import streamlit as st
st.set_page_config(
    layout="centered",  # Can be "centered" or "wide". In the future also "dashboard", etc.
    initial_sidebar_state="auto",  # Can be "auto", "expanded", "collapsed"
    page_title='Extractive Summarization',  # String or None. Strings get appended with "• Streamlit".
    page_icon='./favicon.png',  # String, anything supported by st.image, or None.
)
import pandas as pd
import numpy as np
import os
import sys
sys.path.append(os.path.abspath('./'))
import streamlit_apps_config as config
from streamlit_ner_output import show_html2, jsl_display_annotations, get_color

import sparknlp
from sparknlp.base import *
from sparknlp.annotator import *
from pyspark.sql import functions as F
from sparknlp_display import NerVisualizer
from pyspark.ml import Pipeline
from pyspark.sql.types import StringType
spark= sparknlp.start()

## Marking down NER Style
st.markdown(config.STYLE_CONFIG, unsafe_allow_html=True)

root_path = config.project_path

########## To Remove the Main Menu Hamburger ########

hide_menu_style = """
        <style>
        #MainMenu {visibility: hidden;}
        </style>
        """
st.markdown(hide_menu_style, unsafe_allow_html=True)

########## Side Bar ########

## loading logo(newer version with href)
import base64
@st.cache(allow_output_mutation=True)
def get_base64_of_bin_file(bin_file):
    with open(bin_file, 'rb') as f:
        data = f.read()
    return base64.b64encode(data).decode()

@st.cache(allow_output_mutation=True)
def get_img_with_href(local_img_path, target_url):
    img_format = os.path.splitext(local_img_path)[-1].replace('.', '')
    bin_str = get_base64_of_bin_file(local_img_path)
    html_code = f'''
        <a href="{target_url}">
            <img height="90%" width="90%" src="data:image/{img_format};base64,{bin_str}" />
        </a>'''
    return html_code

logo_html = get_img_with_href('./jsl-logo.png', 'https://www.johnsnowlabs.com/')
st.sidebar.markdown(logo_html, unsafe_allow_html=True)


#sidebar info
model_name= ["nerdl_fewnerd_100d", "bert_large_token_classifier_ontonote", "ner_mit_movie_complex_distilbert_base_cased", "ner_conll_albert_large_uncased"]
st.sidebar.title("Pretrained model to test")
selected_model = st.sidebar.selectbox("", model_name)

######## Main Page #########
if selected_model == "nerdl_fewnerd_100d":
    app_title= "Detect up to 8 entity types in general domain texts"
    app_description= "Named Entity Recognition model aimed to detect up to 8 entity types from general domain texts. This model was trained on the Few-NERD/inter public dataset using Spark NLP, and it is available in Spark NLP Models hub (https://nlp.johnsnowlabs.com/models)"
    st.title(app_title)
    st.markdown("<h2>"+app_description+"</h2>" , unsafe_allow_html=True)
    st.markdown("**`PERSON`**   **,** **`ORGANIZATION`**    **,**  **`LOCATION`** **,**  **`ART`** **,**  **`BUILDING`** **,**  **`PRODUCT`** **,**  **`EVENT`** **,**  **`OTHER`**", unsafe_allow_html=True)

elif selected_model== "bert_large_token_classifier_ontonote":
    app_title= "Detect up to 18 entity types in general domain texts"
    app_description= "Named Entity Recognition model aimed to detect up to 18 entity types from general domain texts. This model is a fine-tuned BERT model that is ready to use for Named Entity Recognition and achieves state-of-the-art performance for the NER task, and it is available in Spark NLP Models hub (https://nlp.johnsnowlabs.com/models)"
    st.title(app_title)
    st.markdown("<h2>"+app_description+"</h2>" , unsafe_allow_html=True)
    st.markdown("""**`CARDINAL`**   **,** **`DATE`**    **,**  **`EVENT`** **,**  **`FAC`** **,**  **`GPE`** **,**  **`LANGUAGE`** **,**  **`LAW`** **,**  **`LOC`**,
                **`MONEY`**   **,** **`NORP`**    **,**  **`ORDINAL`** **,**  **`ORG`** **,**  **`PERCENT`** **,**  **`PERCENT`** **,**  **`PERSON`** **,**  **`PRODUCT`**,
                **`QUANTITY`**   **,** **`TIME`**    **,**  **`WORK_OF_ART` **""", unsafe_allow_html=True)


elif selected_model== "ner_mit_movie_complex_distilbert_base_cased":
    app_title= "Detect up to 12 entity types in movie domain texts"
    app_description= "Named Entity Recognition model aimed to detect up to 12 entity types from movie domain texts. This model was trained on the MIT Movie Corpus complex queries dataset to detect movie trivia using Spark NLP, and it is available in Spark NLP Models hub (https://nlp.johnsnowlabs.com/models)"
    st.title(app_title)
    st.markdown("<h2>"+app_description+"</h2>" , unsafe_allow_html=True)
    st.markdown("""**`ACTOR`**   **,** **`AWARD`**    **,**  **`CHARACTER_NAME`** **,**  **`DIRECTOR`** **,**  **`GENRE`** **,**  **`OPINION`** **,**  **`ORIGIN`** **,**  **`PLOT`**,
                    **`QUOTE`**   **,** **`RELATIONSHIP`**    **,**  **`SOUNDTRACK`** **,**  **`YEAR`  **""", unsafe_allow_html=True)

elif selected_model=="ner_conll_albert_large_uncased":
    app_title= "Detect up to 4 entity types in general domain texts"
    app_description= "Named Entity Recognition model aimed to detect up to 4 entity types from general domain texts. This model was trained on the CoNLL 2003 text corpus using Spark NLP, and it is available in Spark NLP Models hub (https://nlp.johnsnowlabs.com/models)"
    st.title(app_title)
    st.markdown("<h2>"+app_description+"</h2>" , unsafe_allow_html=True)
    st.markdown("**`PER`**   **,** **`LOC`**    **,**  **`ORG`** **,**  **`MISC` **", unsafe_allow_html=True)

st.subheader("")


#### Running model and creating pipeline
st.cache(allow_output_mutation=True)
def get_pipeline(text):
    documentAssembler = DocumentAssembler()\
        .setInputCol("text")\
        .setOutputCol("document")

    sentenceDetector= SentenceDetector()\
        .setInputCols(["document"])\
        .setOutputCol("sentence")

    tokenizer = Tokenizer()\
        .setInputCols(["sentence"])\
        .setOutputCol("token")


    if selected_model=="nerdl_fewnerd_100d":
        embeddings= WordEmbeddingsModel.pretrained("glove_100d")\
            .setInputCols(["sentence", "token"])\
            .setOutputCol("embeddings")

        ner= NerDLModel.pretrained(selected_model)\
            .setInputCols(["document", "token", "embeddings"])\
            .setOutputCol("ner")


        ner_converter= NerConverter()\
            .setInputCols(["sentence", "token", "ner"])\
            .setOutputCol("ner_chunk")


        pipeline = Pipeline(
            stages = [
                documentAssembler,
                sentenceDetector,
                tokenizer,
                embeddings,
                ner,
                ner_converter
            ])

    elif selected_model=="bert_large_token_classifier_ontonote":
        tokenClassifier = BertForTokenClassification \
            .pretrained('bert_large_token_classifier_ontonote', 'en') \
            .setInputCols(['token', 'document']) \
            .setOutputCol('ner') \
            .setCaseSensitive(True) \
            .setMaxSentenceLength(512)

        ner_converter= NerConverter()\
            .setInputCols(["document", "token", "ner"])\
            .setOutputCol("ner_chunk")
        
        pipeline = Pipeline(
            stages = [
                documentAssembler,
                sentenceDetector,
                tokenizer,
                tokenClassifier,
                ner_converter
            ])

    elif selected_model=="ner_mit_movie_complex_distilbert_base_cased":
        embeddings = DistilBertEmbeddings\
            .pretrained('distilbert_base_cased', 'en')\
            .setInputCols(["token", "document"])\
            .setOutputCol("embeddings")

        ner = NerDLModel.pretrained('ner_mit_movie_complex_distilbert_base_cased', 'en') \
            .setInputCols(['document', 'token', 'embeddings']) \
            .setOutputCol('ner')

        ner_converter= NerConverter()\
            .setInputCols(["document", "token", "ner"])\
            .setOutputCol("ner_chunk")
        
        pipeline = Pipeline(
            stages = [
                documentAssembler,
                sentenceDetector,
                tokenizer,
                embeddings,
                ner,
                ner_converter
            ])

    elif selected_model=="ner_conll_albert_large_uncased":
        embeddings = AlbertEmbeddings\
            .pretrained('albert_large_uncased', 'en')\
            .setInputCols(["document", "token"])\
            .setOutputCol("embeddings")

        ner = NerDLModel.pretrained('ner_conll_albert_large_uncased', 'en') \
            .setInputCols(['document', 'token', 'embeddings']) \
            .setOutputCol('ner')

        ner_converter = NerConverter()\
            .setInputCols(["document","token","ner"])\
            .setOutputCol("ner_chunk")

        pipeline = Pipeline(
            stages = [
                documentAssembler,
                sentenceDetector,
                tokenizer,
                embeddings,
                ner,
                ner_converter
            ])



    empty_df = spark.createDataFrame([[""]]).toDF("text")
    pipeline_model = pipeline.fit(empty_df)

    text_df= spark.createDataFrame(pd.DataFrame({"text": [text]}))
    result= pipeline_model.transform(text_df).toPandas()

    return result


    
if selected_model=="ner_mit_movie_complex_distilbert_base_cased":
    text= st.text_input("Type here your text and press enter to run:", value="It's only appropriate that Solaris, Russian filmmaker Andrei Tarkovsky's psychological sci-fi classic from 1972, contains an equally original and mind-bending score. Solaris explores the inadequacies of time and memory on an enigmatic planet below a derelict space station. To reinforce the film's chilling setting, Tarkovsky commissioned composer Eduard Artemiev to construct an electronic soundscape reflecting planet Solaris' amorphous and mysterious surface")


else:
    text= st.text_input("Type here your text and press enter to run:", value="12 Corazones ('12 Hearts') is Spanish-language dating game show produced in the United States for the television network Telemundo since January 2005, based on its namesake Argentine TV show format. The show is filmed in Los Angeles and revolves around the twelve Zodiac signs that identify each contestant. In 2008, Ho filmed a cameo in the Steven Spielberg feature film The Cloverfield Paradox, as a news pundit.")


#placeholder for warning
placeholder= st.empty()
placeholder.info("processing...")

result= get_pipeline(text)

placeholder.empty()


#Displaying Ner Visualization
df= pd.DataFrame({"ner_chunk": result["ner_chunk"].iloc[0]})

labels_set = set()
for i in df['ner_chunk'].values:
    labels_set.add(i[4]['entity'])
labels_set = list(labels_set)

labels = st.sidebar.multiselect(
        "NER Labels", options=labels_set, default=list(labels_set)
    )

show_html2(text, df, labels, "Text annotated with identified Named Entities")


try_link="""<a href="https://colab.research.google.com/github/JohnSnowLabs/spark-nlp-workshop/blob/master/tutorials/Certification_Trainings/Public/3.SparkNLP_Pretrained_Models.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" style="zoom: 1.3" alt="Open In Colab"/></a>"""
st.sidebar.title('')
st.sidebar.markdown('Try it yourself:')
st.sidebar.markdown(try_link, unsafe_allow_html=True)