import os
import numpy as np
import gradio as gr
import stanza
from simpletransformers.classification import ClassificationModel, ClassificationArgs
import preprocessor as p


def clean_text(text):
    text = text.replace("#", "")
    return p.clean(text)


def softmax(x):
    
    return np.exp(x) / np.sum(np.exp(x), axis=0)


def number_to_sentiment(number):

    sentiments = {
        '0': 'Negative',
        '1': 'Neutral',
        '2': 'Positive'
    }

    return sentiments[str(number)]


def number_to_topic(number):
    topics = {
        '0': 'Abortions',
        '1': 'Taiwan',
        '2': 'Afghanistan',
        '3': 'Insurance',
        '4': 'Undefined'
    }
    return topics[str(number)]


def text_processing(text):

    results = nlp(text)
    text = clean_text(text)
    number_of_sentiments = 0
    number_of_sentences = 0
    for i, sentence in enumerate(results.sentences):
        number_of_sentiments += int(sentence.sentiment)
        number_of_sentences += 1

    sentiment = int(round(number_of_sentiments/number_of_sentences))
    sentiment = number_to_sentiment(sentiment)

    predictions, raw_outputs = model.predict(text)

    print(predictions[0], raw_outputs[0])

    softmax_pred = softmax(raw_outputs[0])
    if softmax_pred.max() > 0.90:
        topic = number_to_topic(softmax_pred.argmax())
        print(softmax_pred.argmax())
    else:
        print(4)
        topic = number_to_topic(4)

    return f'Text topic: {topic}, text sentiment: {sentiment}'


if __name__ == "__main__":

    eval_model_args = ClassificationArgs(max_seq_length=128, use_multiprocessing_for_evaluation=False,
                                         eval_batch_size=1)

    model = ClassificationModel(
        "xlnet", "./", use_cuda=False, args=eval_model_args
    )

    stanza.download('en')
    nlp = stanza.Pipeline('en', processors='sentiment,tokenize,mwt', tokenize_no_ssplit=True)

    with gr.Blocks() as demo:
        with gr.Tab("Get text topic and sentiment"):
            text_input = gr.Textbox(label='Input text', placeholder='Put your text here')
            text_output = gr.Textbox(label='Output', placeholder="Topic and sentiment of the text")
            text_button = gr.Button("Run processing")

        text_button.click(text_processing, inputs=text_input, outputs=text_output)

    demo.launch()