summaryrefslogtreecommitdiffstats
path: root/yt_dlp/extractor/nrk.py
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-08-05 09:06:11 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-08-05 09:06:11 +0000
commitfd5a06560caab95c71a2e2e805efa8d0f3a696a0 (patch)
treee1c600b8612bc4b301e2f51b875fcd835c5008cc /yt_dlp/extractor/nrk.py
parentReleasing progress-linux version 2024.05.27-1~progress7.99u1. (diff)
downloadyt-dlp-fd5a06560caab95c71a2e2e805efa8d0f3a696a0.tar.xz
yt-dlp-fd5a06560caab95c71a2e2e805efa8d0f3a696a0.zip
Merging upstream version 2024.07.01.
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'yt_dlp/extractor/nrk.py')
-rw-r--r--yt_dlp/extractor/nrk.py72
1 files changed, 34 insertions, 38 deletions
diff --git a/yt_dlp/extractor/nrk.py b/yt_dlp/extractor/nrk.py
index 384865a..658ae5f 100644
--- a/yt_dlp/extractor/nrk.py
+++ b/yt_dlp/extractor/nrk.py
@@ -3,7 +3,6 @@ import random
import re
from .common import InfoExtractor
-from ..compat import compat_str
from ..networking.exceptions import HTTPError
from ..utils import (
ExtractorError,
@@ -53,12 +52,12 @@ class NRKBaseIE(InfoExtractor):
msg=MESSAGES.get('ProgramIsGeoBlocked'),
countries=self._GEO_COUNTRIES)
message = data.get('endUserMessage') or MESSAGES.get(message_type, message_type)
- raise ExtractorError('%s said: %s' % (self.IE_NAME, message), expected=True)
+ raise ExtractorError(f'{self.IE_NAME} said: {message}', expected=True)
def _call_api(self, path, video_id, item=None, note=None, fatal=True, query=None):
return self._download_json(
urljoin('https://psapi.nrk.no/', path),
- video_id, note or 'Downloading %s JSON' % item,
+ video_id, note or f'Downloading {item} JSON',
fatal=fatal, query=query)
@@ -85,7 +84,7 @@ class NRKIE(NRKBaseIE):
'title': 'Dompap og andre fugler i Piip-Show',
'description': 'md5:d9261ba34c43b61c812cb6b0269a5c8f',
'duration': 262,
- }
+ },
}, {
# audio
'url': 'http://www.nrk.no/video/PS*154915',
@@ -96,7 +95,7 @@ class NRKIE(NRKBaseIE):
'title': 'Slik høres internett ut når du er blind',
'description': 'md5:a621f5cc1bd75c8d5104cb048c6b8568',
'duration': 20,
- }
+ },
}, {
'url': 'nrk:ecc1b952-96dc-4a98-81b9-5296dc7a98d9',
'only_matching': True,
@@ -156,7 +155,7 @@ class NRKIE(NRKBaseIE):
# known values for preferredCdn: akamai, iponly, minicdn and telenor
manifest = call_playback_api('manifest', {'preferredCdn': 'akamai'})
- video_id = try_get(manifest, lambda x: x['id'], compat_str) or video_id
+ video_id = try_get(manifest, lambda x: x['id'], str) or video_id
if manifest.get('playability') == 'nonPlayable':
self._raise_error(manifest['nonPlayable'])
@@ -216,13 +215,13 @@ class NRKIE(NRKBaseIE):
sub_key = str_or_none(sub.get('language')) or 'nb'
sub_type = str_or_none(sub.get('type'))
if sub_type:
- sub_key += '-%s' % sub_type
+ sub_key += f'-{sub_type}'
subtitles.setdefault(sub_key, []).append({
'url': sub_url,
})
legal_age = try_get(
- data, lambda x: x['legalAge']['body']['rating']['code'], compat_str)
+ data, lambda x: x['legalAge']['body']['rating']['code'], str)
# https://en.wikipedia.org/wiki/Norwegian_Media_Authority
age_limit = None
if legal_age:
@@ -243,13 +242,13 @@ class NRKIE(NRKBaseIE):
'age_limit': age_limit,
'formats': formats,
'subtitles': subtitles,
- 'timestamp': parse_iso8601(try_get(manifest, lambda x: x['availability']['onDemand']['from'], str))
+ 'timestamp': parse_iso8601(try_get(manifest, lambda x: x['availability']['onDemand']['from'], str)),
}
if is_series:
series = season_id = season_number = episode = episode_number = None
programs = self._call_api(
- 'programs/%s' % video_id, video_id, 'programs', fatal=False)
+ f'programs/{video_id}', video_id, 'programs', fatal=False)
if programs and isinstance(programs, dict):
series = str_or_none(programs.get('seriesTitle'))
season_id = str_or_none(programs.get('seasonId'))
@@ -259,7 +258,7 @@ class NRKIE(NRKBaseIE):
if not series:
series = title
if alt_title:
- title += ' - %s' % alt_title
+ title += f' - {alt_title}'
if not season_number:
season_number = int_or_none(self._search_regex(
r'Sesong\s+(\d+)', description or '', 'season number',
@@ -289,7 +288,7 @@ class NRKIE(NRKBaseIE):
class NRKTVIE(InfoExtractor):
IE_DESC = 'NRK TV and NRK Radio'
_EPISODE_RE = r'(?P<id>[a-zA-Z]{4}\d{8})'
- _VALID_URL = r'https?://(?:tv|radio)\.nrk(?:super)?\.no/(?:[^/]+/)*%s' % _EPISODE_RE
+ _VALID_URL = rf'https?://(?:tv|radio)\.nrk(?:super)?\.no/(?:[^/]+/)*{_EPISODE_RE}'
_TESTS = [{
'url': 'https://tv.nrk.no/program/MDDP12000117',
'md5': 'c4a5960f1b00b40d47db65c1064e0ab1',
@@ -306,7 +305,7 @@ class NRKTVIE(InfoExtractor):
}],
'nb-ttv': [{
'ext': 'vtt',
- }]
+ }],
},
},
}, {
@@ -411,7 +410,7 @@ class NRKTVIE(InfoExtractor):
def _real_extract(self, url):
video_id = self._match_id(url)
return self.url_result(
- 'nrk:%s' % video_id, ie=NRKIE.ie_key(), video_id=video_id)
+ f'nrk:{video_id}', ie=NRKIE.ie_key(), video_id=video_id)
class NRKTVEpisodeIE(InfoExtractor):
@@ -461,14 +460,14 @@ class NRKTVEpisodeIE(InfoExtractor):
info = self._search_json_ld(webpage, display_id, default={})
nrk_id = info.get('@id') or self._html_search_meta(
'nrk:program-id', webpage, default=None) or self._search_regex(
- r'data-program-id=["\'](%s)' % NRKTVIE._EPISODE_RE, webpage,
+ rf'data-program-id=["\']({NRKTVIE._EPISODE_RE})', webpage,
'nrk id')
assert re.match(NRKTVIE._EPISODE_RE, nrk_id)
info.update({
'_type': 'url',
'id': nrk_id,
- 'url': 'nrk:%s' % nrk_id,
+ 'url': f'nrk:{nrk_id}',
'ie_key': NRKIE.ie_key(),
'season_number': int(season_number),
'episode_number': int(episode_number),
@@ -483,13 +482,13 @@ class NRKTVSerieBaseIE(NRKBaseIE):
entries = []
for episode in entry_list:
nrk_id = episode.get('prfId') or episode.get('episodeId')
- if not nrk_id or not isinstance(nrk_id, compat_str):
+ if not nrk_id or not isinstance(nrk_id, str):
continue
entries.append(self.url_result(
- 'nrk:%s' % nrk_id, ie=NRKIE.ie_key(), video_id=nrk_id))
+ f'nrk:{nrk_id}', ie=NRKIE.ie_key(), video_id=nrk_id))
return entries
- _ASSETS_KEYS = ('episodes', 'instalments',)
+ _ASSETS_KEYS = ('episodes', 'instalments')
def _extract_assets_key(self, embedded):
for asset_key in self._ASSETS_KEYS:
@@ -514,19 +513,18 @@ class NRKTVSerieBaseIE(NRKBaseIE):
(lambda x: x[assets_key]['_embedded'][assets_key],
lambda x: x[assets_key]),
list)
- for e in self._extract_entries(entries):
- yield e
+ yield from self._extract_entries(entries)
# Find next URL
next_url_path = try_get(
data,
(lambda x: x['_links']['next']['href'],
lambda x: x['_embedded'][assets_key]['_links']['next']['href']),
- compat_str)
+ str)
if not next_url_path:
break
data = self._call_api(
next_url_path, display_id,
- note='Downloading %s JSON page %d' % (assets_key, page_num),
+ note=f'Downloading {assets_key} JSON page {page_num}',
fatal=False)
if not data:
break
@@ -593,7 +591,7 @@ class NRKTVSeasonIE(NRKTVSerieBaseIE):
@classmethod
def suitable(cls, url):
return (False if NRKTVIE.suitable(url) or NRKTVEpisodeIE.suitable(url) or NRKRadioPodkastIE.suitable(url)
- else super(NRKTVSeasonIE, cls).suitable(url))
+ else super().suitable(url))
def _real_extract(self, url):
mobj = self._match_valid_url(url)
@@ -601,14 +599,13 @@ class NRKTVSeasonIE(NRKTVSerieBaseIE):
serie_kind = mobj.group('serie_kind')
serie = mobj.group('serie')
season_id = mobj.group('id') or mobj.group('id_2')
- display_id = '%s/%s' % (serie, season_id)
+ display_id = f'{serie}/{season_id}'
data = self._call_api(
- '%s/catalog/%s/%s/seasons/%s'
- % (domain, self._catalog_name(serie_kind), serie, season_id),
+ f'{domain}/catalog/{self._catalog_name(serie_kind)}/{serie}/seasons/{season_id}',
display_id, 'season', query={'pageSize': 50})
- title = try_get(data, lambda x: x['titles']['title'], compat_str) or display_id
+ title = try_get(data, lambda x: x['titles']['title'], str) or display_id
return self.playlist_result(
self._entries(data, display_id),
display_id, title)
@@ -691,7 +688,7 @@ class NRKTVSeriesIE(NRKTVSerieBaseIE):
return (
False if any(ie.suitable(url)
for ie in (NRKTVIE, NRKTVEpisodeIE, NRKRadioPodkastIE, NRKTVSeasonIE))
- else super(NRKTVSeriesIE, cls).suitable(url))
+ else super().suitable(url))
def _real_extract(self, url):
site, serie_kind, series_id = self._match_valid_url(url).groups()
@@ -700,8 +697,7 @@ class NRKTVSeriesIE(NRKTVSerieBaseIE):
size_prefix = 'p' if is_radio else 'embeddedInstalmentsP'
series = self._call_api(
- '%s/catalog/%s/%s'
- % (domain, self._catalog_name(serie_kind), series_id),
+ f'{domain}/catalog/{self._catalog_name(serie_kind)}/{series_id}',
series_id, 'serie', query={size_prefix + 'ageSize': 50})
titles = try_get(series, [
lambda x: x['titles'],
@@ -719,8 +715,8 @@ class NRKTVSeriesIE(NRKTVSerieBaseIE):
season_url = urljoin(url, season.get('href'))
if not season_url:
season_name = season.get('name')
- if season_name and isinstance(season_name, compat_str):
- season_url = 'https://%s.nrk.no/serie/%s/sesong/%s' % (domain, series_id, season_name)
+ if season_name and isinstance(season_name, str):
+ season_url = f'https://{domain}.nrk.no/serie/{series_id}/sesong/{season_name}'
if season_url:
entries.append(self.url_result(
season_url, ie=NRKTVSeasonIE.ie_key(),
@@ -777,7 +773,7 @@ class NRKRadioPodkastIE(InfoExtractor):
def _real_extract(self, url):
video_id = self._match_id(url)
return self.url_result(
- 'nrk:%s' % video_id, ie=NRKIE.ie_key(), video_id=video_id)
+ f'nrk:{video_id}', ie=NRKIE.ie_key(), video_id=video_id)
class NRKPlaylistBaseIE(InfoExtractor):
@@ -790,7 +786,7 @@ class NRKPlaylistBaseIE(InfoExtractor):
webpage = self._download_webpage(url, playlist_id)
entries = [
- self.url_result('nrk:%s' % video_id, NRKIE.ie_key())
+ self.url_result(f'nrk:{video_id}', NRKIE.ie_key())
for video_id in re.findall(self._ITEM_RE, webpage)
]
@@ -831,7 +827,7 @@ class NRKPlaylistIE(NRKPlaylistBaseIE):
class NRKTVEpisodesIE(NRKPlaylistBaseIE):
_VALID_URL = r'https?://tv\.nrk\.no/program/[Ee]pisodes/[^/]+/(?P<id>\d+)'
- _ITEM_RE = r'data-episode=["\']%s' % NRKTVIE._EPISODE_RE
+ _ITEM_RE = rf'data-episode=["\']{NRKTVIE._EPISODE_RE}'
_TESTS = [{
'url': 'https://tv.nrk.no/program/episodes/nytt-paa-nytt/69031',
'info_dict': {
@@ -869,7 +865,7 @@ class NRKSkoleIE(InfoExtractor):
video_id = self._match_id(url)
nrk_id = self._download_json(
- 'https://nrkno-skole-prod.kube.nrk.no/skole/api/media/%s' % video_id,
+ f'https://nrkno-skole-prod.kube.nrk.no/skole/api/media/{video_id}',
video_id)['psId']
- return self.url_result('nrk:%s' % nrk_id)
+ return self.url_result(f'nrk:{nrk_id}')