youtube-dl/youtube_dl/extractor/ellentv.py

109 lines
3.5 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
import json
from .common import InfoExtractor
from ..utils import (
ExtractorError,
NO_DEFAULT,
)
class EllenTVIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?(?:ellentv|ellentube)\.com/videos/(?P<id>[a-z0-9_-]+)'
_TESTS = [{
'url': 'http://www.ellentv.com/videos/0-ipq1gsai/',
'md5': '4294cf98bc165f218aaa0b89e0fd8042',
'info_dict': {
'id': '0_ipq1gsai',
'ext': 'mov',
'title': 'Fast Fingers of Fate',
'description': 'md5:3539013ddcbfa64b2a6d1b38d910868a',
'timestamp': 1428035648,
'upload_date': '20150403',
'uploader_id': 'batchUser',
},
}, {
# not available via http://widgets.ellentube.com/
'url': 'http://www.ellentv.com/videos/1-szkgu2m2/',
'info_dict': {
'id': '1_szkgu2m2',
'ext': 'flv',
'title': "Ellen's Amazingly Talented Audience",
'description': 'md5:86ff1e376ff0d717d7171590e273f0a5',
'timestamp': 1255140900,
'upload_date': '20091010',
'uploader_id': 'ellenkaltura@gmail.com',
},
'params': {
'skip_download': True,
},
}]
def _real_extract(self, url):
video_id = self._match_id(url)
URLS = ('http://widgets.ellentube.com/videos/%s' % video_id, url)
for num, url_ in enumerate(URLS, 1):
webpage = self._download_webpage(
url_, video_id, fatal=num == len(URLS))
default = NO_DEFAULT if num == len(URLS) else None
partner_id = self._search_regex(
r"var\s+partnerId\s*=\s*'([^']+)", webpage, 'partner id',
default=default)
kaltura_id = self._search_regex(
[r'id="kaltura_player_([^"]+)"',
r"_wb_entry_id\s*:\s*'([^']+)",
r'data-kaltura-entry-id="([^"]+)'],
webpage, 'kaltura id', default=default)
if partner_id and kaltura_id:
break
return self.url_result('kaltura:%s:%s' % (partner_id, kaltura_id), 'Kaltura')
class EllenTVClipsIE(InfoExtractor):
IE_NAME = 'EllenTV:clips'
_VALID_URL = r'https?://(?:www\.)?ellentv\.com/episodes/(?P<id>[a-z0-9_-]+)'
_TEST = {
'url': 'http://www.ellentv.com/episodes/meryl-streep-vanessa-hudgens/',
'info_dict': {
'id': 'meryl-streep-vanessa-hudgens',
'title': 'Meryl Streep, Vanessa Hudgens',
},
'playlist_mincount': 7,
}
def _real_extract(self, url):
playlist_id = self._match_id(url)
webpage = self._download_webpage(url, playlist_id)
playlist = self._extract_playlist(webpage)
return {
'_type': 'playlist',
'id': playlist_id,
'title': self._og_search_title(webpage),
'entries': self._extract_entries(playlist)
}
def _extract_playlist(self, webpage):
json_string = self._search_regex(r'playerView.addClips\(\[\{(.*?)\}\]\);', webpage, 'json')
try:
return json.loads('[{' + json_string + '}]')
except ValueError as ve:
raise ExtractorError('Failed to download JSON', cause=ve)
def _extract_entries(self, playlist):
return [
self.url_result(
'kaltura:%s:%s' % (item['kaltura_partner_id'], item['kaltura_entry_id']),
'Kaltura')
for item in playlist]