import pandas as pd
import numpy as np

pd.set_option('display.max_columns', None)

def main():
    process_yt()

def process_yt():
    stream_counts = pd.read_feather(f'../data/yt_streams.feather')
    stream_counts['stream_date'] = stream_counts['as_of'].dt.date
    stream_info = pd.read_feather(f'../data/yt_info.feather')
    stream_info['first_seen'] = stream_info['first_seen'].dt.date
    release_dates = stream_info.groupby('ytid')[['first_seen']].first().reset_index()

    stream_counts = pd.merge(stream_counts, release_dates, left_on='video_ytid', right_on='ytid')
    stream_counts = stream_counts[['ytid', 'comment_count', 'dislike_count', 'favorite_count', 'like_count', 'view_count', 'stream_date', 'first_seen']]

    stream_counts['days_after_first_seen'] =  (stream_counts['stream_date'] - stream_counts['first_seen'])/pd.Timedelta('1d')

    stream_counts[stream_counts['days_after_first_seen'] < 0].groupby('ytid')['days_after_first_seen'].first()
    stream_counts = stream_counts.sort_values('stream_date')
    first_seen_after = stream_counts.groupby('ytid')['days_after_first_seen'].first()
    counts_per_track = stream_counts.groupby('ytid').size()
    remove_tracks = counts_per_track[counts_per_track < 15].index.values
    # 3024

    stream_counts = stream_counts[~stream_counts.ytid.isin(remove_tracks)]
    stream_counts_change = stream_counts.copy()

    sc_change = stream_counts_change[stream_counts_change['days_after_first_seen'] < 100].set_index('ytid')[['view_count', 'stream_date','days_after_first_seen']].sort_values('stream_date')
     
    sc_change['view'] = sc_change['view_count'].astype(np.float64)
    a = sc_change.groupby('ytid')['view'].diff()



def process_soundcloud():
    idfield = 'scid'
    stream_counts = pd.read_feather('../data/sound_cloud_streams.feather')
    stream_counts['stream_date'] = stream_counts['as_of']

    stream_info = pd.read_feather('../data/sound_cloud_info.feather')
    release_dates = stream_info.groupby(idfield)[['created_at']].first().reset_index()

    stream_counts = pd.merge(stream_counts, release_dates, left_on='track_scid', right_on=idfield)
    stream_counts['days_after_first_seen'] =  (stream_counts['stream_date'] - stream_counts['created_at'])/pd.Timedelta('1d')

    first_seen_after = stream_counts.groupby(idfield)['days_after_first_seen'].first()
    counts_per_track = stream_counts.groupby(idfield).size()
    remove_tracks = counts_per_track[counts_per_track < 15].index.values
    stream_counts = stream_counts[~stream_counts[idfield].isin(remove_tracks)]
    stream_counts_change = stream_counts.copy()

    stream_counts_change = stream_counts_change.sort_values([idfield, 'stream_date'])
    stream_counts_change['view_count'] = stream_counts_change['playback_count']
    stream_counts_change['view_count_day'] = stream_counts_change.groupby(idfield)['playback_count'].diff()
    stream_counts_change['daydiff'] = stream_counts_change.groupby(idfield)['stream_date'].diff()/pd.Timedelta('1d')

    stream_counts_change = stream_counts_change[['scid', 'download_count', 'favoritings_count', 'comment_count', 'stream_date', 'days_after_first_seen', 'daydiff', 'view_count', 'view_count_day']]

    stream_counts_change.reset_index().to_feather('../data/sound_cloud_streams_with_diff.feather')


def normaliseStreams(yt, idfield, max_days = 100):

    yt = stream_counts_change
    yt_n = yt.copy()
    yt_n = yt_n[yt_n['days_after_first_seen'] <= max_days]
    yt_n = yt_n[yt_n['days_after_first_seen'] >= 0]
    
    
    total_count_per_track = yt_n.groupby(idfield)[['view_count_day']].sum().reset_index().rename(columns = {'view_count_day':'total_view_count'})
    yt_n = pd.merge(yt_n, total_count_per_track, on= idfield)

    yt_n['normalised_count_day'] = yt_n['view_count_day']/yt_n['total_view_count']

    return yt_n