import gensim as gensim
import smart_open
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LinearRegression
from stop_words import get_stop_words
from sklearn.cluster import KMeans
from gensim.models.doc2vec import Doc2Vec
import os
import pandas as pd

def read_train_file(inDirectory):
    colnames = ['start_date', 'end_date', 'title', 'sort_title', 'data']
    df_train = pd.read_csv(inDirectory, sep="\t", names=colnames)
    return df_train[:5000]

def read_evaluate_file(inDirectory):
    colnames = ['data']
    df_train = pd.read_csv(inDirectory, sep="\t", names=colnames)
    return df_train[:5000]

def train_date_mean(df):
    date_mean = (df['start_date'] + df['end_date']) / 2
    return date_mean

def preper_data(df):
        document_list = list()
        for line in df:
            tokens = gensim.utils.simple_preprocess(line, min_len=2, max_len=15)
            filtered_sentence = []
            for word in tokens:
                if word not in get_stop_words('polish'):
                    filtered_sentence.append(word)
            document_list.append(filtered_sentence)
        return document_list

df = read_train_file('train/train.tsv')
date_mean_df = train_date_mean(df)[:5000]
vectorizer = TfidfVectorizer(stop_words=get_stop_words('polish'))
train_vectorized_corpus = vectorizer.fit_transform(df['data'])
reg = LinearRegression().fit(train_vectorized_corpus, date_mean_df)

df_evaluate = read_evaluate_file('dev-0/in.tsv')
evaluate_vectorized_corpus = vectorizer.transform(df_evaluate['data'])
evaluate = reg.predict(evaluate_vectorized_corpus)
with open("dev-0/out.tsv", 'w') as file:
    for e in evaluate:
        file.write("%i\n" % e)