Datasets:

---

akumoth

/

peewee-issues

like 0

[]

[]

[ "That's a very nice API. Depending on the class name (a string) is something I try to shy away from, as errors can occur despite the code appearing to be perfectly valid, for instance a typo or you happen to change the model name in one place and not the other.\n\nI like the API, though, and agree that it would be worth supporting if done in a backwards-compatible way. Thanks for the patch, looking at it now.\n", "Awesome. Thanks, I've merged your patch, really appreciate you taking the time to fix this one and implement such a clean patch.\n" ]

[]

[]

[ "Thanks for including the traceback. Looks like the exception itself got cut off. Could you please include that? Will make debugging a bit easier... I'm just guessing here, but is psycopg2 already providing a uuid and peewee is treating it like it's a string?\n", "All code with all traceback:\n\n```\n>>> from application.db.models import Department, User\n>>> from application.db.models import db\n>>> from peewee import fn\n>>> k = Department.select(Department.prefix, fn.COUNT(User.id).alias('total_users')).join(User).group_by(Department.prefix).dicts()\n>>> for t in k:\n... print t\n... \nTraceback (most recent call last):\n File \"<input>\", line 1, in <module>\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 2047, in next\n obj = self.qrw.iterate()\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 2107, in iterate\n return self.process_row(row)\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 2196, in process_row\n res[column] = func(row[i])\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 1105, in python_value\n return None if value is None else uuid.UUID(value)\n File \"/usr/lib/python2.7/uuid.py\", line 133, in __init__\n hex = hex.replace('urn:', '').replace('uuid:', '')\nAttributeError: 'long' object has no attribute 'replace'\n```\n", "Oh of course! Sorry, yes. For now you can work around this by calling:\n\n``` python\nfn.COUNT(User.id).coerce(False).alias('total_users')\n```\n\nBut I am working up a patch to fix this.\n", "Fixed in b84ca0c8bc724b8cccd0b5bbd557abc04885c478\n" ]

[]

[]

[ "Is this in reference to a particular issue?\n", "**Amazing！**\n", "This patch seems more suitable for a subclass of BooleanField that is designed to handle things like strings or untrusted data.\n" ]

[]

[]

[ "Actually, this functionality does exist, but is currently marked as deprecated and no longer documented. You can view the docs [on the 2.0.2 build](http://docs.peewee-orm.com/en/2.0.2/peewee/querying.html#SelectQuery.filter), however.\n\nIf you use the alternative `filter()` API, you can use Django-style kwargs lookups:\n\n``` python\n# Works pretty much like you expect:\nactive = User.filter(active=True)\nsome_users = active.filter(username__in=('charlie', 'huey', 'zaizee'))\n\n# You can do \"OR\" using `DQ`:\neditors = User.filter(DQ(is_admin=True) | DQ(is_staff=True))\n```\n" ]

[]

[]

[ "It would be much better! If I recall correctly, you can also use psycopg2.mogrify to get the final query. Unfortunately pysqlite doesnt implement this:\n- http://bugs.python.org/issue9506\n- https://github.com/ghaering/pysqlite/issues/20#issuecomment-26258837\n\nLooks like mysql-python doesn't implement this either.\n\nSo, because it's just postgres, I think I'll favor lowest common denominator here.\n" ]

[]

[]

[ "For that case I'd suggest using the [execution context helper](http://docs.peewee-orm.com/en/latest/peewee/api.html#ExecutionContext).\n\nHard to say more, though. So you're using the \"ReadSlaveModel\"...may I ask what for? Do you have a large number of read replicas and a lot of database load?\n", "No. I just want to figure out whether some operation done when execute a transaction. Maybe when in a transaction, the ReadSlaveModel should return the master connection.\n", "I'm not sure I follow, can you explain giving me a concrete example?\n", "We did not user orm before. In our system, we have two mysql user, db_reader and db_writer. We use db_reader to execute read-only statement, and db_writer execute other statement. When use peewee, I think the ReadSlaveModel can solve this situation. \n", "@hzweveryday -- thanks for the explanation of your database setup. Hmm... I know the `ReadSlaveModel` option sounds like the right solution, but I'm not sure how to go about ensuring the proper isolation between transactions/connections. This comes down to my lack of familiarity with the ins-and-outs of MySQL..\n\nI'm not actually sure what benefits having two particular database users provides you, if your application is freely using both the reader and the writer. If they are both connected to the same database, perhaps you could just use the db_writer exclusively?\n\nI'd suggest that you go ahead and try using the `ReadSlaveModel` if you plan to maintain a strict separation between the two users/connections. Otherwise, I'd suggest using the connection pool with the db_writer user.\n", "Thank you! Now I use PooledMySQLDatabase. This should be a good solution.\n" ]

[]

[]

[ "Here's the plan for the fix:\n\n> If the user selects the foreign key field and it is null, then it's a safe bet to skip object creation. Same, as you mentioned, if they select the related object's primary key. Beyond that it gets murky. I'll look into implementing this.\n\n**Edit: fix has been merged**\n" ]

[]

[]

[ "That was _fast_! How do you do it? :smiley: \n", "Also, thanks!\n" ]

[]

[]

[ "I came up with the following hacks/workarounds:\n\n``` python\n q = Task.select(Task.machine,\n fn.COUNT(Task.id).alias('total'),\n fn.COUNT(SQL('t1.id) FILTER (WHERE t1.status = %s', 'running')).alias('running')) \\\n .group_by(Task.machine)\n\n```\n\n``` python\n q = Task.select(Task.machine,\n fn.COUNT(Task.machine).alias('total'),\n fn.COUNT(SQL('CASE WHEN t1.status = %s THEN t1.id END', 'running')).alias('running')) \\\n .group_by(Task.machine)\n```\n", "There is support for `Case`: http://docs.peewee-orm.com/en/latest/peewee/playhouse.html#case\n\nFor other arbitrary ones, you can do:\n\n``` python\nClause(\n fn.COUNT(Task.id),\n fn.FILTER(Clause(SQL('WHERE'), Task.status == 'running'))\n)\n```\n" ]

[]

[]

[ "Your implementation is better. :+1: \n\nThank you :smiley:\n" ]

[]

[]

[]

[]

[]

[]

[]

[]

[ "You need a trailing comma after (('article_id', 'content_md5'), False) -- otherwise, since it's the only item in the outer-most parentheses, Python treats it as a normal expression rather than a tuple.\n" ]

[]

[]

[ "This will require a bit more work.\nOn PyPy we can use psycopg2-cffi.\nLet me clone that branch. The build will pass soon.\n", "I think pip is being silly here. Lemme check.\n", "Now it skips the damm Cython generation step. I'll get back to you on that.\n", "Closing for now, if you plan to re-open would you mind rebasing the existing commits into a single commit?\n", "I wouldn't. It doesn't work yet though.\n", "I mean I wouldn't mind rebasing...\n", "Cool, I guess I should ask: is this something you plan on continuing to hack on until it's all working? If so be happy to reopen and you can rebase when it's closer to being done.\n" ]

[]

[]

[ "What's going on in this PR? I'm not sure I understand the differences here.\n", "It caches the compilation of the C extensions and also doesn't download pip dependencies.\nThe build as a result should be faster.\n", "Just looking at the build times it doesn't appear to have made a difference.\n", "Did a quick skim of the travis cache docs, I'm going to pass on this one.\n" ]

[]

[]

[ "This is a nice PR, very clean, but the Django-style querying is deprecated and I do not plan on adding new functionality as it will be removed soon in the next major version...which hopefully isn't too far away!\n" ]

[]

[]

[ "I'm not able to replicate this locally. Here's the code I'm using:\n\n``` python\nimport os\nfrom peewee import *\n\nif os.path.exists('/tmp/x.db'):\n os.unlink('/tmp/x.db')\n\ndb = SqliteDatabase('/tmp/x.db')\n\nclass Ticket(Model):\n number = CharField(unique=True)\n modified = DateTimeField()\n subject = CharField()\n queue = CharField()\n assignee = CharField(null=True)\n status = CharField()\n ismodified = BooleanField(default=False)\n newtime = DateTimeField(null=True)\n #account_name = CharField(default='')\n #account_number = CharField(default='')\n class Meta:\n database = db\n\nTicket.create_table()\n\ndef main():\n from playhouse.migrate import SqliteMigrator, migrate\n migrator = SqliteMigrator(db)\n account_name_field = CharField(default='')\n account_number_field = CharField(default='')\n with db.transaction():\n migrate(\n migrator.add_column('Ticket', 'account_name', account_name_field),\n migrator.add_column('Ticket', 'account_number', account_number_field),\n )\n\nif __name__ == '__main__':\n main()\n```\n\nIf I just import the module, then the DB is created and has the following schema:\n\n``` sql\n-- ran .fullschema --indent in the sqlite console\nCREATE TABLE \"ticket\"(\n \"id\" INTEGER NOT NULL PRIMARY KEY,\n \"number\" VARCHAR(255) NOT NULL,\n \"modified\" DATETIME NOT NULL,\n \"subject\" VARCHAR(255) NOT NULL,\n \"queue\" VARCHAR(255) NOT NULL,\n \"assignee\" VARCHAR(255),\n \"status\" VARCHAR(255) NOT NULL,\n \"ismodified\" INTEGER NOT NULL,\n \"newtime\" DATETIME\n);\nCREATE UNIQUE INDEX \"ticket_number\" ON \"ticket\"(\"number\");\n```\n\nAfter running the migration, I have the following schema:\n\n``` sql\nCREATE TABLE \"ticket\"(\n \"id\" INTEGER NOT NULL PRIMARY KEY,\n \"number\" VARCHAR(255) NOT NULL,\n \"modified\" DATETIME NOT NULL,\n \"subject\" VARCHAR(255) NOT NULL,\n \"queue\" VARCHAR(255) NOT NULL,\n \"assignee\" VARCHAR(255),\n \"status\" VARCHAR(255) NOT NULL,\n \"ismodified\" INTEGER NOT NULL,\n \"newtime\" DATETIME,\n \"account_name\" VARCHAR(255) NOT NULL,\n \"account_number\" VARCHAR(255) NOT NULL\n);\nCREATE UNIQUE INDEX \"ticket_number\" ON \"ticket\"(\"number\");\n```\n\nBased on this quick test it seems to be working correctly. Using peewee@49eeec7c14be38f0ba21b428e94f716c45009230 with Python 2.7.12 and SQLite 3.14.\n", "Can you provide any additional info, or try my script out on your own machine?\n", "Please re-open if there's something I've missed or if the above code that I shared reproduces the issue on your computer.\n" ]

[]

[]

[ "What is the use-case for this?\n" ]

[]

[]

[ "Not enough info.\n" ]

[]

[]

[ "[From one of their devs](https://github.com/cockroachdb/cockroachdb-python/issues/21#issuecomment-229755195),\n\n> Cockroach actually supports the Postgres wire protocol, so the Postgres driver builtin to Peewee should work (we support psycopg2 without issue). That said, we have found that a few drivers use some Postgres-specific SQL extensions (for instance, use of pg_catalog instead of the standard information_schema), so current levels of support for them vary.\n\nThis means that peewee, configured with a modified version of the `PostgresqlDatabase`, should be able to work with Cockroach -- at least in some sense.\n\nGoing to close this out as I do not plan on implementing support myself.\n\nThose interested can also follow along here:\n\nhttps://github.com/cockroachdb/cockroachdb-python/issues/21\n" ]

[]

[]

[ "Hey @sin5 - can you please take the 5 extra seconds and learn how to format your code? Did you even look at your issue after creating it?\n", "Please read the documentation on submitting bugs: http://docs.peewee-orm.com/en/latest/peewee/contributing.html#bugs\n\nHere's a link to GitHub's documentation:\n\nhttps://help.github.com/articles/getting-started-with-writing-and-formatting-on-github/\n", "Specifically, though, and this is a bad API -- I will be addressing it in a subsequent release -- calling `database.create_table` will _only_ create the table. Calling `Model.create_table()`, on the other hand, will create the table and the indexes.\n\nHowever,\n\nTo create multiple tables, and their indexes, with the dependency graph resolved to ensure the tables are created in the proper order, use the `Database.create_tables()` method.\n\nIn fact, I pretty much just recommend that people **always** use `database.create_tables([AModel, AnotherModel])`.\n", "Also read the big fat warning in the docs:\n\nhttp://docs.peewee-orm.com/en/latest/peewee/api.html#Database.create_table\n", "```\n sorry for my code format. and thanks for your answer：）\n```\n\n—You are receiving this because you were mentioned.Reply to this email directly, view it on GitHub, or mute the thread.\n", "No probs!\n" ]

[]

[]

[ "How does this issue manifest itself?\n", "For example, I have a log table, and is split by date table, every time you insert new data, checks the table exists, then call `create_table` methods, parameters `fail_silently = True`, it does not exist is automatically created to perform the `execute_sql` method.\n\nthe emergence of a large number of `show tables; commit;` \n`show tables` operation is not whether it is necessary to `commit`?\n", "See #727, #806, #814 for why Peewee tends to issue commits with MySQL. In general I've found that users of the database are confused when it appears results are \"cached\" or they're not seeing changes. This happens when they have a long running process, for instance, and are issues reads with no commits.\n\nIf you would like to avoid the extra `COMMIT` statements, I'd suggest wrapping the code in a transaction using:\n\n``` python\nwith db.atomic():\n # your code here.\n```\n\nAlso, I don't know anything about your use case but it seems incredibly inefficient to be checking for the table existence before every insert statement when the data being inserted is log records. Just my .02 but that seems problematic.\n" ]

[]

[]

[ "Please read the documentation on submitting bugs: http://docs.peewee-orm.com/en/latest/peewee/contributing.html#bugs\n\nSpecifically, take the extra 5 seconds and please format your code:\n\nhttps://help.github.com/articles/getting-started-with-writing-and-formatting-on-github/\n" ]

[]

[]

[ "Cool, thanks!\n" ]

[]

[]

[]

[]

[]

[ "I would appreciate it if you respected my request that you not post things on my project.\n\nYou have your own fork, please use it.\n", "i'm sorry, but you don't own github. i'm allowed to discuss this project with other users/devs (just like anyone else). you don't have to be involved in that discussion.\n", "Derek, you seem to misunderstand free speech or your rights on GitHub.\n\nWhile I might not agree with Charles, I definitely think it is his right to\nask you not to post here.\nOn Jun 17, 2016 7:41 PM, \"Derek Anderson\" [email protected] wrote:\n\n> that is patently false. you've definitely deleted comments before issues\n> were closed. and you've not deleted other's comments after issues were\n> closed where you've deleted mine. (even when my comment was just \"i agree\n> w/ person XYZ\"). i don't even know how many, as GH doesn't notify me when\n> it happens, nor do i have any record of them (other than me luckily having\n> left the browser open for that one screenshot)\n> \n> besides, even the premise for your deletion is invalid. where is there a\n> rule of \"you can't comment after an issue is closed\"? where else would one\n> ever say \"IMO this is still and issue and should be re-opened, because\n> reason XYZ\"?\n> \n> refrain from commenting, creating tickets or posting here.\n> \n> sorry, but no. if i have something to say that's respectful and on topic,\n> i intend to reserve the right to say it. freedom of speech and whatnot.\n> \n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub\n> https://github.com/coleifer/peewee/issues/993#issuecomment-226819371,\n> or mute the thread\n> https://github.com/notifications/unsubscribe/AAEXLPEfY81fTY9jkYow2gQrxNrkJD2vks5qMs4rgaJpZM4I4daH\n> .\n", "What am I supposed to think, @keredson . You are a toxic presence on my project, and I don't want you to keep harassing me.\n", "Why can't you respect that, one person to another?\n" ]

[]

[]

[]

[]

[]

[]

[]

[]

[ "That's very odd. So peewee is expecting to get back the primary key, and since Postgres supports the \"RETURNING\" clause, peewee says \"INSERT INTO my_table (cola, colb, colc) VALUES (vala, valb, valc) RETURNING primary_key\".\n\nFor whatever reason, it appears that the database cursor is not returning anything.\n\nCan you share the SQL that is generated by Peewee when it runs that above insert? You should be able to do:\n\n```\nSOOSItem.insert(**data).sql()\n```\n\nAny other ideas of what might be going on here?\n", "Closing due to lack of response.\n", "Please comment here and I'll reopen.\n", "Just saw this happen when adding a trigger on a table changed the semantics of INSERT statements (inserts were not returning anything then). Dropping the trigger resolved the problem.", "Aha! Many thanks @uzak !" ]

[]

[]

[ "I think the \"first version\" you referred to is missing from the comment. I see your table definition and that calling `db.create_indexes()` creates the indexes, but not the part that may not have been working?\n", "Sorry, nevermind.\nAfter having unrelatedly restarted postgres, it works.\nWhat I meant was, that calling `create_tables()` only created the table, but not the index for the `name` field. I had to call `create_indexes()` separately.\n", "Cool, glad to hear it's working correctly! Thanks for the update.\n" ]

[]

[]

[ "I've committed a fix for this, thanks for bringing it to my attention!\n" ]

[]

[]

[ "Aha!\n\nIt seems like we can use `alias()` to resolve \"conflicts\":\n\n``` python\nuser_a = User.alias()\nuser_b = User.alias()\n\nquery = (Match.select(Match, user_a, user_b)\n .join(user_a, on=Match.user_a)\n .switch(Match)\n .join(user_b, on=Match.user_b)\n .where((Match.user_a == user) | (Match.user_b == user)))\n```\n\nWorked for me, but I'm not sure, whether this is intended usage of `alias()` :)\n", "Yep, that's what the `alias()` helper is for -- allowing multiple table references in a single query .\n" ]

[]

[]

[ "Originally this is a design decision I borrowed from Django's ORM. I believe the idea is to that the API should reinforce the concept that the query result wrapper is not a list, and should not be treated as such. That said, I can see how this might cause confusion since the other `__getitem__` behaviors are available. I'll address this with a patch, using yours as a reference. Thanks!\n\nDiscussion on django issue tracker: https://code.djangoproject.com/ticket/13089\n", "Thanks again @HughMacdonald\n", "Cool - thanks for that.\n" ]

[]

[]

[ "So this error indicates that the `mysql` Python driver is not installed or is not import-able. Here is the code peewee executes at the top of the module:\n\n``` python\ntry:\n import MySQLdb as mysql # prefer the C module.\nexcept ImportError:\n try:\n import pymysql as mysql\n except ImportError:\n mysql = None\n```\n\nYou've triggered the \"mysql=None\" bit, meaning peewee could not import either driver.\n" ]

[]

[]

[ "q = MsgLog.select().where(MsgLog.master_msg_id.startswith(\"%s.\" % chat_id)).order_by(MsgLog.id.desc())\n\nThe problem is you're ordering by MsgLog.id but MsgLog.master_msg_id is the primary key and the model (while it has an \"id\" attribute), does not consider it a valid field.\n\nTo fix this, just change your \"order_by()\" call to order on a field that's defined on the MsgLog model. Or alternatively, you could make \"master_msg_id\" have \"unique=True\" instead of being a primary key.\n", "I think I may make some modifications to Peewee to drop the `id` attribute if a different field is used as the foreign key.\n", "@coleifer deleting other user's comments just because you don't agree with the technical viewpoint is very uncool.\n", "The truth is that I removed your comments because they were not constructive, and I was tired of having to defend aspects of my project because you felt entitled to a technical discussion because you happened to disagree with it. When I pointed to you that the things you disagreed with were documented and obviously intentional (via git log), you suggested I was trying to justify, post-hoc, bad design decisions, and was being an irresponsible maintainer. How is that constructive?\n", "Reopening as I intend to fix. \n" ]

[]

[]

[ "I'm kind of curious what kind of model structure you'd have to have to run into this kind of thing... I understand that there may be a bug here, but I'm not sure how you would actually encounter it in practice?\n", "I run into it not only once, last time we had save parameters from external API call in the model Inherited from the base model. In this case, the log of actions associated with a subscription entity.\nOur deveveloper did'nt know the _id thing and just save it like that:\n\nif v.validate(self.message):\n FortumoLog.create(\n subscription = self.subscription\n **self.message)\n\nSo about the stucture: we have several log entities for different APIs and one general subscription model.\n\nNobody has noticed that one of the parameters is not stored (it just stored for history), but sometimes it drove to crashes our flask-admin, only on the prod with disabled debugging.) with some error \n\"Failed to find field for filter: %s\"\nSo I think it may save a few hours for someone.\n", "Fixed 34e56cb887751453680641695bdec1fc6518b2cf\n" ]

[]

[]

[ ":+1: thanks!\n" ]

[]

[]

[ "so i have a feeling that i'm going to get a response along the lines of \"we preload all the data in case the user wants to access/use it later\", similar to the rationale given for why `first()` doesn't use a `limit 1`. i just would like to point out that both of these are directly counter to the stated \"peewee doesn't do things based on what it 'thinks' you might want\" policy.\n", "> \"peewee doesn't do things based on what it 'thinks' you might want\"\n\nActually I think that current behavior is in line with this policy. If I wanted a database count query, I would've called the `count()` method instead of using `len`. \n", "It actually was, and has been documented, for some time:\n\n45e096e94afb3c4496ac961f64ada374d27b2529\n" ]

[]

[]

[ "That all would depend on your databases default isolation level. Also recall please how, as a general rule, peewee doesn't use magic or do things based on what it \"thinks\" you might want. You have the tools to do your prefetching in an atomic block. If you always want this behavior, write a wrapper.\n", "> That all would depend on your databases default isolation level.\n\ni don't think that's true. two subsequent queries outside of a transaction don't have any consistency guarantees between them at any isolation level on any database.\n\n> peewee doesn't use magic or do things based on what it \"thinks\" you might want\n\nall APIs are doing what they think you want to a certain degree, in the sense that they have reasonable defaults and behavior. anyone who makes a single API call to a peewee method would expect it to return an internally consistent collection of objects. i shouldn't have to know / worry about internal implementation details in order to get correct data, vs. this obvious race condition. \"it's usually right\" arguments aside.\n", "> i shouldn't have to know / worry about internal implementation details in order to get correct data\n\nIn that case you should wrap your code in an `atomic()` block.\n", "i get that that's an option. i'm arguing that that should be the default because as is the results are only \"usually\" right. :)\n" ]

[]

[]

[ "Why put an alias on the column when specifying the `on` column? Seems like a pretty contrived example unless I'm misunderstanding how this applies to a larger class of bugs?\n", "It's only contrived because I simplified it to the smallest possible example for the purpose of the bug report. I hit it by accident in a larger query. I worked around it by not using the alias.\n", "Well, what's even the point of putting an aliased column in the on parameter?\n", "I don't know that there is a point to it here in this simplified example. But clearly it should not be doing what it's doing. (Surprising the user with O(n) queries.) Probably it should effectively ignore the alias, or second to that throw an exception saying you can't used aliased FKs in a join.\n", "I should clarify: when you put an alias on the `on` column (either an expression or a field), that is treated as an instruction to Peewee to use that alias for the joined instance when reconstructing the model graph.\n\nSo, when you write `on=Article.author.alias('something')` you're clearly instructing Peewee that you want the Article's author to be accessible as \"article.something\".\n\nIf you go back and re-check your code, you'll see that accessing it does not result in a query:\n\n``` python\nx = Article.select(Article, Person).join(Person, on=Article.author.alias('something')).first()\nprint x.something.name\n```\n", "hmmm... ok, i can get behind that. +1\n", "I'm relieved. Thanks for the approval.\n" ]

[]

[]

[]

[]

[]

[ "> Maybe we could toss in a extend=False kwarg to the method that\n\ni can see the use, but a definite +1 on not changing the default behavior if this is done.\n", "Of course. Default behavior is critical, especially when considering that this is likely a heavily-used feature of the ORM.\n" ]

[]

[]

[ "There's some chicanery going on with Peewee in regards to implementing the `rowid` field. I'll take another look at the implementation and see if I can clean it up.\n", "Thanks! Yes, the field gets removed here: \nhttps://github.com/coleifer/peewee/blob/master/playhouse/sqlite_ext.py#L90\n\nDo extension tests not get run by the CI? In theory the build should have failed. (I had to add `from playhouse.tests.test_sqlite_ext import *` to `tests.py` get them to run locally).\n", "> Do extension tests not get run by the CI?\n\nThey should be getting run. The .travis.yml file specifies `./runtests.py -a` which includes all extensions (except those for which a dependency is missing, which shouldn't be the case for the sqlite extension tests).\n", "Yes, looks like they ran there and passed. I suspect it's because some code paths don't get executed on the CI, for instance:\n\n``` python\n insert_many = sqlite3 and sqlite3.sqlite_version_info >= (3, 7, 11, 0)\n # ...\n def execute(self):\n insert_with_loop = (\n self._is_multi_row_insert and\n self._query is None and\n self._returning is None and\n not self.database.insert_many)\n if insert_with_loop:\n return self._insert_with_loop()\n```\n\nAnyway, with the lastest master everything works as expected and I can remove my workaround, thanks!\n" ]

[]

[]

[ "Oof...this implementation feels very fragile and hacky, and I don't think the extra lines of code are worth the feature of being able to specify an alias name.\n\nGoing to pass on this for now. The 3.0 stuff I'm working on can do this as a matter of course, so it'll land then.\n", "+1 on it for the rewrite. including the alias equality?\n" ]

[]

[]

[ "In rewriting the SQL generation, I've got this feature covered. The implementation of this in peewee 2.8 just feels a bit too hacky to me to want to merge.\n" ]

[]

[]

[ "I came across this while trying to work around another bug (insert_many + FTS5). Will follow up with another PR.\n", "Awesome catch, thank you!\n" ]

[]

[]

[]

[]

[]

[ "Thanks for sharing the changes. Closing as wontfix per earlier discussion.\n" ]

[]

[]

[]

[]

[]

[ "force_insert=True\n" ]

[]

[]

[ "force_insert=True\nThe logging helped a ton--didn't find that till I looked at https://github.com/coleifer/peewee/issues/961 but super helpful.\n\nThx.\n" ]

[]

[]

[ "Aliased models are frequently used with correlated subqueries, making the possibility of inferring the correct thing to do quite a bit more complicated than it might seem.\n\nPhilosophically, I'm opposed to adding _any_ heuristics into `peewee`. Peewee is incredibly easy to understand because it does what you tell it to do. For people who are comfortable writing SQL this is a big deal because there are no surprises in store when you go to try out your query. You don't need to know a bunch of specialized APIs, because peewee provides the right simple tools that you can use them to build up more complex tools.\n", "Take a look at this post I wrote to get a longer take on these ideas: http://charlesleifer.com/blog/shortcomings-in-the-django-orm-and-a-look-at-peewee-a-lightweight-alternative/\n", "yep, read that about a week and a half ago. one of the reasons i gave peewee a real hard look, and i very much agree w/ it. i did a lot of django work ~2006-2008 and know well how glitchy their ORM can be. (tho to be fair, then it was light years ahead of anything else, when everything was full on \"don't worry about the database, just pretend it's a giant in memory object graph\" hell.) then (when thrown back into an all-java world), i did internally for my company just what you did in 2.0. pure SQL constructs w/ an immutable builder API (all java).\n\nanyway, not meaning to life story here.\n\n> Philosophically, I'm opposed to adding any heuristics into peewee.\n\nagree that's a good philosophically. \n\nbtw, i found having dealing w/ the alias objects outside the scope of the query a significant source of user complaints. i tried a couple different techniques to ease the pain. this was one. another was having the aliases, but not having to be tied to a specific object instance. like:\n\n``` java\nTableA.ALL.join(TableB.as(\"x\")).where(TableB.as(\"x\").COLUMN_Y.eq(\"whatever\"));\n```\n\n(the `ALL` for me is like your `select()`)\n\nnote the first `as(\"x\")` is returning effectively what you call a model alias, where the second is (after the column is referenced) returning a bound field. this way people weren't having to declare an external variable outside their query. (granted, this declaration was more of a PITA in java)\n\nanyway... :)\n", "Thanks for the input. Although the whole way peewee does model aliases currently is a bit of a lucky hack (see c118e434), I like using descriptive variable names for my table aliases. I've found it much easier to read and keep track of my queries when I can think of subsequent table references as distinct row sources. Nothing wrong with `.as('alias')`, but the repetition does create an equally easy avenue for one to forget to include the alias.\n", "hah. been there.\n\nyeah, the references also worked as how yours do as well. you could also do:\n\n``` java\nTable tableb = TableB.as(\"x\");\nTableA.ALL.join(tableb).where(tableb.COLUMN_Y.eq(\"whatever\"));\n```\n\nbut it just was never used very much that way.\n" ]

[]

[]

[ "This gets into a little bit of a gray area... I'm mainly opposed to stomping on the backref descriptor. Using `related_name_prefetch` isn't the greatest solution, but it's a convention that's been established by the documentation. Adding the ability to specify an alias to use for the prefetched objects seems like the best solution, but the implementation is already pretty out of control from successive attempts at adding functionality without really re-thinking things.\n\nAnother consideration is that accessing a backref creates a new query each time, so it's not the case that a backref query will ever be cached when accessed directly:\n\n``` python\nIn [9]: note = Note.public().get()\n\nIn [10]: query1 = note.notetag_set\n\nIn [11]: query2 = note.notetag_set\n\nIn [12]: id(query1) == id(query2)\nOut[12]: False\n```\n\nThe bad part is that `aggregate_rows()` behaves differently... `aggregate_rows()` supports the ability to assign prefetched objects to an alias of your choosing, but by default it uses the `related_name` and therefore overwrites it with a `list`. Not good to have `prefetch()` do one thing and `aggregate_rows()` another, but at this point both implementations gross me out so much that I'm not too inclined to change them. Furthermore, this would probably break a lot of people's code, as `prefetch()` and `aggregate_rows()` are pretty commonly-used APIs.\n\nSo, let's table this for now. With the 3.0 release I'm planning on completely dropping the `aggregate_rows()`-type functionality as my testing has never found it to be more performant than prefetch's O(k) or the naive O(n). I'll re-think `prefetch()` to make it less of a special-case and more tightly integrated with the rest of the APIs.\n", "aren't all instance variables stomping on the table references already? the instance `post.title` is a string, while `Post.title` is a column reference. the backref seems no different.\n\n> accessing a backref creates a new query each time\n\nhmmm. i think that would make it easier. but i haven't looked at the code.\n\n> it's a convention that's been established by the documentation\n\nyeah, it'd have to be left there anyway for backwards compatibility obviously. just seeding ideas for 3.0. whatever replaces prefetch IMO should just store it in the backref pseudo-query/list rather than a secondary variable.\n", "> aren't all instance variables stomping on the table references already? the instance post.title is a string, while Post.title is a column reference. the backref seems no different.\n\nThe class attributes are different from the instance attributes. For a reverse relation, it only has meaning if we have an instance with a value in the field the related model references, so again, it's an instances-only thing.\n" ]

[]

[]

[ "I've commented already that the semantics of get() are not going to change, so I will pass on this change.\n", "3.0 man, 3.0! :-)\n", "It's not changing in 3.0 either man.\n" ]

[]

[]

[ "Weird..have you tried enabling the peewee logger and adding logging statements around your db connect/close code? That might help diagnose where things are going wrong.\n\nYou can enable the peewee debugger by following these instructions: http://docs.peewee-orm.com/en/latest/peewee/database.html#logging-queries\n\nThere's another logger in the `peewee.pool` namespace that should get picked up as well, so be sure to keep an eye out for that.\n\nTry enabling the logger and letting me know where the error occurs? Or alternatively, could you share the whole traceback?\n", "Hi @coleifer, I put a print when I call database.open and database.close and both methods are being called and logged.\n\nHere is the entire log. You can see the calls to my backend and the peewee log\n\n``` bash\nNo connection available in pool.\nCreated new connection 23380784.\n('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465147903, u'{\"request_json\": {\"phone\": \"1111111111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"1111111111\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\n('SELECT `t1`.`id`, `t1`.`name`, `t1`.`email`, `t1`.`profile`, `t1`.`cpf`, `t1`.`rg`, `t1`.`birthday`, `t1`.`ddd`, `t1`.`phone`, `t1`.`gender`, `t1`.`device_uuid`, `t1`.`city`, `t1`.`state`, `t1`.`address`, `t1`.`house_number`, `t1`.`neighborhood`, `t1`.`cep`, `t1`.`push_token` FROM `user` AS t1 WHERE (`t1`.`cpf` = %s)', [u'1111111111'])\n('SELECT `t1`.`id`, `t1`.`name`, `t1`.`email`, `t1`.`profile`, `t1`.`cpf`, `t1`.`rg`, `t1`.`birthday`, `t1`.`ddd`, `t1`.`phone`, `t1`.`gender`, `t1`.`device_uuid`, `t1`.`city`, `t1`.`state`, `t1`.`address`, `t1`.`house_number`, `t1`.`neighborhood`, `t1`.`cep`, `t1`.`push_token` FROM `user` AS t1 WHERE (`t1`.`phone` = %s)', [u'1111111111'])\n\n('SELECT `t1`.`id`, `t1`.`proposal_id`, `t1`.`account_id`, `t1`.`card_id`, `t1`.`card_number`, `t1`.`card_validation`, `t1`.`user_id`, `t1`.`has_bank_account`, `t1`.`contract_acceptance_date` FROM `customer` AS t1 WHERE (`t1`.`user_id` = %s)', [None])\n('SELECT `t1`.`id`, `t1`.`owner_id`, `t1`.`phone_invited`, `t1`.`status`, `t1`.`token` FROM `invite` AS t1 WHERE (`t1`.`phone_invited` = %s)', [u'1111111111'])\nMethod name: ORIGINA_CLIENTE\nParameters used: [\"1111111111\", \"11111111\", \"12321\", 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \"11\", 1, {\"empresa\": \"20937849000101\", \"fluxo\": \"ORIGINA_CLIENTE\", \"senha\": \"x\", \"usuario\": \"x\"}]\nParameters used: {}\n('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'ORIGINA_CLIENTE', 1465147906, u'{\"method\": \"GET\", \"custom\": {\"params\": [\"1111111111\", \"11111111\", \"12321\", 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, \"11\", 1, {\"empresa\": \"20937849000101\", \"fluxo\": \"ORIGINA_CLIENTE\", \"senha\": \"x\", \"usuario\": \"x\"}, {}], \"value\": {\"processReturn\": {\"ticket\": \"414D5120514D2E524F434B45542020201D884F57035B1C20\", \"variaveisContexto\": \"\", \"hash\": \"4380401e6d6a09706e6d08e24ebd40f6\"}}}}'])\n('INSERT INTO `user` (`profile`, `cpf`, `ddd`, `phone`, `device_uuid`) VALUES (%s, %s, %s, %s, %s)', [u'buyer', u'1111111111', u'11', u'1111111111', u'12321'])\n('INSERT INTO `rocketresource` (`user_id`, `ticket`, `hash_value`, `token`, `status`) VALUES (%s, %s, %s, %s, %s)', [10, u'414D5120514D2E524F434B45542020201D884F57035B1C20', u'4380401e6d6a09706e6d08e24ebd40f6', u'7749ba85-0d74-430a-a52b-2047177cd9ae', u'waiting_response'])\nReturning 23380784 to pool.\n\n[pid: 38|app: 0|req: 1/1] 127.0.0.1 () {48 vars in 825 bytes} [Sun Jun 5 14:31:43 2016] POST /s/user/Register => generated 155 bytes in 3162 msecs (HTTP/1.0 200) 3 headers in 112 bytes (2 switches on core 0)\n\n\n\n('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'MockService', 1465147928, u'{\"method\": \"GET\", \"custom\": {}}'])\nReturning 23380784 to pool.\n[pid: 38|app: 0|req: 2/2] 127.0.0.1 () {42 vars in 703 bytes} [Sun Jun 5 14:32:08 2016] GET /s/mock/Mock => generated 0 bytes in 18 msecs (HTTP/1.0 403) 3 headers in 110 bytes (1 switches on core 1)\n\n\n\nConnection 23380784 was closed.\nNo connection available in pool.\nCreated new connection 25864240.\n('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465149152, u'{\"request_json\": {\"phone\": \"1111111111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"1111111111\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\nReturning 25864240 to pool.\nERROR:root:(2006, 'MySQL server has gone away')\nTraceback (most recent call last):\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1535, in __call__\n rv = self.handle_exception(request, response, e)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1529, in __call__\n rv = self.router.dispatch(request, response)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1278, in default_dispatcher\n return route.handler_adapter(request, response)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1102, in __call__\n return handler.dispatch()\n File \"./handler/service_handler.py\", line 16, in inner\n return func(*args, **kwargs)\n File \"./handler/service_handler.py\", line 30, in inner\n result = fnc(*a, **k)\n File \"./handler/service_handler.py\", line 54, in dispatch\n raise e\nOperationalError: (2006, 'MySQL server has gone away')\n[pid: 38|app: 0|req: 3/3] 127.0.0.1 () {48 vars in 824 bytes} [Sun Jun 5 14:52:32 2016] POST /s/user/Register => generated 114 bytes in 11 msecs (HTTP/1.0 500) 2 headers in 100 bytes (1 switches on core 0)\n\n```\n", "Very strange... what causes that traceback to happen? It seems like it is not related to the connection pool.\n", "Hi @coleifer, thanks for helping me. I put some logs and the exception is raised when I try to call the create method with one of my models. Check it out:\n\n``` bash\nConnection 16238048 was closed.\nDEBUG:peewee.pool:Connection 16238048 was closed.\nNo connection available in pool.\nDEBUG:peewee.pool:No connection available in pool.\nCreated new connection 139681664633120.\nDEBUG:peewee.pool:Created new connection 139681664633120.\n('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465213386, u'{\"request_json\": {\"phone\": \"1111121111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"2222222222\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\nDEBUG:peewee:('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465213386, u'{\"request_json\": {\"phone\": \"1111121111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"2222222222\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\nReturning 139681664633120 to pool.\nDEBUG:peewee.pool:Returning 139681664633120 to pool.\nERROR:root:(2006, 'MySQL server has gone away')\nTraceback (most recent call last):\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1535, in __call__\n rv = self.handle_exception(request, response, e)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1529, in __call__\n rv = self.router.dispatch(request, response)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1278, in default_dispatcher\n return route.handler_adapter(request, response)\n File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1102, in __call__\n return handler.dispatch()\n File \"./handler/service_handler.py\", line 18, in inner\n return func(*args, **kwargs)\n File \"./handler/service_handler.py\", line 34, in inner\n result = fnc(*a, **k)\n File \"./handler/service_handler.py\", line 47, in dispatch\n self.__saving_log(obj)\n File \"./handler/service_handler.py\", line 75, in __saving_log\n log_helper.log(action=obj.__class__.__name__, method=self.request.method, value=result, request=req)\n File \"./helper/log_helper.py\", line 42, in log\n return Log.create(owner=user, action=action, value=log_json, date=now)\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 4551, in create\n inst.save(force_insert=True)\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 4747, in save\n pk_from_cursor = self.insert(**field_dict).execute()\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3244, in execute\n cursor = self._execute()\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 2656, in _execute\n return self.database.execute_sql(sql, params, self.require_commit)\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3492, in execute_sql\n self.commit()\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3316, in __exit__\n reraise(new_type, new_type(*exc_args), traceback)\n File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3485, in execute_sql\n cursor.execute(sql, params or ())\n File \"/usr/local/lib/python2.7/dist-packages/MySQLdb/cursors.py\", line 174, in execute\n self.errorhandler(self, exc, value)\n File \"/usr/local/lib/python2.7/dist-packages/MySQLdb/connections.py\", line 36, in defaulterrorhandler\n raise errorclass, errorvalue\nOperationalError: (2006, 'MySQL server has gone away')\n[pid: 38|app: 0|req: 4/4] 127.0.0.1 () {48 vars in 843 bytes} [Mon Jun 6 08:43:06 2016] POST /s/user/Register => generated 114 bytes in 12 msecs (HTTP/1.0 500) 2 headers in 100 bytes (1 switches on core 1)\n```\n", "Cool thanks! So hmmm...are there a set of steps you follow to trigger the\nerror or does it just seem to happen sporadically? What values are you\nusing for your MySQL server connection timeout, and what is the\nstale_timeout on the peewee pool? My thought is if it's random, then you've\njust got to turn the pool stale timeout to a value lower than mysqls\ntimeout. Another option would be to stop using the pool and see how that\ngoes. Is this db getting a lot of traffic?\nOn Jun 6, 2016 6:49 AM, \"Felipe Volpone\" [email protected] wrote:\n\n> Hi @coleifer https://github.com/coleifer, thanks for helping me. I put\n> some logs and the exception is raised when I try to call the create method\n> with one of my models. Check it out:\n> \n> Connection 16238048 was closed.\n> DEBUG:peewee.pool:Connection 16238048 was closed.\n> No connection available in pool.\n> DEBUG:peewee.pool:No connection available in pool.\n> Created new connection 139681664633120.\n> DEBUG:peewee.pool:Created new connection 139681664633120.\n> ('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465213386, u'{\"request_json\": {\"phone\": \"1111121111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"2222222222\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\n> DEBUG:peewee:('INSERT INTO `log` (`owner`, `action`, `date`, `value`) VALUES (%s, %s, %s, %s)', [None, u'RegisterService', 1465213386, u'{\"request_json\": {\"phone\": \"1111121111\", \"meta\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"type\": \"buyer\", \"cpf\": \"2222222222\"}, \"mobile_meta_info\": {\"rooted\": \"false\", \"app\": \"1.0\", \"appversion\": \"1.0\", \"Carrier\": \"Claro\", \"device_id\": \"12321\", \"OSVersion\": \"4.4\", \"gps\": \"13.03;15.033\"}, \"method\": \"POST\", \"custom\": {}}'])\n> Returning 139681664633120 to pool.\n> DEBUG:peewee.pool:Returning 139681664633120 to pool.\n> ERROR:root:(2006, 'MySQL server has gone away')\n> Traceback (most recent call last):\n> File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1535, in **call**\n> rv = self.handle_exception(request, response, e)\n> File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1529, in **call**\n> rv = self.router.dispatch(request, response)\n> File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1278, in default_dispatcher\n> return route.handler_adapter(request, response)\n> File \"/usr/local/lib/python2.7/dist-packages/webapp2.py\", line 1102, in **call**\n> return handler.dispatch()\n> File \"./handler/service_handler.py\", line 18, in inner\n> return func(_args, *_kwargs)\n> File \"./handler/service_handler.py\", line 34, in inner\n> result = fnc(_a, *_k)\n> File \"./handler/service_handler.py\", line 47, in dispatch\n> self.**saving_log(obj)\n> File \"./handler/service_handler.py\", line 75, in __saving_log\n> log_helper.log(action=obj.__class**.**name**, method=self.request.method, value=result, request=req)\n> File \"./helper/log_helper.py\", line 42, in log\n> return Log.create(owner=user, action=action, value=log_json, date=now)\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 4551, in create\n> inst.save(force_insert=True)\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 4747, in save\n> pk_from_cursor = self.insert(*_field_dict).execute()\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3244, in execute\n> cursor = self._execute()\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 2656, in _execute\n> return self.database.execute_sql(sql, params, self.require_commit)\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3492, in execute_sql\n> self.commit()\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3316, in __exit__\n> reraise(new_type, new_type(_exc_args), traceback)\n> File \"/usr/local/lib/python2.7/dist-packages/peewee.py\", line 3485, in execute_sql\n> cursor.execute(sql, params or ())\n> File \"/usr/local/lib/python2.7/dist-packages/MySQLdb/cursors.py\", line 174, in execute\n> self.errorhandler(self, exc, value)\n> File \"/usr/local/lib/python2.7/dist-packages/MySQLdb/connections.py\", line 36, in defaulterrorhandler\n> raise errorclass, errorvalue\n> OperationalError: (2006, 'MySQL server has gone away')\n> [pid: 38|app: 0|req: 4/4] 127.0.0.1 () {48 vars in 843 bytes} [Mon Jun 6 08:43:06 2016] POST /s/user/Register => generated 114 bytes in 12 msecs (HTTP/1.0 500) 2 headers in 100 bytes (1 switches on core 1)\n> \n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub\n> https://github.com/coleifer/peewee/issues/961#issuecomment-223937057,\n> or mute the thread\n> https://github.com/notifications/unsubscribe/AAHUprAoWZngQWMLNHKtYIc01mfr7-kiks5qJAligaJpZM4Ituf3\n> .\n", "Yes, when I do these steps, the error always happens:\n1. Run my uwsgi server\n2. Do the first request, who will use peewee to do some queries etc\n3. I wait for three minutes (180 seconds) and do another request to my backend: everything works fine. When I do the third request to my backend is when the error happens. In this third request I get the error \"mysql server has gone away\". Also, I can do 20, 30 or even 40 request before the 180 seconds, but the second after this time, shit happens.\n\nMySQL configuration:\nwait_timeout 180\ninteractive_timeout 180\n\nMy peewee db settings:\n\n``` python\nimport peewee\n\ndatabase = PooledMySQLDatabase(settings.db_DBNAME, host=settings.db_HOST, user=settings.db_USER,\n passwd=settings.db_PASSWORD, max_connections=8, stale_timeout=110,\n threadlocals=True)\n\n```\n\nThe database don't receive any traffic, just these requests that I said.\n\nI tried to use the MySQLDatabase with this config:\n\n``` python\nfrom playhouse.shortcuts import RetryOperationalError\n\ndatabase = peewee.MySQLDatabase(settings.db_DBNAME, host=settings.db_HOST, user=settings.db_USER,\n passwd=settings.db_PASSWORD, threadlocals=True)\n\n```\n\nIt took more time (almost 10 minutes) to get the error, but it happened. \n", "Doing more tests, it's seems that if I leave mysql \"without use\" the problem happens. If I call my backend frequently, the error don't happen. \n", "Hi @coleifer, I fixed it creating a uwsgi spooler that hit my API and do a simple select in the database each 3 minutes. Now, I'll change the timeout configuration at MySQL to 8 hours (the default) and increase the frequency of my spooler. This is not a \"best practice\" solution, but it's the only thing that worked after almost 10 days trying everything. \n\nAnyway, very thank you for your effort creating the project and maintaining it. :+1: \n", "Glad you found something that works...I'm not very familiar with MySQL and am not too sure what might be going on. At any rate, feel free to comment with new details if you find anything out!\n", "I solved it in the following way:\r\n```\r\ndef refresh_db_connection(fn):\r\n # It's fix to avoid error \"MySQL was gone away\".\r\n # Here we check whether our current db connection is accessible or not (if not, refresh it)\r\n db.get_conn().ping(True)\r\n\r\n@refresh_db_connection\r\ndef register_nickname(nickname=\"\"):\r\n pass\r\n```\r\nJust handled the decorator to refresh the connection if it was closed." ]

[]

[]

[ "I'd like to preserve the behavior where multiple calls to `first()` does not yield multiple queries. I _do_ agree, though, that there is benefit in applying a limit and that a second `peek()` method might more correctly. I've gone ahead and implemented this in a separate patch. You can compare notes to see how I've done this a bit differently: eeb6d4d727da8536906a00c490f94352465e90bb\n", "+1\n" ]

[]

[]

[ "Sorry but that has always been a design decision I've liked, and I'm not going to change it.\n", "This is documented anyways, did you even bother to read it?\n\nhttp://docs.peewee-orm.com/en/latest/peewee/api.html#Model.get\n", "just because the quirkiness of the API is documented doesn't mean there isn't room for improvement. and i've seen no documentation/description on _why_ you like this this design decision over other alternatives.\n", "Who do you think you are?\n\n> just because the quirkiness of the API is documented doesn't mean there isn't room for improvement.\n\nYou're judging the API as being _quirky_, but don't you think that's a bit rude or presumptuous, especially given that you've only been using the library for a very very short while?\n", "> Who do you think you are?\n\na developer who generally likes your API, wants to use it and contribute back improvements. is that not welcome here?\n\nlooking back at your accepted external PRs i see many regarding minor bug fixes and documentation tweaks, but nothing beyond that, nor any other non-trivial collaborators. at the risk of being blunt: is this project just your baby/territory and all worthwhile ideas flow only from your brain, and all questions regarding even the most trivial design decisions personally offensive to you?\n", "I think you've got it backwards. You're new to the project, yet have submitted 5 issues and 4 pull-requests. Three of these issues were not actual issues - in two cases you had not read the docs, and in another (failing tests) you didn't see that it had been patched almost 2 weeks ago with a very clear commit message. The other two (which includes this one), you state that you disagree with the semantics of methods **which are clearly documented**. You call the APIs quirky, or create issues with titles implying that the implementations are incorrect -- that rubs me the wrong way.\n\nRegarding the pull-requests, you submitted a patch I found to be functionally useless, inelegantly implemented, potentially inefficient, and error-prone (are you sure you got _all_ the SQL reserved words? what about differences in database engines? what about folks who want to add a backend for a new database that has additional reserved words?). Oh, and best of all, you got your 3 buddies at work to come by and put emojis on it...or maybe it's just a coincidence you're all from houston, tx (@smtheard, @neopunisher, @ajbisberg).\n\nYou even have the gall to tell me you think it's a bad idea for me to improve my library by rewriting parts of it for a new major version.\n\nDude, you've been here for like a week, you're embarrassing yourself.\n", "The build is passing locally and on Travis ci.\n", "This is why [Ecto](https://github.com/elixir-ecto/ecto) chose to rename `get` to `first`. \n" ]

[]

[]

[ "hah! these unit tests are infuriating. not sure how to reproduce this locally. going to try some test commits...\n", "How is this any better? I'm going to pass.\n", "it's simpler and eliminates a class of bugs where one might forget to do the \"if (does this key exist?) append else set\" conditional check.\nhttps://docs.python.org/2/library/collections.html#defaultdict-objects\n" ]

[]

[]

[ "This was fixed on May 20th, e4bd8e6cb04abc826143ac64dfd9ef1bc186f070\n", "ic, thanks.\n\ni'm still getting 4 failures in `playhouse.tests.test_extra_fields`:\n\n```\n======================================================================\nERROR: test_valid_password (playhouse.tests.test_extra_fields.TestPasswordFields)\n----------------------------------------------------------------------\nTraceback (most recent call last):\n File \"/home/derek/projects/peewee/playhouse/tests/test_extra_fields.py\", line 143, in setUp\n super(TestPasswordFields, self).setUp()\n File \"/home/derek/projects/peewee/playhouse/tests/base.py\", line 285, in setUp\n test_db.create_tables(self.requires)\n File \"/home/derek/projects/peewee/peewee.py\", line 3574, in create_tables\n create_model_tables(models, fail_silently=safe)\n File \"/home/derek/projects/peewee/peewee.py\", line 4940, in create_model_tables\n m.create_table(**create_table_kwargs)\n File \"/home/derek/projects/peewee/peewee.py\", line 4624, in create_table\n db.create_table(cls)\n File \"/home/derek/projects/peewee/peewee.py\", line 3571, in create_table\n return self.execute_sql(*qc.create_table(model_class, safe))\n File \"/home/derek/projects/peewee/peewee.py\", line 3492, in execute_sql\n self.commit()\n File \"/home/derek/projects/peewee/peewee.py\", line 3316, in __exit__\n reraise(new_type, new_type(*exc_args), traceback)\n File \"/home/derek/projects/peewee/peewee.py\", line 3485, in execute_sql\n cursor.execute(sql, params or ())\nOperationalError: attempt to write a readonly database\n```\n\nboth off of 2.8.1 w/ that commit cherry-picked, and your current master.\n" ]

[]

[]

[ "No plans on my part, but I would be willing to accept a patch that implemented it somewhat along the same lines as the `postgres_ext` or `sqlite_ext` playhouse modules.\n" ]