ 0
@param b: b ∈ R
@param c: c ∈ R
@return:
"""
if a == 0:
raise ParameterError(parameter_name="a", message="a cannot be zero!")
uprint(f"Equation:")
uprint(f" {a}x² + {b}x + {c} = 0")
delta = b**2 - 4 * a * c
if delta < 0:
return None
x1 = (-b + delta**0.5) / (2 * a)
if delta == 0:
return x1, x1
x2 = (-b - delta**0.5) / (2 * a)
return x1, x2
if __name__ == "__main__":
window_config = FnExecuteWindowConfig(
title="Equation Solver",
icon="mdi6.function-variant",
execute_button_text="Solve",
size=(350, 450),
document_dock_visible=False,
show_function_result=True,
function_result_message="roots: {}",
default_parameter_group_name="Equation Parameters",
# 隐藏`OutputDock`窗口
output_dock_visible=False,
# 因为隐藏了`OutputDock`窗口，所以无需将函数调用结果及函数异常信息打印到输出浏览器中
print_function_error=False,
print_function_result=False,
# 隐藏清除按钮和清楚选框
clear_button_visible=False,
clear_checkbox_visible=False,
)
def on_action_about(wind: FnExecuteWindow, action: Action):
messagebox.show_text_file(
wind,
text_file="./about.html",
text_format="html",
title="About",
)
action_about = Action(text="About", on_triggered=on_action_about)
menu_help = Menu(title="Help", actions=[action_about])
adapter = GUIAdapter()
adapter.add(
equation_solver_5,
window_menus=[menu_help],
window_config=window_config,
widget_configs={
"a": FloatSpinBoxConfig(
default_value=1.0,
decimals=5,
step=0.00005,
),
"b": FloatSpinBoxConfig(decimals=5, step=0.00005),
"c": FloatSpinBoxConfig(decimals=5, step=0.00005),
},
)
adapter.run()
| 2,693 | Python | .py | 75 | 27.186667 | 78 | 0.615631 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,367 | qdarkstyle_example.py | zimolab_PyGUIAdapter/examples/qdarkstyle_example.py | from datetime import datetime
from qtpy.QtWidgets import QApplication
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t
def app_style_example(
arg1: str, arg2: int, arg3: float, arg4: bool, arg5: text_t, arg6: datetime
):
"""
This example requires [QDarkStyleSheet](https://github.com/ColinDuquesnoy/QDarkStyleSheet).
Please install it before you run the example.
<br />
e.g. using `pip`:
> `pip install qdarkstyle`
@param arg1: arg1 description
@param arg2: arg2 description
@param arg3: arg3 description
@param arg4: arg4 description
@param arg5: arg5 description
@param arg6: arg6 description
@return:
@params
[arg6]
calendar_popup = true
@end
"""
pass
if __name__ == "__main__":
import qdarkstyle
def on_app_start(app: QApplication):
# this will be called after the instantiation of QApplication.
print("app started")
app.setStyleSheet(qdarkstyle.load_stylesheet())
adapter = GUIAdapter(on_app_start=on_app_start)
adapter.add(app_style_example)
adapter.run()
| 1,141 | Python | .py | 35 | 27.628571 | 95 | 0.705775 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,368 | equation_solver_4.py | zimolab_PyGUIAdapter/examples/equation_solver_4.py | from typing import Optional
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.widgets import FloatSpinBoxConfig
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def equation_solver_4(
a: float = 1.0, b: float = 0.0, c: float = 0.0
) -> Optional[tuple]:
"""A simple equation solver for equations like:
**ax^2 + bx + c = 0** (a, b, c ∈ **R** and a ≠ 0)
@param a: a ∈ R and a ≠ 0
@param b: b ∈ R
@param c: c ∈ R
@return:
"""
if a == 0:
raise ParameterError(parameter_name="a", message="a cannot be zero!")
uprint(f"Equation:")
uprint(f" {a}x² + {b}x + {c} = 0")
delta = b**2 - 4 * a * c
if delta < 0:
return None
x1 = (-b + delta**0.5) / (2 * a)
if delta == 0:
return x1, x1
x2 = (-b - delta**0.5) / (2 * a)
return x1, x2
if __name__ == "__main__":
window_config = FnExecuteWindowConfig(
title="Equation Solver",
icon="mdi6.function-variant",
execute_button_text="Solve",
size=(350, 550),
document_dock_visible=False,
output_dock_initial_size=(None, 100),
show_function_result=True,
function_result_message="roots: {}",
default_parameter_group_name="Equation Parameters",
)
adapter = GUIAdapter()
adapter.add(
equation_solver_4,
window_config=window_config,
widget_configs={
"a": FloatSpinBoxConfig(
default_value=1.0,
decimals=5,
step=0.00005,
),
"b": FloatSpinBoxConfig(decimals=5, step=0.00005),
"c": FloatSpinBoxConfig(decimals=5, step=0.00005),
},
)
adapter.run()
| 1,828 | Python | .py | 55 | 26.109091 | 77 | 0.593607 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,369 | custom_icon_and_name_example.py | zimolab_PyGUIAdapter/examples/custom_icon_and_name_example.py | from pyguiadapter.adapter import GUIAdapter
def function_1(arg: int):
"""
description of function_1
"""
pass
def function_2(arg: int):
"""
description of function_2
"""
pass
def function_3(arg: int):
"""
description of function_3
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(function_1, display_name="Barcode Generator", icon="ei.barcode")
adapter.add(function_2, display_name="QRCode Generator", icon="ei.qrcode")
adapter.add(function_3, display_name="Generator Service", icon="mdi.web")
adapter.run()
| 605 | Python | .py | 22 | 23.045455 | 80 | 0.66087 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,370 | fn_select_window_event_example.py | zimolab_PyGUIAdapter/examples/fn_select_window_event_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.utils import messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnselect import FnSelectWindow
def on_window_create(window: FnSelectWindow):
print("on_create")
def on_window_show(window: FnSelectWindow):
print("on_show")
def on_window_hide(window: FnSelectWindow):
print("on_hide")
def on_window_close(window: FnSelectWindow) -> bool:
print("on_close")
ret = messagebox.show_question_message(
window,
title="Confirm Quit",
message="Are you sure to quit?",
buttons=messagebox.Yes | messagebox.No,
)
if ret == messagebox.Yes:
return True
return False
def on_window_destroy(window: FnSelectWindow):
print("on_destroy")
def event_example_3():
pass
if __name__ == "__main__":
event_listener = SimpleWindowEventListener(
on_create=on_window_create,
on_show=on_window_show,
on_hide=on_window_hide,
on_close=on_window_close,
on_destroy=on_window_destroy,
)
adapter = GUIAdapter()
adapter.add(event_example_3)
adapter.run(show_select_window=True, select_window_listener=event_listener)
| 1,237 | Python | .py | 36 | 29.083333 | 79 | 0.713564 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,371 | exception_example.py | zimolab_PyGUIAdapter/examples/exception_example.py | from pyguiadapter.adapter import GUIAdapter
def foo(a: int, b: int) -> float:
return float(a) / b
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(foo)
adapter.run()
| 199 | Python | .py | 7 | 24.571429 | 43 | 0.643617 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,372 | fn_select_window_toolbar_example.py | zimolab_PyGUIAdapter/examples/fn_select_window_toolbar_example.py | from pyguiadapter.action import Action
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.toolbar import ToolBar, ToolButtonTextUnderIcon
from pyguiadapter.utils import messagebox
from pyguiadapter.windows.fnselect import FnSelectWindow
def on_action_test(window: FnSelectWindow, action: Action):
messagebox.show_info_message(
window, message=f"Action Triggered!(Action: {action.text})"
)
def on_action_close(window: FnSelectWindow, _: Action):
ret = messagebox.show_question_message(
window,
message="Are you sure to close the application?",
buttons=messagebox.Yes | messagebox.No,
)
if ret == messagebox.Yes:
window.close()
action_test = Action(
text="Test", icon="fa.folder-open", on_triggered=on_action_test, shortcut="Ctrl+O"
)
action_close = Action(
text="Close", icon="fa.close", on_triggered=on_action_close, shortcut="Ctrl+Q"
)
toolbar = ToolBar(
actions=[action_test, action_close],
floatable=True,
button_style=ToolButtonTextUnderIcon,
)
def foo():
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(foo)
adapter.run(show_select_window=True, select_window_toolbar=toolbar)
| 1,218 | Python | .py | 34 | 31.676471 | 86 | 0.734015 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,373 | save_load_parameters_example.py | zimolab_PyGUIAdapter/examples/save_load_parameters_example.py | import enum
import json
from typing import Any, Dict
from pyguiadapter.action import Action
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.extend_types import color_t, json_obj_t
from pyguiadapter.menu import Menu
from pyguiadapter.utils import PyLiteralType, messagebox, filedialog, io
from pyguiadapter.windows.fnexec import FnExecuteWindow
def on_action_save_params(window: FnExecuteWindow, _: Action):
# check if the function is executing
if window.is_function_executing():
messagebox.show_warning_message(window, "Function is executing")
return
# get current parameter values from widgets
# some values may be invalid so do not forget to catch errors
try:
params: Dict[str, Any] = window.get_parameter_values()
except ParameterError as e:
# if the error is a ParameterError, we can let the window handle it
window.process_parameter_error(e)
return
except Exception as e:
# if the error is not a ParameterError, we need handle it by ourselves properly
# here we just show the exception message with a message box
messagebox.show_exception_messagebox(
window, e, message="Unable to get the parameters: {}"
)
return
# now we get current parameter values, we can save them to a location like a file in disk.
# So the next topic is how to serialize them properly? It is not a simple problem especially considering that we
# need to load them back later.
# If the all the parameters are basic or simple types, json is a good choice. However, considering the current case
# we are working on, we have to deal with some complex types, like:
# arg5: list
# arg6: dict
# arg7: tuple
# arg8: set
# arg10: json_obj_t
# arg11: WeekDay
# arg12: PyLiteralType
# In order to save them in json files and read them back accurately, a lot of efforts need to be made in
# serialization and deserialization logic.
# As a demo, we decide not to do this job so precisely here. In a real life project, please be careful.
# let user select a file to save the parameters
save_path = filedialog.get_save_file(
window, title="Save Parameters", start_dir="./", filters="JSON files(*.json)"
)
if not save_path:
return
# process arg8 which is a set
# set is not serializable by default, so we have to convert it to a list first
if "arg8" in params:
arg8 = params["arg8"]
params["arg8"] = list(arg8)
# process arg11 which is an enum of type WeekDay
if "arg11" in params:
arg11 = params["arg11"]
params["arg11"] = arg11.value
try:
# process arg12 which is a PyLiteralType
if "arg12" in params:
arg12 = params["arg12"]
params["arg12"] = json.dumps(arg12, ensure_ascii=False)
serialized_params = json.dumps(params, ensure_ascii=False)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to serialize the parameters: "
)
return
try:
io.write_text_file(save_path, serialized_params)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to save the parameters: "
)
return
messagebox.show_info_message(window, "Parameters have been saved!")
def on_action_load_params(window: FnExecuteWindow, _: Action):
# check if the function is executing
if window.is_function_executing():
messagebox.show_warning_message(window, "Function is executing")
return
load_path = filedialog.get_open_file(
window, title="Load Parameters", start_dir="./", filters="JSON files(*.json)"
)
if not load_path:
return
try:
content = io.read_text_file(load_path)
params = json.loads(content)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to load the parameters: "
)
return
if not isinstance(params, dict):
messagebox.show_critical_message(window, message="Invalid parameters format!")
return
# some parameters require special processing to be converted back
try:
# arg7: tuple
if "arg7" in params:
arg7 = params["arg7"]
params["arg7"] = tuple(arg7)
# arg8: set
if "arg8" in params:
arg8 = params["arg8"]
params["arg8"] = set(arg8)
# arg9: color_t (which is actually a tuple of 3 or 4 elements in this demo)
if "arg9" in params:
arg9 = params["arg9"]
params["arg9"] = tuple(arg9)
# arg11: WeekDay (a Enum class)
if "arg11" in params:
arg11 = params["arg11"]
params["arg11"] = WeekDay(arg11)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Invalid parameters format: "
)
return
try:
window.set_parameter_values(params)
except ParameterError as e:
window.process_parameter_error(e)
return
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to set the parameters: "
)
return
else:
messagebox.show_info_message(window, "Parameters have been loaded!")
class WeekDay(enum.Enum):
Monday = 1
Tuesday = 2
Wednesday = 3
Thursday = 4
Friday = 5
Saturday = 6
Sunday = 7
def load_save_example(
arg1: int,
arg2: str,
arg3: bool,
arg4: float,
arg5: list,
arg6: dict,
arg7: tuple,
arg8: set,
arg9: color_t,
arg10: json_obj_t,
arg11: WeekDay,
ar12: PyLiteralType,
):
uprint("arg1=", arg1)
uprint("arg2=", arg2)
uprint("arg3=", arg3)
uprint("arg4=", arg4)
uprint("arg5=", arg5)
uprint("arg6=", arg6)
uprint("arg7=", arg7)
uprint("arg8=", arg8)
uprint("arg9=", arg9)
uprint("arg10=", arg10)
uprint("arg11=", arg11)
uprint("ar12=", ar12)
if __name__ == "__main__":
action_save_params = Action(
text="Save Parameters",
icon="fa.save",
shortcut="Ctrl+S",
on_triggered=on_action_save_params,
)
action_load_params = Action(
text="Load Parameters",
icon="fa.folder-open",
shortcut="Ctrl+L",
on_triggered=on_action_load_params,
)
file_menu = Menu(
title="File",
actions=[action_save_params, action_load_params],
)
adapter = GUIAdapter()
adapter.add(load_save_example, window_menus=[file_menu])
adapter.run()
| 6,833 | Python | .py | 190 | 29.157895 | 119 | 0.648853 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,374 | uinput_builtin_dialog_example.py | zimolab_PyGUIAdapter/examples/uinput_builtin_dialog_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter import uinput
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.extend_types import choices_t
# noinspection SpellCheckingInspection
def uinput_example(inputs: choices_t):
"""
Example of getting user inputs inside the function
@param inputs: choose what you want to get from user
@return:
@params
[inputs]
choices = ["int", "str", "text", "float", "item", "color", "dir", "file", "save_file", "files", "json object", "python literal"]
columns = 2
@end
"""
if "int" in inputs:
value = uinput.get_int(title="Input Integer", label="Enter an integer:")
uprint("User inputs: ", value)
if "str" in inputs:
value = uinput.get_string(title="Input Text", label="Enter a string:")
uprint("User inputs: ", value)
if "text" in inputs:
value = uinput.get_text(title="Input Text", label="Enter a string:")
uprint("User inputs: ", value)
if "float" in inputs:
value = uinput.get_float(title="Input Float", label="Enter a float:")
uprint("User inputs: ", value)
if "item" in inputs:
value = uinput.get_selected_item(
items=["Item 1", "Item 2", "Item 3", "Item 4"],
title="Select Item",
label="Select an item:",
)
uprint("User inputs: ", value)
if "color" in inputs:
value = uinput.get_color(title="Select Color", alpha_channel=True)
uprint("User inputs: ", value)
if "dir" in inputs:
value = uinput.get_existing_directory(title="Select Directory")
uprint("User inputs: ", value)
if "file" in inputs:
value = uinput.get_open_file(title="Select File")
uprint("User inputs: ", value)
if "save_file" in inputs:
value = uinput.get_save_file(title="Select File")
uprint("User inputs: ", value)
if "files" in inputs:
value = uinput.get_open_files(title="Select Files")
uprint("User inputs: ", value)
if "json object" in inputs:
value = uinput.get_json_object(title="Input Json Object")
uprint("User inputs: ", value, f" {type(value)}")
if "python literal" in inputs:
value = uinput.get_py_literal(title="Input Python Literal")
uprint("User inputs: ", value, f" {type(value)}")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(uinput_example)
adapter.run()
| 2,474 | Python | .py | 60 | 34.533333 | 132 | 0.630814 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,375 | more_widgets_example.py | zimolab_PyGUIAdapter/examples/more_widgets_example.py | import enum
from datetime import datetime, date, time
from typing import List, Tuple, Literal
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.extend_types import (
int_t,
int_dial_t,
int_slider_t,
float_t,
file_t,
directory_t,
files_t,
color_t,
string_list_t,
plain_dict_t,
json_obj_t,
text_t,
key_sequence_t,
choices_t,
choice_t,
)
class WeekDays(enum.Enum):
Monday = 1
Tuesday = 2
Wednesday = 3
Thursday = 4
Friday = 5
Saturday = 6
Sunday = 7
def more_widgets_example(
arg1: int,
arg2: float,
arg3: str,
arg4: bool,
arg5: List[int],
arg6: Tuple,
arg7: dict,
arg9: set,
arg10: int_t,
arg11: int_dial_t,
arg12: int_slider_t,
arg13: float_t,
arg14: file_t,
arg15: directory_t,
arg16: files_t,
arg17: Literal["a", "b", "c"],
arg18: color_t,
arg19: string_list_t,
arg20: plain_dict_t,
arg21: json_obj_t,
arg22: text_t,
arg23: datetime,
arg24: date,
arg25: time,
arg26: key_sequence_t,
arg27: choices_t,
arg28: choice_t,
arg29: WeekDays,
):
"""
@params
[arg27]
choices = ["a", "b", "c", "d", "e", "f"]
columns = 2
[arg28]
choices = ["a", "b", "c", "d", "e", "f"]
editable = true
@end
"""
uprint("arg1=", arg1)
uprint("arg2=", arg2)
uprint("arg3=", arg3)
uprint("arg4=", arg4)
uprint("arg5=", arg5)
uprint("arg6=", arg6)
uprint("arg7=", arg7)
uprint("arg9=", arg9)
uprint("arg10=", arg10)
uprint("arg11=", arg11)
uprint("arg12=", arg12)
uprint("arg13=", arg13)
uprint("arg14=", arg14)
uprint("arg15=", arg15)
uprint("arg16=", arg16)
uprint("arg17=", arg17)
uprint("arg18=", arg18)
uprint("arg19=", arg19)
uprint("arg20=", arg20)
uprint("arg21=", arg21)
uprint("arg22=", arg22)
uprint("arg23=", arg23)
uprint("arg24=", arg24)
uprint("arg25=", arg25)
uprint("arg26=", arg26)
uprint("arg27=", arg27)
uprint("arg28=", arg28)
uprint("arg29=", arg29)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(more_widgets_example)
adapter.run()
| 2,275 | Python | .py | 102 | 17.627451 | 47 | 0.593432 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,376 | global_style_example_2.py | zimolab_PyGUIAdapter/examples/global_style_example_2.py | import os.path
from datetime import datetime
from pyguiadapter import utils
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t
def app_style_example(
arg1: str, arg2: int, arg3: float, arg4: bool, arg5: text_t, arg6: datetime
):
"""
This example shows how to apply a global stylesheet to the application.
First, read the stylesheet from a qss file. In this example, [Ubuntu.qss](https://github.com/GTRONICK/QSS/blob/master/Ubuntu.qss)
will be used. This qss file is from GTRONICK's [QSS](https://github.com/GTRONICK/QSS) repo.
Then, pass the qss content to the GUIAdapter constructor's **global_style** argument.
"""
pass
if __name__ == "__main__":
QSS_FILE = os.path.join(os.path.dirname(__file__), "Ubuntu.qss")
def load_stylesheet() -> str:
"""
This function will be called after the QApplication instance being created
"""
return utils.read_text_file(QSS_FILE)
adapter = GUIAdapter(global_stylesheet=load_stylesheet)
adapter.add(app_style_example)
adapter.run()
| 1,099 | Python | .py | 25 | 39.16 | 133 | 0.714017 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,377 | equation_solver_3.py | zimolab_PyGUIAdapter/examples/equation_solver_3.py | from typing import Optional
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def equation_solver_3(
a: float = 1.0, b: float = 0.0, c: float = 0.0
) -> Optional[tuple]:
"""A simple equation solver for equations like:
**ax^2 + bx + c = 0** (a, b, c ∈ **R** and a ≠ 0)
@param a: a ∈ R and a ≠ 0
@param b: b ∈ R
@param c: c ∈ R
@return:
@params
[a]
decimals = 5
step = 0.00005
[b]
decimals = 5
step = 0.00005
[c]
decimals = 5
step = 0.00005
@end
"""
if a == 0:
raise ParameterError(parameter_name="a", message="a cannot be zero!")
uprint(f"Equation:")
uprint(f" {a}x² + {b}x + {c} = 0")
delta = b**2 - 4 * a * c
if delta < 0:
return None
x1 = (-b + delta**0.5) / (2 * a)
if delta == 0:
return x1, x1
x2 = (-b - delta**0.5) / (2 * a)
return x1, x2
if __name__ == "__main__":
window_config = FnExecuteWindowConfig(
title="Equation Solver",
icon="mdi6.function-variant",
execute_button_text="Solve",
size=(350, 550),
document_dock_visible=False,
output_dock_initial_size=(None, 100),
show_function_result=True,
function_result_message="roots: {}",
default_parameter_group_name="Equation Parameters",
)
adapter = GUIAdapter()
adapter.add(equation_solver_3, window_config=window_config)
adapter.run()
| 1,603 | Python | .py | 53 | 24.471698 | 77 | 0.603934 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,378 | menu_toolbar_example.py | zimolab_PyGUIAdapter/examples/menu_toolbar_example.py | from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.toolbar import ToolBar
from pyguiadapter.windows.fnexec import FnExecuteWindow
from pyguiadapter.utils import messagebox, filedialog
def on_action_about(window: FnExecuteWindow, _: Action):
messagebox.show_info_message(
parent=window,
message="This is an example of toolbar and menu with custom actions.",
title="About",
)
def on_action_close(window: FnExecuteWindow, _: Action):
ret = messagebox.show_question_message(
window, "Are you sure you want to quit?", buttons=messagebox.Yes | messagebox.No
)
if ret == messagebox.Yes:
window.close()
def on_action_open(window: FnExecuteWindow, _: Action):
ret = filedialog.get_open_file(
window,
title="Open File",
start_dir="./",
filters="JSON files(*.json);;Python files(*.py);;All files(*.*)",
)
if not ret:
return
messagebox.show_info_message(window, f"File will be opened: {ret}")
def on_action_save(window: FnExecuteWindow, _: Action):
ret = filedialog.get_save_file(
window,
title="Save File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if not ret:
return
messagebox.show_info_message(window, f"File will be saved: {ret}")
def on_action_auto_theme(window: FnExecuteWindow, _: Action, checked: bool):
if checked:
messagebox.show_info_message(window, "Auto theme is selected.")
def on_action_light_theme(window: FnExecuteWindow, _: Action, checked: bool):
if checked:
messagebox.show_info_message(window, "Light theme is selected.")
def on_action_dark_theme(window: FnExecuteWindow, _: Action, checked: bool):
if checked:
messagebox.show_info_message(window, "Dark theme is selected.")
action_about = Action(
text="About",
icon="fa.info-circle",
on_triggered=on_action_about,
)
action_open = Action(
text="Open",
icon="fa.folder-open",
shortcut="Ctrl+O",
on_triggered=on_action_open,
)
action_save = Action(
text="Save",
icon="fa.save",
shortcut="Ctrl+S",
on_triggered=on_action_save,
)
action_close = Action(
text="Quit",
icon="fa.close",
shortcut="Ctrl+Q",
on_triggered=on_action_close,
)
action_auto_them = Action(
text="Auto",
checkable=True,
checked=True,
on_toggled=on_action_auto_theme,
)
action_light_theme = Action(
text="Light",
checkable=True,
on_toggled=on_action_light_theme,
)
action_dark_theme = Action(
text="Dark",
checkable=True,
on_toggled=on_action_dark_theme,
)
submenu_theme = Menu(
title="Theme",
actions=[action_auto_them, action_light_theme, action_dark_theme],
exclusive=True,
)
menu_file = Menu(
title="File",
actions=[
action_open,
action_save,
Separator(),
action_close,
Separator(),
submenu_theme,
],
)
menu_help = Menu(
title="Help",
actions=[action_about],
)
def menu_toolbar_example(arg1: int, arg2: str, arg3: bool):
"""
This example shows how to add and config toolbar and menus to the window.
@param arg1:
@param arg2:
@param arg3:
@return:
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(
menu_toolbar_example,
window_menus=[menu_file, menu_help],
window_toolbar=ToolBar(
actions=[action_open, action_save, Separator(), action_close]
),
)
adapter.run()
| 3,637 | Python | .py | 125 | 24.04 | 88 | 0.663128 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,379 | clipboard_example.py | zimolab_PyGUIAdapter/examples/clipboard_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter import uclipboard
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.extend_types import text_t
def clipboard_example(text: text_t):
uprint("selection support:", uclipboard.supports_selection())
uprint("owns selection:", uclipboard.owns_selection())
uprint("owns clipboard:", uclipboard.owns_clipboard())
uprint("selection text:")
uprint(uclipboard.get_selection_text())
uprint("clipboard text:")
uprint(uclipboard.get_text())
uprint("set clipboard text...")
uclipboard.set_text(text)
uprint("clipboard text:")
uprint(uclipboard.get_text())
uprint("======================")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(clipboard_example)
adapter.run()
| 818 | Python | .py | 21 | 34.904762 | 65 | 0.713745 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,380 | style_override_example.py | zimolab_PyGUIAdapter/examples/style_override_example.py | """
This example requires PyQtDarkTheme. Please install it before you run this example.
"""
from datetime import datetime
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import text_t
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig, OutputBrowserConfig
def app_style_example(
arg1: str, arg2: int, arg3: float, arg4: bool, arg5: text_t, arg6: datetime
):
"""
This example requires [PyQtDarkTheme](https://github.com/5yutan5/PyQtDarkTheme).
Please install it before you run the example.
<br />
e.g. using `pip`:
> `pip install pyqtdarktheme`
<br />
The style of output browser will be overridden with **OutputBrowserConfig.stylesheet**
@param arg1: arg1 description
@param arg2: arg2 description
@param arg3: arg3 description
@param arg4: arg4 description
@param arg5: arg5 description
@param arg6: arg6 description
@return:
"""
pass
if __name__ == "__main__":
import qdarktheme
def on_app_start(app):
# this will be called after the instantiation of QApplication.
print("app started")
qdarktheme.setup_theme("dark")
adapter = GUIAdapter(on_app_start=on_app_start)
adapter.add(
app_style_example,
window_config=FnExecuteWindowConfig(
output_browser_config=OutputBrowserConfig(
stylesheet="""
background-color: "#380C2A";
color: "#FFFFFF";
font-family: "Consolas";
font-size: 12pt;
"""
)
),
)
adapter.run()
| 1,623 | Python | .py | 48 | 27.041667 | 90 | 0.660051 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,381 | parameter_anchor_example.py | zimolab_PyGUIAdapter/examples/parameter_anchor_example.py | from datetime import datetime, date, time
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.extend_types import string_list_t
from pyguiadapter.widgets import IntSpinBoxConfig
from pyguiadapter.windows import DocumentBrowserConfig
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def parameter_anchor_example(
a: int,
b: float,
c: bool,
d: str,
e: datetime,
f: date,
h: time,
i: list,
j: tuple,
k: dict,
l: string_list_t,
):
"""
This example show how to use parameter anchor.
<br>
<h3>Parameter Anchors</h3>
<a href="#param=a">click to jump to parameter a.</a>
<br>
<a href="#param=b">click to jump to parameter b.</a>
<br>
<a href="#param=c">click to jump to parameter c.</a>
<br>
<a href="#param=d">click to jump to parameter d.</a>
<br>
<a href="#param=e">click to jump to parameter e.</a>
<br>
<a href="#param=f">click to jump to parameter f.</a>
<br>
<a href="#param=h">click to jump to parameter h.</a>
<br>
<a href="#param=i">click to jump to parameter i.</a>
<br>
<a href="#param=j">click to jump to parameter j.</a>
<br>
<a href="#param=k">click to jump to parameter k.</a>
<br>
<a href="#param=l">click to jump to parameter l.</a>
<h3>Group Anchors</h3>
<a href="#group=">click to jump to default group</a>
<br>
<a href="#group=Group A">click to jump to group Group A.</a>
<br>
<a href="#group=Group B">click to jump to group Group B.</a>
<br>
<a href="#group=Group C">click to jump to group Group C.</a>
Args:
a: description of a.
b: description of b.
c: description of c.
d: description of d.
e: description of e.
f: description of f.
l: description of l.
k: description of k.
j: description of j.
i: description of i.
h: description of h.
Returns:
"""
pass
if __name__ == "__main__":
a_conf = IntSpinBoxConfig(group="Group A")
b_conf = IntSpinBoxConfig(group="Group B")
c_conf = IntSpinBoxConfig(group="Group C")
adapter = GUIAdapter()
adapter.add(
parameter_anchor_example,
document_format="html",
widget_configs={"a": a_conf, "b": b_conf, "c": c_conf},
window_config=FnExecuteWindowConfig(
document_browser_config=DocumentBrowserConfig(
parameter_anchor=True, group_anchor=True
)
),
)
adapter.run()
| 2,533 | Python | .py | 83 | 24.771084 | 64 | 0.617021 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,382 | fn_execute_window_event_example_2.py | zimolab_PyGUIAdapter/examples/fn_execute_window_event_example_2.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.utils import messagebox
from pyguiadapter.windows.fnexec import (
FnExecuteWindow,
SimpleFnExecuteWindowEventListener,
)
def on_execute_start(window: FnExecuteWindow):
print("on_execute_start()")
def on_execute_result(window: FnExecuteWindow, result) -> bool:
print(f"on_execute_result(): {result}")
messagebox.show_info_message(window, message=f"Result: {result}", title="Result")
return False
def on_execute_error(window: FnExecuteWindow, error) -> bool:
print(f"on_execute_error(): {error}")
messagebox.show_exception_messagebox(window, error)
return False
def on_execute_finish(window: FnExecuteWindow):
print("on_execute_finish()")
def event_example_3(a: int = 1, b: int = 1):
return a / b
if __name__ == "__main__":
event_listener = SimpleFnExecuteWindowEventListener(
on_execute_start=on_execute_start,
on_execute_result=on_execute_result,
on_execute_error=on_execute_error,
on_execute_finish=on_execute_finish,
)
adapter = GUIAdapter()
adapter.add(event_example_3, window_listener=event_listener)
adapter.run()
| 1,189 | Python | .py | 30 | 35.033333 | 85 | 0.728858 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,383 | function_groups_example.py | zimolab_PyGUIAdapter/examples/function_groups_example.py | from pyguiadapter.adapter import GUIAdapter
def mp4_encoder():
"""
MP4 Encoder
"""
pass
def mp3_encoder():
"""
MP3 Encoder
"""
pass
def avi_encoder():
"""
AVI Encoder
"""
pass
def ogg_encoder():
"""
OGG Encoder
"""
pass
def avi_decoder():
"""
AVI Decoder
"""
pass
def ogg_decoder():
"""
OGG Decoder
"""
pass
def mp3_decoder():
"""
MP3 Decoder
"""
pass
def mp4_decoder():
"""
MP4 Decoder
"""
pass
def universal_settings():
"""
Universal Settings
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(universal_settings)
adapter.add(mp4_encoder, group="Encoders")
adapter.add(mp3_encoder, group="Encoders")
adapter.add(avi_encoder, group="Encoders")
adapter.add(ogg_encoder, group="Encoders")
adapter.add(avi_decoder, group="Decoders")
adapter.add(ogg_decoder, group="Decoders")
adapter.add(mp3_decoder, group="Decoders")
adapter.add(mp4_decoder, group="Decoders")
adapter.run()
| 1,102 | Python | .py | 58 | 14.413793 | 46 | 0.609375 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,384 | hide_docks_example.py | zimolab_PyGUIAdapter/examples/hide_docks_example.py | from pyguiadapter.adapter import GUIAdapter, udialog
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def hide_docks_example(a: int, b: str, c: bool):
if a < 0:
raise ParameterError("a", "a must >= 0")
if b == "":
raise ValueError("invalid value of b")
udialog.show_info_messagebox(f"Receive: {a}, {b}, {c}")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(
hide_docks_example,
window_config=FnExecuteWindowConfig(
size=(300, 400),
document_dock_visible=False,
output_dock_visible=False,
clear_button_visible=False,
clear_checkbox_visible=False,
execute_button_text="Start",
default_parameter_group_name="Parameters",
),
)
adapter.run()
| 871 | Python | .py | 24 | 28.791667 | 61 | 0.635824 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,385 | menu_example_2.py | zimolab_PyGUIAdapter/examples/menu_example_2.py | import json
from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.toolbar import ToolBar
from pyguiadapter.utils import filedialog, inputdialog, messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
def menu_example():
pass
###################Action Callbacks#########################
def on_action_open(window: FnExecuteWindow, _: Action):
print("on_action_open()")
ret = filedialog.get_open_file(
parent=window,
title="Open File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be opened: {ret}")
def on_action_save(window: FnExecuteWindow, _: Action):
print("on_action_save()")
ret = filedialog.get_save_file(
parent=window,
title="Save File",
start_dir="./",
filters="JSON files(*.json);;All files(*.*)",
)
if ret:
messagebox.show_info_message(window, f"File will be saved to: {ret}")
def on_action_settings(window: FnExecuteWindow, _: Action):
default_settings = {
"opt1": 1,
"opt2": "2",
"opt3": True,
}
new_settings = inputdialog.input_json_object(
parent=window,
title="Settings",
icon="fa.cog",
size=(600, 400),
ok_button_text="Save",
cancel_button_text="Cancel",
initial_text=json.dumps(default_settings, indent=4, ensure_ascii=False),
auto_indent=True,
indent_size=4,
auto_parentheses=True,
line_wrap_mode=inputdialog.LineWrapMode.WidgetWidth,
line_wrap_width=88,
)
if isinstance(new_settings, dict):
messagebox.show_info_message(window, f"new settings: {new_settings}")
def on_action_confirm_quit(window: FnExecuteWindow, _: Action, checked: bool):
print("on_action_confirm_close(): ", checked)
def on_action_close(window: FnExecuteWindow, _: Action):
print("on_action_close()")
window.close()
def on_action_about(window: FnExecuteWindow, _: Action):
print("on_action_about()")
about_text = """
<h1>PyGUIAdapter V2</h1>
<p>PyGUIAdapter is a GUI lib for those who want make GUI application without writing GUI code!</p>
<p>
You can access the source code <a href="https://github.com/zimolab/PyGUIAdapter">here</a>!
</p>
"""
messagebox.show_text_content(
window,
text_content=about_text,
text_format="html",
title="About PyGUIAdapter",
icon="fa.info-circle",
)
def on_action_license(window: FnExecuteWindow, _: Action):
print("on_action_license()")
license_file = "../../LICENSE"
messagebox.show_text_file(
window,
text_file=license_file,
text_format="plaintext",
title="License",
icon="fa.copyright",
)
###################~Action Callbacks#########################
if __name__ == "__main__":
###################Actions#############################
action_open = Action(
text="Open",
icon="fa.folder-open",
on_triggered=on_action_open,
shortcut="Ctrl+O",
)
action_save = Action(
text="Save",
icon="fa.save",
on_triggered=on_action_save,
shortcut="Ctrl+S",
)
action_settings = Action(
text="Settings",
icon="fa.cog",
on_triggered=on_action_settings,
shortcut="Ctrl+,",
)
action_quit = Action(
text="Quit",
icon="fa.power-off",
on_triggered=on_action_close,
shortcut="Ctrl+Q",
)
action_confirm_quit = Action(
text="Confirm Quit",
icon="fa.question-circle",
checkable=True,
checked=False,
on_toggled=on_action_confirm_quit,
)
action_about = Action(
text="About",
icon="fa.info-circle",
on_triggered=on_action_about,
)
action_license = Action(
text="License",
icon="fa.copyright",
on_triggered=on_action_license,
)
###################~Actions#############################
####################Menus#############################
submenu_help = Menu(
title="Help",
actions=[action_about, action_license],
)
menu_file = Menu(
title="File",
actions=[
action_open,
action_save,
Separator(),
submenu_help,
Separator(),
action_quit,
],
)
menu_settings = Menu(
title="Settings",
actions=[action_settings, Separator(), action_confirm_quit],
)
menus = [menu_file, menu_settings]
###################~Menus#############################
################Window Event Listener###################
def on_window_create(window: FnExecuteWindow):
print("on_window_create()")
# make action_confirm_quit checked after the select window is created
window.set_action_state(action_confirm_quit, True)
def on_window_close(window: FnExecuteWindow) -> bool:
# get the state of action_confirm_quit
# if it is checked, show a question message box to ask if the user really wants to close the window
# if it is not checked, return True to close the window directly.
state = window.get_action_state(action_confirm_quit)
if state:
# access the
ret = messagebox.show_question_message(
window,
message="Do you really want to close the window?",
title="Quit",
buttons=messagebox.Yes | messagebox.No,
)
return ret == messagebox.Yes
return True
window_listener = SimpleWindowEventListener(
on_create=on_window_create, on_close=on_window_close
)
#################Window Event Listener##################
adapter = GUIAdapter()
adapter.add(
menu_example,
window_menus=menus,
window_toolbar=ToolBar(
actions=[
action_open,
action_save,
Separator(),
action_settings,
Separator(),
action_about,
action_license,
]
),
window_listener=window_listener,
)
adapter.run()
| 6,463 | Python | .py | 193 | 25.761658 | 107 | 0.575 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,386 | equation_solver.py | zimolab_PyGUIAdapter/examples/equation_solver.py | from typing import Optional
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
def equation_solver(a: float = 1.0, b: float = 0.0, c: float = 0.0) -> Optional[tuple]:
"""A simple equation solver for equations like:
**ax^2 + bx + c = 0** (a, b, c ∈ **R** and a ≠ 0)
@param a: a ∈ R and a ≠ 0
@param b: b ∈ R
@param c: c ∈ R
@return:
"""
if a == 0:
raise ParameterError(parameter_name="a", message="a cannot be zero!")
uprint(f"Equation:")
uprint(f" {a}x² + {b}x + {c} = 0")
delta = b**2 - 4 * a * c
if delta < 0:
return None
x1 = (-b + delta**0.5) / (2 * a)
if delta == 0:
return x1, x1
x2 = (-b - delta**0.5) / (2 * a)
return x1, x2
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(equation_solver)
adapter.run()
| 939 | Python | .py | 28 | 28.25 | 87 | 0.583614 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,387 | udialog_std_messagebox_example.py | zimolab_PyGUIAdapter/examples/udialog_std_messagebox_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter import udialog
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.extend_types import text_t
from pyguiadapter.utils import messagebox
def dialog_example(
info_message: text_t,
warning_message: text_t,
error_message: text_t,
question_message: text_t,
):
if info_message:
udialog.show_info_messagebox(
text=info_message,
title="Information",
buttons=messagebox.Ok | messagebox.No,
default_button=messagebox.Ok,
)
if warning_message:
udialog.show_warning_messagebox(
text=warning_message,
title="Warning",
buttons=messagebox.Ok | messagebox.No,
default_button=messagebox.Ok,
)
if error_message:
udialog.show_critical_messagebox(
text=error_message,
title="Error",
buttons=messagebox.Ok | messagebox.No,
default_button=messagebox.Ok,
)
if question_message:
answer = udialog.show_question_messagebox(
text=question_message,
title="Question",
buttons=messagebox.Yes | messagebox.No,
default_button=messagebox.No,
)
if answer == messagebox.Yes:
uprint("Your Choice: Yes")
udialog.show_info_messagebox("You Choose Yes!", title="Answer")
else:
uprint("Your Choice: No")
udialog.show_info_messagebox("You Choose No!", title="Answer")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(dialog_example)
adapter.run()
| 1,673 | Python | .py | 49 | 25.571429 | 75 | 0.629561 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,388 | output_logger_example.py | zimolab_PyGUIAdapter/examples/output_logger_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import Logger, LoggerConfig
logger = Logger(
config=LoggerConfig(
info_color="green",
debug_color="blue",
warning_color="yellow",
critical_color="red",
fatal_color="magenta",
)
)
def output_log_msg(
info_msg: str = "info message",
debug_msg: str = "debug message",
warning_msg: str = "warning message",
critical_msg: str = "critical message",
fatal_msg: str = "fatal message",
):
logger.info(info_msg)
logger.debug(debug_msg)
logger.warning(warning_msg)
logger.critical(critical_msg)
logger.fatal(fatal_msg)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(output_log_msg)
adapter.run()
| 784 | Python | .py | 27 | 24.111111 | 61 | 0.668442 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,389 | uoutput_example.py | zimolab_PyGUIAdapter/examples/uoutput_example.py | from pyguiadapter.adapter import GUIAdapter, uoutput
def output_log_msg(
info_msg: str = "info message",
debug_msg: str = "debug message",
warning_msg: str = "warning message",
critical_msg: str = "critical message",
fatal_msg: str = "fatal message",
):
uoutput.info(info_msg)
uoutput.debug(debug_msg)
uoutput.warning(warning_msg)
uoutput.critical(critical_msg)
uoutput.fatal(fatal_msg)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(output_log_msg, document_format="html")
adapter.run()
| 559 | Python | .py | 17 | 28.588235 | 55 | 0.685874 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,390 | cancel_function_example.py | zimolab_PyGUIAdapter/examples/cancel_function_example.py | import time
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.ucontext import is_function_cancelled
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.windows.fnexec import FnExecuteWindowConfig
def cancel_function_example(target: int = 10, delay_per_iter: float = 0.5):
"""
@params
[target]
min_value = 1
[delay_per_iter]
min_value = 0.001
step = 0.5
decimals = 3
@end
"""
for i in range(target):
if is_function_cancelled():
break
uprint("process: ", i)
time.sleep(delay_per_iter)
uprint("done!")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(
cancel_function_example,
cancelable=True,
window_config=FnExecuteWindowConfig(disable_widgets_on_execute=True),
)
adapter.run()
| 855 | Python | .py | 30 | 23.133333 | 77 | 0.672372 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,391 | html_docstring_document_example.py | zimolab_PyGUIAdapter/examples/html_docstring_document_example.py | from pyguiadapter.adapter import GUIAdapter
def function_2(arg1: int, arg2: str, arg3: bool):
"""
<h3>Description</h3>
<p>
This is the document of the <b>function_3</b>. And by default this document will automatically
appear in the <strong>document area</strong>.
</p>
<p>
The format of the document is <b>Markdown</b> by default.
The <b>plaintext</b> and <b>html</b> formats are also supported.
</p>
<hr>
<h3>Arguments</h3>
<p>This function needs 3 arguments:</p>
<ul>
<li><b>arg1</b>: Balabala....</li>
<li><b>arg2</b>: Balabala....</li>
<li><b>arg3</b>: Balabala....</li>
</ul>
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(function_2, document_format="html")
adapter.run()
| 799 | Python | .py | 26 | 26 | 98 | 0.611979 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,392 | dock_property_example.py | zimolab_PyGUIAdapter/examples/dock_property_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.windows.fnexec import RightDockWidgetArea, FnExecuteWindowConfig
def dock_area_example_1():
"""
This example shows how to change the dock area and initial size of the output and document docks.
@return:
"""
pass
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(
dock_area_example_1,
window_config=FnExecuteWindowConfig(
document_dock_initial_size=(614, 538),
output_dock_initial_area=RightDockWidgetArea,
output_dock_initial_size=(None, 230),
),
)
adapter.run()
| 637 | Python | .py | 19 | 27.263158 | 101 | 0.675896 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,393 | confirm_quit_example.py | zimolab_PyGUIAdapter/examples/confirm_quit_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.utils import messagebox
from pyguiadapter.window import SimpleWindowEventListener
from pyguiadapter.windows.fnexec import FnExecuteWindow
def foo(arg1: int, arg2: str, arg3: bool):
pass
def on_close(window: FnExecuteWindow) -> bool:
ret = messagebox.show_question_message(
window, message="Are you sure you want to quit?"
)
if ret == messagebox.Yes:
# when `on_close()` returns True, the window will be closed
return True
else:
messagebox.show_info_message(window, message="Quit cancelled by user!")
# when `on_close()` returns False, the window will not be closed
# in other words, the close event will be ignored
return False
if __name__ == "__main__":
# create a window listener that listens for the `on_close` event
event_listener = SimpleWindowEventListener(on_close=on_close)
adapter = GUIAdapter()
# add the `foo` function to the adapter with the `window_listener` argument set to the event_listener
adapter.add(foo, window_listener=event_listener)
adapter.run()
| 1,139 | Python | .py | 25 | 40.28 | 105 | 0.722674 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,394 | simple_load_save_example.py | zimolab_PyGUIAdapter/examples/simple_load_save_example.py | import json
from typing import Dict, Any
from pyguiadapter.action import Action
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
from pyguiadapter.extend_types import color_t
from pyguiadapter.menu import Menu
from pyguiadapter.utils import messagebox, filedialog
from pyguiadapter.windows.fnexec import FnExecuteWindow
def simple_load_save_example(
arg1: int,
arg2: float,
arg3: bool,
arg4: str,
arg5: color_t,
):
"""
This example shows how to save current parameter values to a json file and load a parameter values from a json file.
@param arg1:
@param arg2:
@param arg3:
@param arg4:
@param arg5:
@return:
"""
uprint("arg1=", arg1)
uprint("arg2=", arg2)
uprint("arg3=", arg3)
uprint("arg4=", arg4)
uprint("arg5=", arg5)
def on_save_params(window: FnExecuteWindow, _: Action):
# Step 1: obtain current parameter values from widgets
#
# if the current input in the widgets of some parameter is invalid, the get_parameter_values() method may raise a
# exception. A good practice is to catch the exception and handle it properly:
# - for ParameterError, the FnExecuteWindow has a builtin logic to deal with it, so just call the
# process_parameter_error() method and let the window do the job.
#
# - for other exceptions, we need handle it by ourselves. Here we choose to show the exception message with a
# message box to the user.
try:
params: Dict[str, Any] = window.get_parameter_values()
except ParameterError as e:
window.process_parameter_error(e)
return
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to get the parameters: "
)
return
# Step2: serialize the parameter values and save them to a json file
#
# In this example, because we don't use any complex types, we can use simply json.dump() to do the serialization.
# However, If your function contains parameters of complex types, such as list, tuple, set, dict, enum, then
# serialization and deserialization must be considered very carefully.
#
save_file = filedialog.get_save_file(
window, "Save Parameters", filters="JSON files(*.json)"
)
if not save_file:
return
try:
with open(save_file, "w") as f:
json.dump(params, f)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to save the parameters: "
)
else:
messagebox.show_info_message(window, "Parameters have been saved!")
def on_load_params(window: FnExecuteWindow, _: Action):
# Step 1: load the parameter values from a json file
file = filedialog.get_open_file(
window, "Load Parameters", filters="JSON files(*.json)"
)
if not file:
return
try:
with open(file, "r") as f:
params: Dict[str, Any] = json.load(f)
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to load the parameters: "
)
return
if not isinstance(params, dict):
messagebox.show_critical_message(window, message="Invalid parameters format!")
return
# Step2: set the parameter values to the widgets
try:
window.set_parameter_values(params)
except ParameterError as e:
window.process_parameter_error(e)
return
except Exception as e:
messagebox.show_exception_messagebox(
window, e, message="Unable to set the parameters: "
)
else:
messagebox.show_info_message(window, "Parameters have been loaded!")
if __name__ == "__main__":
action_save_params = Action(
text="Save Parameters",
icon="fa.save",
shortcut="Ctrl+S",
on_triggered=on_save_params,
)
action_load_params = Action(
text="Load Parameters",
icon="fa.folder-open",
shortcut="Ctrl+L",
on_triggered=on_load_params,
)
menu = Menu(
title="File",
actions=[action_save_params, action_load_params],
)
adapter = GUIAdapter()
adapter.add(simple_load_save_example, window_menus=[menu])
adapter.run()
| 4,378 | Python | .py | 121 | 29.958678 | 120 | 0.672011 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,395 | fn_select_window_config_example.py | zimolab_PyGUIAdapter/examples/fn_select_window_config_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.windows import DocumentBrowserConfig
from pyguiadapter.windows.fnselect import FnSelectWindowConfig
def fn1():
"""
This example shows how config the **function select window**
"""
pass
def fn2():
"""
This example shows how config the **function select window**
"""
pass
def fn3():
"""
This example shows how config the **function select window**
"""
pass
def fn4():
"""
This example shows how config the **function select window**
"""
pass
if __name__ == "__main__":
select_window_config = FnSelectWindowConfig(
title="My Tool Kit",
icon="fa5s.tools",
default_fn_group_name="Group 1",
default_fn_group_icon="fa.desktop",
fn_group_icons={
"Group 2": "fa.mobile",
"Group 3": "fa.cloud",
},
size=(600, 400),
icon_size=32,
icon_mode=True,
select_button_text="Go!",
document_browser_width=400,
document_browser_config=DocumentBrowserConfig(),
always_on_top=True,
)
adapter = GUIAdapter()
adapter.add(fn1)
adapter.add(fn2)
adapter.add(fn3, group="Group 2")
adapter.add(fn4, group="Group 3")
adapter.run(select_window_config=select_window_config)
| 1,338 | Python | .py | 47 | 22.446809 | 64 | 0.633307 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,396 | udialog_custom_messagebox_example.py | zimolab_PyGUIAdapter/examples/udialog_custom_messagebox_example.py | from datetime import date, datetime
from typing import Any
from uuid import uuid1
from qtpy.QtWidgets import QWidget, QVBoxLayout, QLabel, QDialogButtonBox
from pyguiadapter.adapter import GUIAdapter, BaseCustomDialog
from pyguiadapter.adapter import udialog
from pyguiadapter.adapter.uoutput import uprint
from pyguiadapter.exceptions import ParameterError
class UserInfoDialog(BaseCustomDialog):
def __init__(
self,
parent: QWidget,
username: str,
nickname: str,
user_id: str,
birthdate: date,
join_time: datetime,
**kwargs,
):
super().__init__(parent, **kwargs)
self.setWindowTitle("Confirm")
self._confirmed = False
self._user_info = {
"username": username,
"nickname": nickname,
"user_id": user_id,
"birthdate": birthdate,
"join_time": join_time,
}
self._button_box = QDialogButtonBox(self)
self._button_box.setStandardButtons(
QDialogButtonBox.Ok | QDialogButtonBox.Cancel
)
self._button_box.accepted.connect(self._on_accepted)
self._button_box.rejected.connect(self._on_rejected)
layout = QVBoxLayout()
username_label = QLabel(self)
username_label.setText(f"username: {username}")
layout.addWidget(username_label)
nickname_label = QLabel(self)
nickname_label.setText(f"nickname: {nickname}")
layout.addWidget(nickname_label)
user_id_label = QLabel(self)
user_id_label.setText(f"user_id: {user_id}")
layout.addWidget(user_id_label)
birthdate_label = QLabel(self)
birthdate_label.setText(f"birthdate: {birthdate}")
layout.addWidget(birthdate_label)
join_time_label = QLabel(self)
join_time_label.setText(f"join_time: {join_time}")
layout.addWidget(join_time_label)
layout.addWidget(self._button_box)
self.setLayout(layout)
def get_result(self) -> Any:
if self._confirmed:
return self._user_info
return None
def _on_accepted(self) -> None:
self._confirmed = True
self.accept()
def _on_rejected(self) -> None:
self._confirmed = False
self.reject()
def add_user_example(
username: str,
nickname: str,
user_id: str,
birth_date: date,
join_time: datetime,
):
if not username:
raise ParameterError("username", "username is empty")
if not user_id:
udialog.show_warning_messagebox(
"user_id is empty, a random one will be generated!"
)
user_id = uuid1().hex
result = udialog.show_custom_dialog(
UserInfoDialog,
username=username,
nickname=nickname,
user_id=user_id,
birthdate=birth_date,
join_time=join_time,
)
if result is not None:
udialog.show_info_messagebox(f"user added!")
uprint(result)
return
udialog.show_info_messagebox(f"user not added!")
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(add_user_example)
adapter.run()
| 3,170 | Python | .py | 94 | 26.031915 | 73 | 0.636809 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,397 | process_integers.py | zimolab_PyGUIAdapter/examples/process_integers.py | from typing import List, Literal
from pyguiadapter.adapter import GUIAdapter
def process_integers(
integers: List[int], operation: Literal["sum", "max"] = "max"
) -> int:
"""
Process some integers.
@param integers: integers for the accumulator
@param operation: sum the integers (default: find the max)
"""
func = max if operation == "max" else sum
return func(integers)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(process_integers)
adapter.run()
| 514 | Python | .py | 16 | 28.0625 | 65 | 0.68357 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,398 | print_image_example.py | zimolab_PyGUIAdapter/examples/print_image_example.py | from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.adapter.uoutput import print_image
from pyguiadapter.extend_types import file_t
def print_image_example(
image_path: file_t,
image_type: str = "jpeg",
width: int = None,
height: int = None,
embed_base64: bool = True,
centered: bool = True,
):
print_image(
image_path,
image_type,
width=width,
height=height,
embed_base64=embed_base64,
centered=centered,
)
if __name__ == "__main__":
adapter = GUIAdapter()
adapter.add(print_image_example)
adapter.run()
| 611 | Python | .py | 23 | 21.391304 | 52 | 0.659247 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
2,288,399 | dock_operation_example.py | zimolab_PyGUIAdapter/examples/dock_operation_example.py | from pyguiadapter.action import Action, Separator
from pyguiadapter.adapter import GUIAdapter
from pyguiadapter.menu import Menu
from pyguiadapter.windows.fnexec import (
FnExecuteWindow,
BottomDockWidgetArea,
)
def dock_operation_example() -> None:
pass
def on_toggle_document_dock(win: FnExecuteWindow, action: Action):
win.set_document_dock_property(visible=not win.is_document_dock_visible())
def on_toggle_output_dock(win: FnExecuteWindow, action: Action):
win.set_output_dock_property(visible=not win.is_output_dock_visible())
def on_tabify_docks(win: FnExecuteWindow, action: Action):
win.tabify_docks()
def on_move_output_area(win: FnExecuteWindow, action: Action):
if win.is_output_dock_floating():
win.set_output_dock_property(floating=False)
win.set_output_dock_property(area=BottomDockWidgetArea)
def on_float_output_dock(win: FnExecuteWindow, action: Action):
win.set_output_dock_property(floating=True)
def main():
action_document_dock = Action(
text="Toggle Document Dock",
on_triggered=on_toggle_document_dock,
)
action_output_dock = Action(
text="Toggle Output Dock",
on_triggered=on_toggle_output_dock,
)
action_tabify_docks = Action(
text="Tabify Docks",
on_triggered=on_tabify_docks,
)
action_move_output_area = Action(
text="Move Output Area",
on_triggered=on_move_output_area,
)
action_float_output_dock = Action(
text="Float Output Dock",
on_triggered=on_float_output_dock,
)
menu_views = Menu(
title="Views",
actions=[
action_document_dock,
action_output_dock,
Separator(),
action_tabify_docks,
action_move_output_area,
action_float_output_dock,
],
)
##########
adapter = GUIAdapter()
adapter.add(dock_operation_example, window_menus=[menu_views])
adapter.run()
if __name__ == "__main__":
main()
| 2,024 | Python | .py | 59 | 28.152542 | 78 | 0.678297 | zimolab/PyGUIAdapter | 8 | 0 | 0 | GPL-3.0 | 9/5/2024, 10:48:26 PM (Europe/Amsterdam) |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.