import subprocess
import sys

def install(package):
    subprocess.check_call([sys.executable, "-m", "pip", "install", package])


install('youtube-transcript-api')
install('pytube')
install('thirdai[neural_db]')
# !pip install -Uqq youtube-transcript-api
# !pip install -Uqq pytube
# !pip3 install -Uqq thirdai[neural_db]

from youtube_transcript_api import YouTubeTranscriptApi
import pandas as pd
from pytube import Playlist
import re
import time

def get_video_links_from_playlist(playlist_url):
    playlist = Playlist(playlist_url)
    
    video_links = [video.watch_url for video in playlist.videos]
    return video_links

def extract_video_id(video_url):
    # Regular expression pattern to match YouTube video ID
    pattern = r"(?:v=|v\/|embed\/|youtu.be\/|\/v\/|\/e\/|watch\?v=|youtube.com\/user\/[^#]*#([^\/]*?\/)*)((\w|-){11})"

    match = re.search(pattern, video_url)
    if match:
        return match.group(2)
    else:
        return None

def create_csv(transcript_lst):
    
    flat_data = [item for sublist in transcript_lst for item in sublist]
    df = pd.DataFrame(flat_data)
    df['end_time'] = df['start'] + df['duration']
    df = df[['text', 'start', 'duration', 'end_time', 'video_serial_number']]
    return df

def create_data(playlist_link):
    all_playlist_videos_lst = get_video_links_from_playlist(playlist_link)
    # create individual video id list
    video_id_lst = []
    for video_link in all_playlist_videos_lst:
        video_id_lst.append(extract_video_id(video_link))
    # get transcripts for all video_ids
    transcript_lst = []
    for video_serial in range(len(video_id_lst)):
        video_id = video_id_lst[video_serial]
        try:
            video_transcript = YouTubeTranscriptApi.get_transcript(video_id)
            new_transcript = []
            for i in range(0, len(video_transcript), 4):
                start_time = video_transcript[i]['start']
                duration = 0
                text = ''
                for j in range(4):
                    index = i + j
                    if index < len(video_transcript):
                        text += video_transcript[index]['text']
                        text += ' '
                        duration += video_transcript[index]['duration']
                dict = {'text' : text,
                        'start' : start_time,
                        'duration' : duration,
                       'video_serial_number':video_serial+1}
                # print('serail_number: ',video_serial)
                new_transcript.append(dict)
            transcript_lst.append(new_transcript)
        except:
            continue 

    # make dataframe from transcript list
    if len(transcript_lst)!=0:
        csv_file = create_csv(transcript_lst)
        csv_file['text'] = csv_file['text'].str.replace(r'\n', ' ')
        return csv_file
    return 0