PyBreak.py

import string, nltk

question_to_id_mapping, id_to_question_mapping, ans_id_mappings = {}, {}, {}
TfIdfVector, tfidf_values = [], []

DEFAULT_ANSWER = "I am sorry! I couldn't find answers for your problem. My knowledge is limited"

def compute_similarity(query):
    global sentence_tokens, DEFAULT_ANSWER
    global TfIdfVector, tfidf_values
    from sklearn.feature_extraction.text import TfidfVectorizer
    from sklearn.metrics.pairwise import cosine_similarity

    tfidf_query = TfIdfVector.transform([query])[0]
    similarity_values = cosine_similarity(tfidf_query, tfidf_values)
    flattened = similarity_values.flatten()
    flattened.sort()
    req_tfidf = flattened[-2]
    
    if req_tfidf == 0:
        return DEFAULT_ANSWER
    else:
        indexes = similarity_values.argsort()[0][::-1][:10]
        return list(map(lambda idx: sentence_tokens[idx], indexes))

def get_answers(query):
    global question_to_id_mapping, ans_id_mappings

    print(' ' * 40 + '\rAlmost done! Just arranging content for you...', end='\r')
    questions = compute_similarity(query)
    if questions == DEFAULT_ANSWER:
        return DEFAULT_ANSWER
    
    ids = list(map(lambda qn: question_to_id_mapping[qn[:-1]], questions))
    results = {}
    for question_id in ids:
        try:
            results[question_id] = ans_id_mappings[question_id]
        except:
            pass
    
    return results

def read_data():
    global sentence_tokens, question_to_id_mapping, id_to_question_mapping, ans_id_mappings
    global TfIdfVector, tfidf_values
    import pickle

    print("Please wait while I am thinking...", end='\r')
    question_to_id_mapping, id_to_question_mapping = pickle.load(open('data/question_data.pickle', 'rb'))
    ans_id_mappings, sentence_tokens = pickle.load(open('data/answer_data.pickle', 'rb'))
    TfIdfVector, tfidf_values = pickle.load(open('data/model.pickle', 'rb'))

    for question_id in id_to_question_mapping.keys():
        if question_id not in ans_id_mappings:
            continue
        for i, answer1 in enumerate(ans_id_mappings[question_id]):
            for j, answer2 in enumerate(ans_id_mappings[question_id]):
                if answer1['score'] < answer2['score']:
                    temp = ans_id_mappings[question_id][i]
                    ans_id_mappings[question_id][i] = ans_id_mappings[question_id][j]
                    ans_id_mappings[question_id][j] = temp

def beautify_print(matches):
    global id_to_question_mapping
    import html2text

    for question_id in matches.keys():
        question = id_to_question_mapping[question_id]

        if question_id not in ans_id_mappings:
            continue
        
        print('#' * 80)
        print('Question title: ' + question['title'])
        print('#' * 80)
        print('Author: User #' + question['author'])
        print('\nDescription: \n' + html2text.html2text(question['body'])); print('-' * 80)

        for i, answer in enumerate(ans_id_mappings[question_id]):
            print('[Answer ' + str(i+1) + '] by User #' + answer['author'] + ':', end='\n\n')
            print(html2text.html2text(answer['body']))
            print('-' * 80)

def PyBreak(your_code):
    def PyBreak_internal(*args, **kwargs):
        try:
            return your_code(*args, **kwargs)
        except Exception as err:
            import traceback, sys

            traceback.print_exc(file=sys.stderr)
            exception = str(err)

            if input('\n[PyBreak] Get help from stackoverflow? [y/n]: ').lower() == 'y':
                print('-' * 80, end='\n')
                read_data()
                matches = get_answers(exception)
                print(" " * 40 + "\rHere's some content from StackOverflow")
                print('-' * 80, end='\n\n')

                beautify_print(matches)
            else:
                print('Good luck with your error!')
    return PyBreak_internal