yt-dlp/youtube_dl/extractor/teamfourstar.py
felix 8953319916 [screenwavemedia] Remove extractor
Rewrite TeamFourStar and Normalboots extractors in terms of JWPlatform
2016-11-28 23:17:56 +07:00

48 lines
2.4 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from .jwplatform import JWPlatformIE
from ..utils import unified_strdate
class TeamFourStarIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?teamfourstar\.com/(?P<id>[a-z0-9\-]+)/?'
_TEST = {
'url': 'http://teamfourstar.com/tfs-abridged-parody-episode-1-2/',
'info_dict': {
'id': '0WdZO31W',
'title': 'TFS Abridged Parody Episode 1',
'description': 'Episode 1: The Return of Raditz! … Wait…\nCast\nMasakoX Goku, Roshi\nLanipator Piccolo, Radditz, Krillin, Vegeta\nVegeta3986 Radditz, Yamcha, Oolong, Gohan\nHbi2k Farmer with Shotgun\nMegami33 Bulma, Puar\nTakahata101 Nappa\nKaiserNeko SpacePod\nSongs\nMorgenstemning by Edvard Hagerup Grieg\nCha-La-Head-Cha-La by Kageyama Hiranobu\nWE DO NOT OWN DRAGONBALL. DragonBall is Owned by TOEI ANIMATION, Ltd. and Licensed by FUNimation Productions, Ltd.. All Rights Reserved. DragonBall, DragonBall Z, DragonBall GT and all logos, character names and distinctive likenesses thereof are trademarks of TOEI ANIMATION, Ltd.\nThis is nothing more than a Parody made for entertainment purposes only.',
'ext': 'mp4',
'timestamp': 1394168400,
'upload_date': '20080508',
},
}
def _real_extract(self, url):
display_id = self._match_id(url)
webpage = self._download_webpage(url, display_id)
jwplatform_url = JWPlatformIE._extract_url(webpage)
video_title = self._html_search_regex(
r'<h1 class="entry-title">(?P<title>.+?)</h1>',
webpage, 'title')
video_date = unified_strdate(self._html_search_regex(
r'<span class="meta-date date updated">(?P<date>.+?)</span>',
webpage, 'date', fatal=False))
video_description = self._html_search_regex(
r'(?s)<div class="content-inner">.*?(?P<description><p>.+?)</div>',
webpage, 'description', fatal=False)
video_thumbnail = self._og_search_thumbnail(webpage)
return {
'_type': 'url_transparent',
'display_id': display_id,
'title': video_title,
'description': video_description,
'upload_date': video_date,
'thumbnail': video_thumbnail,
'url': jwplatform_url,
}