1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-16 06:15:46 +00:00

[pokemon] add extractor for watch.pokemon.com

This commit is contained in:
ssaqua 2020-04-28 20:15:50 +10:00
parent 1bc1520adc
commit 1a082f5aa8
No known key found for this signature in database
GPG key ID: A590B1193655DEEE
2 changed files with 71 additions and 1 deletions

View file

@ -874,7 +874,10 @@ from .pluralsight import (
PluralsightCourseIE,
)
from .podomatic import PodomaticIE
from .pokemon import PokemonIE
from .pokemon import (
PokemonIE,
PokemonWatchIE,
)
from .polskieradio import (
PolskieRadioIE,
PolskieRadioCategoryIE,

View file

@ -5,8 +5,11 @@ import re
from .common import InfoExtractor
from ..utils import (
ExtractorError,
extract_attributes,
int_or_none,
js_to_json,
merge_dicts,
)
@ -69,3 +72,67 @@ class PokemonIE(InfoExtractor):
'episode_number': int_or_none(video_data.get('data-video-episode')),
'ie_key': 'LimelightMedia',
}
class PokemonWatchIE(InfoExtractor):
_VALID_URL = r'https?://watch\.pokemon\.com/[a-z]{2}-[a-z]{2}/player\.html\?id=(?P<id>[a-z0-9]{32})'
_API_URL = 'https://www.pokemon.com/api/pokemontv/v2/channels/{0:}'
_TESTS = [{
'url': 'https://watch.pokemon.com/en-us/player.html?id=8309a40969894a8e8d5bc1311e9c5667',
'md5': '62833938a31e61ab49ada92f524c42ff',
'info_dict': {
'id': '8309a40969894a8e8d5bc1311e9c5667',
'ext': 'mp4',
'title': 'Lillier and the Staff!',
'description': 'md5:338841b8c21b283d24bdc9b568849f04',
}
}, {
'url': 'https://watch.pokemon.com/de-de/player.html?id=b3c402e111a4459eb47e12160ab0ba07',
'only_matching': True
}]
def _extract_media(self, channel_array, video_id):
for channel in channel_array:
for media in channel.get('media'):
if media.get('id') == video_id:
return media
return None
def _real_extract(self, url):
video_id = self._match_id(url)
info = {
'_type': 'url',
'id': video_id,
'url': 'limelight:media:%s' % video_id,
'ie_key': 'LimelightMedia',
}
# API call can be avoided entirely if we are listing formats
if self._downloader.params.get('listformats', False):
return info
webpage = self._download_webpage(url, video_id)
build_vars = self._parse_json(self._search_regex(
r'(?s)buildVars\s*=\s*({.*?})', webpage, 'build vars'),
video_id, transform_source=js_to_json)
region = build_vars.get('region')
channel_array = self._download_json(self._API_URL.format(region), video_id)
video_data = self._extract_media(channel_array, video_id)
if video_data is None:
raise ExtractorError(
'Video %s does not exist' % video_id, expected=True)
info['_type'] = 'url_transparent'
images = video_data.get('images')
return merge_dicts(info, {
'title': video_data.get('title'),
'description': video_data.get('description'),
'thumbnail': images.get('medium') or images.get('small'),
'series': 'Pokémon',
'season_number': int_or_none(video_data.get('season')),
'episode': video_data.get('title'),
'episode_number': int_or_none(video_data.get('episode')),
})