0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-25 02:45:12 +00:00
yt-dlp/yt_dlp/extractor/nfb.py
rrgomes 4b8b0dded8
[ie/nfb] Add support for onf.ca and series (#8997)
Closes #8198
Authored by: bashonly, rrgomes

Co-authored-by: bashonly <88596187+bashonly@users.noreply.github.com>
2024-01-31 18:00:15 +00:00

300 lines
13 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.

from .common import InfoExtractor
from ..utils import (
int_or_none,
join_nonempty,
merge_dicts,
parse_count,
url_or_none,
urljoin,
)
from ..utils.traversal import traverse_obj
class NFBBaseIE(InfoExtractor):
_VALID_URL_BASE = r'https?://(?:www\.)?(?P<site>nfb|onf)\.ca'
_GEO_COUNTRIES = ['CA']
def _extract_ep_data(self, webpage, video_id, fatal=False):
return self._search_json(
r'const\s+episodesData\s*=', webpage, 'episode data', video_id,
contains_pattern=r'\[\s*{(?s:.+)}\s*\]', fatal=fatal) or []
def _extract_ep_info(self, data, video_id, slug=None):
info = traverse_obj(data, (lambda _, v: video_id in v['embed_url'], {
'description': ('description', {str}),
'thumbnail': ('thumbnail_url', {url_or_none}),
'uploader': ('data_layer', 'episodeMaker', {str}),
'release_year': ('data_layer', 'episodeYear', {int_or_none}),
'episode': ('data_layer', 'episodeTitle', {str}),
'season': ('data_layer', 'seasonTitle', {str}),
'season_number': ('data_layer', 'seasonTitle', {parse_count}),
'series': ('data_layer', 'seriesTitle', {str}),
}), get_all=False)
return {
**info,
'id': video_id,
'title': join_nonempty('series', 'episode', from_dict=info, delim=' - '),
'episode_number': int_or_none(self._search_regex(
r'[/-]e(?:pisode)?-?(\d+)(?:[/-]|$)', slug or video_id, 'episode number', default=None)),
}
class NFBIE(NFBBaseIE):
IE_NAME = 'nfb'
IE_DESC = 'nfb.ca and onf.ca films and episodes'
_VALID_URL = [
rf'{NFBBaseIE._VALID_URL_BASE}/(?P<type>film)/(?P<id>[^/?#&]+)',
rf'{NFBBaseIE._VALID_URL_BASE}/(?P<type>series?)/(?P<id>[^/?#&]+/s(?:ea|ai)son\d+/episode\d+)',
]
_TESTS = [{
'note': 'NFB film',
'url': 'https://www.nfb.ca/film/trafficopter/',
'info_dict': {
'id': 'trafficopter',
'ext': 'mp4',
'title': 'Trafficopter',
'description': 'md5:060228455eb85cf88785c41656776bc0',
'thumbnail': r're:^https?://.*\.jpg$',
'uploader': 'Barrie Howells',
'release_year': 1972,
'duration': 600.0,
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'ONF film',
'url': 'https://www.onf.ca/film/mal-du-siecle/',
'info_dict': {
'id': 'mal-du-siecle',
'ext': 'mp4',
'title': 'Le mal du siècle',
'description': 'md5:1abf774d77569ebe603419f2d344102b',
'thumbnail': r're:^https?://.*\.jpg$',
'uploader': 'Catherine Lepage',
'release_year': 2019,
'duration': 300.0,
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'NFB episode with English title',
'url': 'https://www.nfb.ca/series/true-north-inside-the-rise-of-toronto-basketball/season1/episode9/',
'info_dict': {
'id': 'true-north-episode9-true-north-finale-making-it',
'ext': 'mp4',
'title': 'True North: Inside the Rise of Toronto Basketball - Finale: Making It',
'description': 'We catch up with each player in the midst of their journey as they reflect on their road ahead.',
'series': 'True North: Inside the Rise of Toronto Basketball',
'release_year': 2018,
'season': 'Season 1',
'season_number': 1,
'episode': 'Finale: Making It',
'episode_number': 9,
'uploader': 'Ryan Sidhoo',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'ONF episode with French title',
'url': 'https://www.onf.ca/serie/direction-nord-la-montee-du-basketball-a-toronto/saison1/episode9/',
'info_dict': {
'id': 'direction-nord-episode-9',
'ext': 'mp4',
'title': 'Direction nord La montée du basketball à Toronto - Finale : Réussir',
'description': 'md5:349a57419b71432b97bf6083d92b029d',
'series': 'Direction nord La montée du basketball à Toronto',
'release_year': 2018,
'season': 'Saison 1',
'season_number': 1,
'episode': 'Finale : Réussir',
'episode_number': 9,
'uploader': 'Ryan Sidhoo',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'NFB episode with French title (needs geo-bypass)',
'url': 'https://www.nfb.ca/series/etoile-du-nord/saison1/episode1/',
'info_dict': {
'id': 'etoile-du-nord-episode-1-lobservation',
'ext': 'mp4',
'title': 'Étoile du Nord - L\'observation',
'description': 'md5:161a4617260dee3de70f509b2c9dd21b',
'series': 'Étoile du Nord',
'release_year': 2023,
'season': 'Saison 1',
'season_number': 1,
'episode': 'L\'observation',
'episode_number': 1,
'uploader': 'Patrick Bossé',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'ONF episode with English title (needs geo-bypass)',
'url': 'https://www.onf.ca/serie/north-star/season1/episode1/',
'info_dict': {
'id': 'north-star-episode-1-observation',
'ext': 'mp4',
'title': 'North Star - Observation',
'description': 'md5:c727f370839d8a817392b9e3f23655c7',
'series': 'North Star',
'release_year': 2023,
'season': 'Season 1',
'season_number': 1,
'episode': 'Observation',
'episode_number': 1,
'uploader': 'Patrick Bossé',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'NFB episode with /film/ URL and English title (needs geo-bypass)',
'url': 'https://www.nfb.ca/film/north-star-episode-1-observation/',
'info_dict': {
'id': 'north-star-episode-1-observation',
'ext': 'mp4',
'title': 'North Star - Observation',
'description': 'md5:c727f370839d8a817392b9e3f23655c7',
'series': 'North Star',
'release_year': 2023,
'season': 'Season 1',
'season_number': 1,
'episode': 'Observation',
'episode_number': 1,
'uploader': 'Patrick Bossé',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'ONF episode with /film/ URL and French title (needs geo-bypass)',
'url': 'https://www.onf.ca/film/etoile-du-nord-episode-1-lobservation/',
'info_dict': {
'id': 'etoile-du-nord-episode-1-lobservation',
'ext': 'mp4',
'title': 'Étoile du Nord - L\'observation',
'description': 'md5:161a4617260dee3de70f509b2c9dd21b',
'series': 'Étoile du Nord',
'release_year': 2023,
'season': 'Saison 1',
'season_number': 1,
'episode': 'L\'observation',
'episode_number': 1,
'uploader': 'Patrick Bossé',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'Season 2 episode w/o episode num in id, extract from json ld',
'url': 'https://www.onf.ca/film/liste-des-choses-qui-existent-saison-2-ours',
'info_dict': {
'id': 'liste-des-choses-qui-existent-saison-2-ours',
'ext': 'mp4',
'title': 'La liste des choses qui existent - L\'ours en peluche',
'description': 'md5:d5e8d8fc5f3a7385a9cf0f509b37e28a',
'series': 'La liste des choses qui existent',
'release_year': 2022,
'season': 'Saison 2',
'season_number': 2,
'episode': 'L\'ours en peluche',
'episode_number': 12,
'uploader': 'Francis Papillon',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}, {
'note': 'NFB film /embed/player/ page',
'url': 'https://www.nfb.ca/film/afterlife/embed/player/',
'info_dict': {
'id': 'afterlife',
'ext': 'mp4',
'title': 'Afterlife',
'description': 'md5:84951394f594f1fb1e62d9c43242fdf5',
'release_year': 1978,
'duration': 420.0,
'uploader': 'Ishu Patel',
'thumbnail': r're:^https?://.*\.jpg$',
},
'params': {'skip_download': 'm3u8'},
}]
def _real_extract(self, url):
site, type_, slug = self._match_valid_url(url).group('site', 'type', 'id')
# Need to construct the URL since we match /embed/player/ URLs as well
webpage, urlh = self._download_webpage_handle(f'https://www.{site}.ca/{type_}/{slug}/', slug)
# type_ can change from film to serie(s) after redirect; new slug may have episode number
type_, slug = self._match_valid_url(urlh.url).group('type', 'id')
embed_url = urljoin(f'https://www.{site}.ca', self._html_search_regex(
r'<[^>]+\bid=["\']player-iframe["\'][^>]*\bsrc=["\']([^"\']+)', webpage, 'embed url'))
video_id = self._match_id(embed_url) # embed url has unique slug
player = self._download_webpage(embed_url, video_id, 'Downloading player page')
if 'MESSAGE_GEOBLOCKED' in player:
self.raise_geo_restricted(countries=self._GEO_COUNTRIES)
formats, subtitles = self._extract_m3u8_formats_and_subtitles(
self._html_search_regex(r'source:\s*\'([^\']+)', player, 'm3u8 url'),
video_id, 'mp4', m3u8_id='hls')
if dv_source := self._html_search_regex(r'dvSource:\s*\'([^\']+)', player, 'dv', default=None):
fmts, subs = self._extract_m3u8_formats_and_subtitles(
dv_source, video_id, 'mp4', m3u8_id='dv', preference=-2, fatal=False)
for fmt in fmts:
fmt['format_note'] = 'described video'
formats.extend(fmts)
self._merge_subtitles(subs, target=subtitles)
info = {
'id': video_id,
'title': self._html_search_regex(
r'<[^>]+\bid=["\']titleHeader["\'][^>]*>\s*<h1[^>]*>\s*([^<]+?)\s*</h1>',
webpage, 'title', default=None),
'description': self._html_search_regex(
r'<[^>]+\bid=["\']tabSynopsis["\'][^>]*>\s*<p[^>]*>\s*([^<]+)',
webpage, 'description', default=None),
'thumbnail': self._html_search_regex(
r'poster:\s*\'([^\']+)', player, 'thumbnail', default=None),
'uploader': self._html_search_regex(
r'<[^>]+\bitemprop=["\']name["\'][^>]*>([^<]+)', webpage, 'uploader', default=None),
'release_year': int_or_none(self._html_search_regex(
r'<[^>]+\bitemprop=["\']datePublished["\'][^>]*>([^<]+)',
webpage, 'release_year', default=None)),
} if type_ == 'film' else self._extract_ep_info(self._extract_ep_data(webpage, video_id, slug), video_id)
return merge_dicts({
'formats': formats,
'subtitles': subtitles,
}, info, self._search_json_ld(webpage, video_id, default={}))
class NFBSeriesIE(NFBBaseIE):
IE_NAME = 'nfb:series'
IE_DESC = 'nfb.ca and onf.ca series'
_VALID_URL = rf'{NFBBaseIE._VALID_URL_BASE}/(?P<type>series?)/(?P<id>[^/?#&]+)/?(?:[?#]|$)'
_TESTS = [{
'url': 'https://www.nfb.ca/series/true-north-inside-the-rise-of-toronto-basketball/',
'playlist_mincount': 9,
'info_dict': {
'id': 'true-north-inside-the-rise-of-toronto-basketball',
},
}, {
'url': 'https://www.onf.ca/serie/la-liste-des-choses-qui-existent-serie/',
'playlist_mincount': 26,
'info_dict': {
'id': 'la-liste-des-choses-qui-existent-serie',
},
}]
def _entries(self, episodes):
for episode in traverse_obj(episodes, lambda _, v: NFBIE.suitable(v['embed_url'])):
mobj = NFBIE._match_valid_url(episode['embed_url'])
yield self.url_result(
mobj[0], NFBIE, **self._extract_ep_info([episode], mobj.group('id')))
def _real_extract(self, url):
site, type_, series_id = self._match_valid_url(url).group('site', 'type', 'id')
season_path = 'saison' if type_ == 'serie' else 'season'
webpage = self._download_webpage(
f'https://www.{site}.ca/{type_}/{series_id}/{season_path}1/episode1', series_id)
episodes = self._extract_ep_data(webpage, series_id, fatal=True)
return self.playlist_result(self._entries(episodes), series_id)