mirror of
https://github.com/yt-dlp/yt-dlp
synced 2024-12-27 21:59:17 +01:00
parent
20b91b9b63
commit
600e900300
1 changed files with 22 additions and 20 deletions
|
@ -14,6 +14,7 @@ from ..utils import (
|
|||
orderedSet,
|
||||
parse_codecs,
|
||||
qualities,
|
||||
str_or_none,
|
||||
try_get,
|
||||
unified_timestamp,
|
||||
update_url_query,
|
||||
|
@ -49,35 +50,35 @@ class ZDFBaseIE(InfoExtractor):
|
|||
|
||||
def _extract_format(self, video_id, formats, format_urls, meta):
|
||||
format_url = url_or_none(meta.get('url'))
|
||||
if not format_url:
|
||||
return
|
||||
if format_url in format_urls:
|
||||
if not format_url or format_url in format_urls:
|
||||
return
|
||||
format_urls.add(format_url)
|
||||
mime_type = meta.get('mimeType')
|
||||
ext = determine_ext(format_url)
|
||||
|
||||
mime_type, ext = meta.get('mimeType'), determine_ext(format_url)
|
||||
if mime_type == 'application/x-mpegURL' or ext == 'm3u8':
|
||||
formats.extend(self._extract_m3u8_formats(
|
||||
new_formats = self._extract_m3u8_formats(
|
||||
format_url, video_id, 'mp4', m3u8_id='hls',
|
||||
entry_protocol='m3u8_native', fatal=False))
|
||||
entry_protocol='m3u8_native', fatal=False)
|
||||
elif mime_type == 'application/f4m+xml' or ext == 'f4m':
|
||||
formats.extend(self._extract_f4m_formats(
|
||||
update_url_query(format_url, {'hdcore': '3.7.0'}), video_id, f4m_id='hds', fatal=False))
|
||||
new_formats = self._extract_f4m_formats(
|
||||
update_url_query(format_url, {'hdcore': '3.7.0'}), video_id, f4m_id='hds', fatal=False)
|
||||
else:
|
||||
f = parse_codecs(meta.get('mimeCodec'))
|
||||
format_id = ['http']
|
||||
for p in (meta.get('type'), meta.get('quality')):
|
||||
if p and isinstance(p, compat_str):
|
||||
format_id.append(p)
|
||||
if not f and meta.get('type'):
|
||||
data = meta['type'].split('_')
|
||||
if try_get(data, lambda x: x[2]) == ext:
|
||||
f = {'vcodec': data[0], 'acodec': data[1]}
|
||||
f.update({
|
||||
'url': format_url,
|
||||
'format_id': '-'.join(format_id),
|
||||
'format_note': meta.get('quality'),
|
||||
'language': meta.get('language'),
|
||||
'quality': qualities(self._QUALITIES)(meta.get('quality')),
|
||||
'preference': -10,
|
||||
'format_id': '-'.join(filter(str_or_none, ('http', meta.get('type'), meta.get('quality')))),
|
||||
})
|
||||
formats.append(f)
|
||||
new_formats = [f]
|
||||
formats.extend(merge_dicts(f, {
|
||||
'format_note': ', '.join(filter(None, (meta.get('quality'), meta.get('class')))),
|
||||
'language': meta.get('language'),
|
||||
'language_preference': 10 if meta.get('class') == 'main' else -10 if meta.get('class') == 'ad' else -1,
|
||||
'quality': qualities(self._QUALITIES)(meta.get('quality')),
|
||||
}) for f in new_formats)
|
||||
|
||||
def _extract_ptmd(self, ptmd_url, video_id, api_token, referrer):
|
||||
ptmd = self._call_api(
|
||||
|
@ -106,9 +107,10 @@ class ZDFBaseIE(InfoExtractor):
|
|||
'type': f.get('type'),
|
||||
'mimeType': f.get('mimeType'),
|
||||
'quality': quality.get('quality'),
|
||||
'class': track.get('class'),
|
||||
'language': track.get('language'),
|
||||
})
|
||||
self._sort_formats(formats)
|
||||
self._sort_formats(formats, ('hasaud', 'res', 'quality', 'language_preference'))
|
||||
|
||||
duration = float_or_none(try_get(
|
||||
ptmd, lambda x: x['attributes']['duration']['value']), scale=1000)
|
||||
|
|
Loading…
Reference in a new issue