parent
c5f01bf7d4
commit
00a3e47bf5
2 changed files with 124 additions and 0 deletions
|
@ -276,6 +276,7 @@ from .brilliantpala import (
|
|||
)
|
||||
from .businessinsider import BusinessInsiderIE
|
||||
from .bundesliga import BundesligaIE
|
||||
from .bundestag import BundestagIE
|
||||
from .buzzfeed import BuzzFeedIE
|
||||
from .byutv import BYUtvIE
|
||||
from .c56 import C56IE
|
||||
|
|
123
yt_dlp/extractor/bundestag.py
Normal file
123
yt_dlp/extractor/bundestag.py
Normal file
|
@ -0,0 +1,123 @@
|
|||
import re
|
||||
from functools import partial
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..networking.exceptions import HTTPError
|
||||
from ..utils import (
|
||||
ExtractorError,
|
||||
bug_reports_message,
|
||||
clean_html,
|
||||
format_field,
|
||||
get_element_text_and_html_by_tag,
|
||||
int_or_none,
|
||||
url_or_none,
|
||||
)
|
||||
from ..utils.traversal import traverse_obj
|
||||
|
||||
|
||||
class BundestagIE(InfoExtractor):
|
||||
_VALID_URL = [
|
||||
r'https?://dbtg\.tv/[cf]vid/(?P<id>\d+)',
|
||||
r'https?://www\.bundestag\.de/mediathek/?\?(?:[^#]+&)?videoid=(?P<id>\d+)',
|
||||
]
|
||||
_TESTS = [{
|
||||
'url': 'https://dbtg.tv/cvid/7605304',
|
||||
'info_dict': {
|
||||
'id': '7605304',
|
||||
'ext': 'mp4',
|
||||
'title': '145. Sitzung vom 15.12.2023, TOP 24 Barrierefreiheit',
|
||||
'description': 'md5:321a9dc6bdad201264c0045efc371561',
|
||||
},
|
||||
}, {
|
||||
'url': 'https://www.bundestag.de/mediathek?videoid=7602120&url=L21lZGlhdGhla292ZXJsYXk=&mod=mediathek',
|
||||
'info_dict': {
|
||||
'id': '7602120',
|
||||
'ext': 'mp4',
|
||||
'title': '130. Sitzung vom 18.10.2023, TOP 1 Befragung der Bundesregierung',
|
||||
'description': 'Befragung der Bundesregierung',
|
||||
},
|
||||
}, {
|
||||
'url': 'https://www.bundestag.de/mediathek?videoid=7604941#url=L21lZGlhdGhla292ZXJsYXk/dmlkZW9pZD03NjA0OTQx&mod=mediathek',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
'url': 'http://dbtg.tv/fvid/3594346',
|
||||
'only_matching': True,
|
||||
}]
|
||||
|
||||
_OVERLAY_URL = 'https://www.bundestag.de/mediathekoverlay'
|
||||
_INSTANCE_FORMAT = 'https://cldf-wzw-od.r53.cdn.tv1.eu/13014bundestagod/_definst_/13014bundestag/ondemand/3777parlamentsfernsehen/archiv/app144277506/145293313/{0}/{0}_playlist.smil/playlist.m3u8'
|
||||
|
||||
_SHARE_URL = 'https://webtv.bundestag.de/player/macros/_x_s-144277506/shareData.json?contentId='
|
||||
_SHARE_AUDIO_REGEX = r'/\d+_(?P<codec>\w+)_(?P<bitrate>\d+)kb_(?P<channels>\w+)_\w+_\d+\.(?P<ext>\w+)'
|
||||
_SHARE_VIDEO_REGEX = r'/\d+_(?P<codec>\w+)_(?P<width>\w+)_(?P<height>\w+)_(?P<bitrate>\d+)kb_\w+_\w+_\d+\.(?P<ext>\w+)'
|
||||
|
||||
def _bt_extract_share_formats(self, video_id):
|
||||
share_data = self._download_json(
|
||||
f'{self._SHARE_URL}{video_id}', video_id, note='Downloading share format JSON')
|
||||
if traverse_obj(share_data, ('status', 'code', {int})) != 1:
|
||||
self.report_warning(format_field(
|
||||
share_data, [('status', 'message', {str})],
|
||||
'Share API response: %s', default='Unknown Share API Error')
|
||||
+ bug_reports_message())
|
||||
return
|
||||
|
||||
for name, url in share_data.items():
|
||||
if not isinstance(name, str) or not url_or_none(url):
|
||||
continue
|
||||
|
||||
elif name.startswith('audio'):
|
||||
match = re.search(self._SHARE_AUDIO_REGEX, url)
|
||||
yield {
|
||||
'format_id': name,
|
||||
'url': url,
|
||||
'vcodec': 'none',
|
||||
**traverse_obj(match, {
|
||||
'acodec': 'codec',
|
||||
'audio_channels': ('channels', {{'mono': 1, 'stereo': 2}.get}),
|
||||
'abr': ('bitrate', {int_or_none}),
|
||||
'ext': 'ext',
|
||||
}),
|
||||
}
|
||||
|
||||
elif name.startswith('download'):
|
||||
match = re.search(self._SHARE_VIDEO_REGEX, url)
|
||||
yield {
|
||||
'format_id': name,
|
||||
'url': url,
|
||||
**traverse_obj(match, {
|
||||
'vcodec': 'codec',
|
||||
'tbr': ('bitrate', {int_or_none}),
|
||||
'width': ('width', {int_or_none}),
|
||||
'height': ('height', {int_or_none}),
|
||||
'ext': 'ext',
|
||||
}),
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
formats = []
|
||||
result = {'id': video_id, 'formats': formats}
|
||||
|
||||
try:
|
||||
formats.extend(self._extract_m3u8_formats(
|
||||
self._INSTANCE_FORMAT.format(video_id), video_id, m3u8_id='instance'))
|
||||
except ExtractorError as error:
|
||||
if isinstance(error.cause, HTTPError) and error.cause.status == 404:
|
||||
raise ExtractorError('Could not find video id', expected=True)
|
||||
self.report_warning(f'Error extracting hls formats: {error}', video_id)
|
||||
formats.extend(self._bt_extract_share_formats(video_id))
|
||||
if not formats:
|
||||
self.raise_no_formats('Could not find suitable formats', video_id=video_id)
|
||||
|
||||
result.update(traverse_obj(self._download_webpage(
|
||||
self._OVERLAY_URL, video_id,
|
||||
query={'videoid': video_id, 'view': 'main'},
|
||||
note='Downloading metadata overlay', fatal=False,
|
||||
), {
|
||||
'title': (
|
||||
{partial(get_element_text_and_html_by_tag, 'h3')}, 0,
|
||||
{partial(re.sub, r'<span[^>]*>[^<]+</span>', '')}, {clean_html}),
|
||||
'description': ({partial(get_element_text_and_html_by_tag, 'p')}, 0, {clean_html}),
|
||||
}))
|
||||
|
||||
return result
|
Loading…
Reference in a new issue