mirror of
https://github.com/l1ving/youtube-dl
synced 2024-12-22 12:13:09 +08:00
[peertube] Improve extraction
This commit is contained in:
parent
0e30a7b973
commit
ac379fa236
@ -8,6 +8,7 @@ from ..compat import compat_str
|
|||||||
from ..utils import (
|
from ..utils import (
|
||||||
int_or_none,
|
int_or_none,
|
||||||
parse_resolution,
|
parse_resolution,
|
||||||
|
str_or_none,
|
||||||
try_get,
|
try_get,
|
||||||
unified_timestamp,
|
unified_timestamp,
|
||||||
url_or_none,
|
url_or_none,
|
||||||
@ -423,26 +424,30 @@ class PeerTubeIE(InfoExtractor):
|
|||||||
(?P<id>%s)
|
(?P<id>%s)
|
||||||
''' % (_INSTANCES_RE, _UUID_RE)
|
''' % (_INSTANCES_RE, _UUID_RE)
|
||||||
_TESTS = [{
|
_TESTS = [{
|
||||||
'url': 'https://peertube.cpy.re/videos/watch/2790feb0-8120-4e63-9af3-c943c69f5e6c',
|
'url': 'https://framatube.org/videos/watch/9c9de5e8-0a1e-484a-b099-e80766180a6d',
|
||||||
'md5': '80f24ff364cc9d333529506a263e7feb',
|
'md5': '9bed8c0137913e17b86334e5885aacff',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '2790feb0-8120-4e63-9af3-c943c69f5e6c',
|
'id': '9c9de5e8-0a1e-484a-b099-e80766180a6d',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': 'wow',
|
'title': 'What is PeerTube?',
|
||||||
'description': 'wow such video, so gif',
|
'description': '**[Want to help to translate this video?](https://weblate.framasoft.org/projects/what-is-peertube-video/)**\r\n\r\n**Take back the control of your videos! [#JoinPeertube](https://joinpeertube.org)**\r\n*A decentralized video hosting network, based on free/libre software!*\r\n\r\n**Animation Produced by:** [LILA](https://libreart.info) - [ZeMarmot Team](https://film.zemarmot.net)\r\n*Directed by* Aryeom\r\n*Assistant* Jehan\r\n**Licence**: [CC-By-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)\r\n\r\n**Sponsored by** [Framasoft](https://framasoft.org)\r\n\r\n**Music**: [Red Step Forward](http://play.dogmazic.net/song.php?song_id=52491) - CC-By Ken Bushima\r\n\r\n**Movie Clip**: [Caminades 3: Llamigos](http://www.caminandes.com/) CC-By Blender Institute\r\n\r\n**Video sources**: https://gitlab.gnome.org/Jehan/what-is-peertube/',
|
||||||
'thumbnail': r're:https?://.*\.(?:jpg|png)',
|
'thumbnail': r're:https?://.*\.(?:jpg|png)',
|
||||||
'timestamp': 1519297480,
|
'timestamp': 1538391166,
|
||||||
'upload_date': '20180222',
|
'upload_date': '20181001',
|
||||||
'uploader': 'Luclu7',
|
'uploader': 'Framasoft',
|
||||||
'uploader_id': '7fc42640-efdb-4505-a45d-a15b1a5496f1',
|
'uploader_id': '3',
|
||||||
'uploder_url': 'https://peertube.nsa.ovh/accounts/luclu7',
|
'uploader_url': 'https://framatube.org/accounts/framasoft',
|
||||||
'license': 'Unknown',
|
'channel': 'Les vidéos de Framasoft',
|
||||||
'duration': 3,
|
'channel_id': '2',
|
||||||
|
'channel_url': 'https://framatube.org/video-channels/bf54d359-cfad-4935-9d45-9d6be93f63e8',
|
||||||
|
'language': 'en',
|
||||||
|
'license': 'Attribution - Share Alike',
|
||||||
|
'duration': 113,
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
'like_count': int,
|
'like_count': int,
|
||||||
'dislike_count': int,
|
'dislike_count': int,
|
||||||
'tags': list,
|
'tags': ['framasoft', 'peertube'],
|
||||||
'categories': list,
|
'categories': ['Science & Technology'],
|
||||||
}
|
}
|
||||||
}, {
|
}, {
|
||||||
'url': 'https://peertube.tamanoir.foucry.net/videos/watch/0b04f13d-1e18-4f1d-814e-4979aa7c9c44',
|
'url': 'https://peertube.tamanoir.foucry.net/videos/watch/0b04f13d-1e18-4f1d-814e-4979aa7c9c44',
|
||||||
@ -484,6 +489,23 @@ class PeerTubeIE(InfoExtractor):
|
|||||||
entries = [peertube_url]
|
entries = [peertube_url]
|
||||||
return entries
|
return entries
|
||||||
|
|
||||||
|
def _get_subtitles(self, host, video_id):
|
||||||
|
video_captions = self._download_json(
|
||||||
|
'https://%s/api/v1/videos/%s/captions' % (host, video_id), video_id, fatal=False)
|
||||||
|
if not isinstance(video_captions, dict):
|
||||||
|
return None
|
||||||
|
|
||||||
|
subtitles = {}
|
||||||
|
for entry in video_captions.get('data'):
|
||||||
|
language_id = try_get(entry, lambda x: x['language']['id'], compat_str)
|
||||||
|
caption_path = str_or_none(entry.get('captionPath'))
|
||||||
|
if language_id and caption_path:
|
||||||
|
caption_url = urljoin('https://%s' % host, entry.get('captionPath'))
|
||||||
|
subtitles.setdefault(language_id, []).append({
|
||||||
|
'url': caption_url,
|
||||||
|
})
|
||||||
|
return subtitles
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
mobj = re.match(self._VALID_URL, url)
|
mobj = re.match(self._VALID_URL, url)
|
||||||
host = mobj.group('host') or mobj.group('host_2')
|
host = mobj.group('host') or mobj.group('host_2')
|
||||||
@ -513,10 +535,25 @@ class PeerTubeIE(InfoExtractor):
|
|||||||
formats.append(f)
|
formats.append(f)
|
||||||
self._sort_formats(formats)
|
self._sort_formats(formats)
|
||||||
|
|
||||||
def account_data(field):
|
video_description = self._download_json(
|
||||||
return try_get(video, lambda x: x['account'][field], compat_str)
|
'https://%s/api/v1/videos/%s/description' % (host, video_id), video_id, fatal=False)
|
||||||
|
|
||||||
category = try_get(video, lambda x: x['category']['label'], compat_str)
|
description = None
|
||||||
|
if isinstance(video_description, dict):
|
||||||
|
description = str_or_none(video_description.get('description'))
|
||||||
|
|
||||||
|
subtitles = self.extract_subtitles(host, video_id)
|
||||||
|
|
||||||
|
def data(section, field, type_):
|
||||||
|
return try_get(video, lambda x: x[section][field], type_)
|
||||||
|
|
||||||
|
def account_data(field, type_):
|
||||||
|
return data('account', field, type_)
|
||||||
|
|
||||||
|
def channel_data(field, type_):
|
||||||
|
return data('channel', field, type_)
|
||||||
|
|
||||||
|
category = data('category', 'label', compat_str)
|
||||||
categories = [category] if category else None
|
categories = [category] if category else None
|
||||||
|
|
||||||
nsfw = video.get('nsfw')
|
nsfw = video.get('nsfw')
|
||||||
@ -528,14 +565,17 @@ class PeerTubeIE(InfoExtractor):
|
|||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'title': title,
|
'title': title,
|
||||||
'description': video.get('description'),
|
'description': description,
|
||||||
'thumbnail': urljoin(url, video.get('thumbnailPath')),
|
'thumbnail': urljoin(url, video.get('thumbnailPath')),
|
||||||
'timestamp': unified_timestamp(video.get('publishedAt')),
|
'timestamp': unified_timestamp(video.get('publishedAt')),
|
||||||
'uploader': account_data('displayName'),
|
'uploader': account_data('displayName', compat_str),
|
||||||
'uploader_id': account_data('uuid'),
|
'uploader_id': str(account_data('id', int)),
|
||||||
'uploder_url': account_data('url'),
|
'uploader_url': url_or_none(account_data('url', compat_str)),
|
||||||
'license': try_get(
|
'channel': channel_data('displayName', compat_str),
|
||||||
video, lambda x: x['licence']['label'], compat_str),
|
'channel_id': str(channel_data('id', int)),
|
||||||
|
'channel_url': url_or_none(channel_data('url', compat_str)),
|
||||||
|
'language': data('language', 'id', compat_str),
|
||||||
|
'license': data('licence', 'label', compat_str),
|
||||||
'duration': int_or_none(video.get('duration')),
|
'duration': int_or_none(video.get('duration')),
|
||||||
'view_count': int_or_none(video.get('views')),
|
'view_count': int_or_none(video.get('views')),
|
||||||
'like_count': int_or_none(video.get('likes')),
|
'like_count': int_or_none(video.get('likes')),
|
||||||
@ -544,4 +584,5 @@ class PeerTubeIE(InfoExtractor):
|
|||||||
'tags': try_get(video, lambda x: x['tags'], list),
|
'tags': try_get(video, lambda x: x['tags'], list),
|
||||||
'categories': categories,
|
'categories': categories,
|
||||||
'formats': formats,
|
'formats': formats,
|
||||||
|
'subtitles': subtitles
|
||||||
}
|
}
|
||||||
|
Loading…
Reference in New Issue
Block a user