mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-26 09:11:25 +01:00
Compare commits
No commits in common. "984f9233c49ce99c2d754c143b95ce5674e3121f" and "2410edc0015a444f8bf68896e2c62337df2a1f29" have entirely different histories.
984f9233c4
...
2410edc001
|
@ -481,7 +481,7 @@ class TestTraversalHelpers:
|
|||
'id': 'name',
|
||||
'data': 'content',
|
||||
'url': 'url',
|
||||
}, all, {subs_list_to_dict(lang=None)}]) == {
|
||||
}, all, {subs_list_to_dict}]) == {
|
||||
'de': [{'url': 'https://example.com/subs/de.ass'}],
|
||||
'en': [{'data': 'content'}],
|
||||
}, 'subs with mandatory items missing should be filtered'
|
||||
|
@ -507,54 +507,6 @@ class TestTraversalHelpers:
|
|||
{'url': 'https://example.com/subs/en1', 'ext': 'ext'},
|
||||
{'url': 'https://example.com/subs/en2', 'ext': 'ext'},
|
||||
]}, '`quality` key should sort subtitle list accordingly'
|
||||
assert traverse_obj([
|
||||
{'name': 'de', 'url': 'https://example.com/subs/de.ass'},
|
||||
{'name': 'de'},
|
||||
{'name': 'en', 'content': 'content'},
|
||||
{'url': 'https://example.com/subs/en'},
|
||||
], [..., {
|
||||
'id': 'name',
|
||||
'url': 'url',
|
||||
'data': 'content',
|
||||
}, all, {subs_list_to_dict(lang='en')}]) == {
|
||||
'de': [{'url': 'https://example.com/subs/de.ass'}],
|
||||
'en': [
|
||||
{'data': 'content'},
|
||||
{'url': 'https://example.com/subs/en'},
|
||||
],
|
||||
}, 'optionally provided lang should be used if no id available'
|
||||
assert traverse_obj([
|
||||
{'name': 1, 'url': 'https://example.com/subs/de1'},
|
||||
{'name': {}, 'url': 'https://example.com/subs/de2'},
|
||||
{'name': 'de', 'ext': 1, 'url': 'https://example.com/subs/de3'},
|
||||
{'name': 'de', 'ext': {}, 'url': 'https://example.com/subs/de4'},
|
||||
], [..., {
|
||||
'id': 'name',
|
||||
'url': 'url',
|
||||
'ext': 'ext',
|
||||
}, all, {subs_list_to_dict(lang=None)}]) == {
|
||||
'de': [
|
||||
{'url': 'https://example.com/subs/de3'},
|
||||
{'url': 'https://example.com/subs/de4'},
|
||||
],
|
||||
}, 'non str types should be ignored for id and ext'
|
||||
assert traverse_obj([
|
||||
{'name': 1, 'url': 'https://example.com/subs/de1'},
|
||||
{'name': {}, 'url': 'https://example.com/subs/de2'},
|
||||
{'name': 'de', 'ext': 1, 'url': 'https://example.com/subs/de3'},
|
||||
{'name': 'de', 'ext': {}, 'url': 'https://example.com/subs/de4'},
|
||||
], [..., {
|
||||
'id': 'name',
|
||||
'url': 'url',
|
||||
'ext': 'ext',
|
||||
}, all, {subs_list_to_dict(lang='de')}]) == {
|
||||
'de': [
|
||||
{'url': 'https://example.com/subs/de1'},
|
||||
{'url': 'https://example.com/subs/de2'},
|
||||
{'url': 'https://example.com/subs/de3'},
|
||||
{'url': 'https://example.com/subs/de4'},
|
||||
],
|
||||
}, 'non str types should be replaced by default id'
|
||||
|
||||
def test_trim_str(self):
|
||||
with pytest.raises(TypeError):
|
||||
|
@ -573,7 +525,7 @@ class TestTraversalHelpers:
|
|||
def test_unpack(self):
|
||||
assert unpack(lambda *x: ''.join(map(str, x)))([1, 2, 3]) == '123'
|
||||
assert unpack(join_nonempty)([1, 2, 3]) == '1-2-3'
|
||||
assert unpack(join_nonempty, delim=' ')([1, 2, 3]) == '1 2 3'
|
||||
assert unpack(join_nonempty(delim=' '))([1, 2, 3]) == '1 2 3'
|
||||
with pytest.raises(TypeError):
|
||||
unpack(join_nonempty)()
|
||||
with pytest.raises(TypeError):
|
||||
|
|
|
@ -72,6 +72,7 @@ from yt_dlp.utils import (
|
|||
intlist_to_bytes,
|
||||
iri_to_uri,
|
||||
is_html,
|
||||
join_nonempty,
|
||||
js_to_json,
|
||||
limit_length,
|
||||
locked_file,
|
||||
|
@ -2157,6 +2158,10 @@ Line 1
|
|||
assert int_or_none(v=10) == 10, 'keyword passed positional should call function'
|
||||
assert int_or_none(scale=0.1)(10) == 100, 'call after partial application should call the function'
|
||||
|
||||
assert callable(join_nonempty(delim=', ')), 'varargs positional should apply partially'
|
||||
assert callable(join_nonempty()), 'varargs positional should apply partially'
|
||||
assert join_nonempty(None, delim=', ') == '', 'passed varargs should call the function'
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
unittest.main()
|
||||
|
|
|
@ -3767,7 +3767,7 @@ class InfoExtractor:
|
|||
""" Merge subtitle dictionaries, language by language. """
|
||||
if target is None:
|
||||
target = {}
|
||||
for d in filter(None, dicts):
|
||||
for d in dicts:
|
||||
for lang, subs in d.items():
|
||||
target[lang] = cls._merge_subtitle_items(target.get(lang, []), subs)
|
||||
return target
|
||||
|
|
|
@ -13,10 +13,7 @@ from ..utils import (
|
|||
unified_timestamp,
|
||||
url_or_none,
|
||||
)
|
||||
from ..utils.traversal import (
|
||||
subs_list_to_dict,
|
||||
traverse_obj,
|
||||
)
|
||||
from ..utils.traversal import traverse_obj
|
||||
|
||||
|
||||
class RutubeBaseIE(InfoExtractor):
|
||||
|
@ -95,11 +92,11 @@ class RutubeBaseIE(InfoExtractor):
|
|||
hls_url, video_id, 'mp4', fatal=False, m3u8_id='hls')
|
||||
formats.extend(fmts)
|
||||
self._merge_subtitles(subs, target=subtitles)
|
||||
self._merge_subtitles(traverse_obj(options, ('captions', ..., {
|
||||
'id': 'code',
|
||||
'url': 'file',
|
||||
'name': ('langTitle', {str}),
|
||||
}, all, {subs_list_to_dict(lang='ru')})), target=subtitles)
|
||||
for caption in traverse_obj(options, ('captions', lambda _, v: url_or_none(v['file']))):
|
||||
subtitles.setdefault(caption.get('code') or 'ru', []).append({
|
||||
'url': caption['file'],
|
||||
'name': caption.get('langTitle'),
|
||||
})
|
||||
return formats, subtitles
|
||||
|
||||
def _download_and_extract_formats_and_subtitles(self, video_id, query=None):
|
||||
|
|
|
@ -216,7 +216,7 @@ def partial_application(func):
|
|||
sig = inspect.signature(func)
|
||||
required_args = [
|
||||
param.name for param in sig.parameters.values()
|
||||
if param.kind in (inspect.Parameter.POSITIONAL_ONLY, inspect.Parameter.POSITIONAL_OR_KEYWORD)
|
||||
if param.kind in (inspect.Parameter.POSITIONAL_ONLY, inspect.Parameter.POSITIONAL_OR_KEYWORD, inspect.Parameter.VAR_POSITIONAL)
|
||||
if param.default is inspect.Parameter.empty
|
||||
]
|
||||
|
||||
|
@ -4837,6 +4837,7 @@ def number_of_digits(number):
|
|||
return len('%d' % number)
|
||||
|
||||
|
||||
@partial_application
|
||||
def join_nonempty(*values, delim='-', from_dict=None):
|
||||
if from_dict is not None:
|
||||
values = (traversal.traverse_obj(from_dict, variadic(v)) for v in values)
|
||||
|
|
|
@ -332,14 +332,14 @@ class _RequiredError(ExtractorError):
|
|||
|
||||
|
||||
@typing.overload
|
||||
def subs_list_to_dict(*, lang: str | None = 'und', ext: str | None = None) -> collections.abc.Callable[[list[dict]], dict[str, list[dict]]]: ...
|
||||
def subs_list_to_dict(*, ext: str | None = None) -> collections.abc.Callable[[list[dict]], dict[str, list[dict]]]: ...
|
||||
|
||||
|
||||
@typing.overload
|
||||
def subs_list_to_dict(subs: list[dict] | None, /, *, lang: str | None = 'und', ext: str | None = None) -> dict[str, list[dict]]: ...
|
||||
def subs_list_to_dict(subs: list[dict] | None, /, *, ext: str | None = None) -> dict[str, list[dict]]: ...
|
||||
|
||||
|
||||
def subs_list_to_dict(subs: list[dict] | None = None, /, *, lang='und', ext=None):
|
||||
def subs_list_to_dict(subs: list[dict] | None = None, /, *, ext=None):
|
||||
"""
|
||||
Convert subtitles from a traversal into a subtitle dict.
|
||||
The path should have an `all` immediately before this function.
|
||||
|
@ -352,7 +352,7 @@ def subs_list_to_dict(subs: list[dict] | None = None, /, *, lang='und', ext=None
|
|||
`quality` The sort order for each subtitle
|
||||
"""
|
||||
if subs is None:
|
||||
return functools.partial(subs_list_to_dict, lang=lang, ext=ext)
|
||||
return functools.partial(subs_list_to_dict, ext=ext)
|
||||
|
||||
result = collections.defaultdict(list)
|
||||
|
||||
|
@ -360,16 +360,10 @@ def subs_list_to_dict(subs: list[dict] | None = None, /, *, lang='und', ext=None
|
|||
if not url_or_none(sub.get('url')) and not sub.get('data'):
|
||||
continue
|
||||
sub_id = sub.pop('id', None)
|
||||
if not isinstance(sub_id, str):
|
||||
if not lang:
|
||||
continue
|
||||
sub_id = lang
|
||||
sub_ext = sub.get('ext')
|
||||
if not isinstance(sub_ext, str):
|
||||
if not ext:
|
||||
sub.pop('ext', None)
|
||||
else:
|
||||
sub['ext'] = ext
|
||||
if sub_id is None:
|
||||
continue
|
||||
if ext is not None and not sub.get('ext'):
|
||||
sub['ext'] = ext
|
||||
result[sub_id].append(sub)
|
||||
result = dict(result)
|
||||
|
||||
|
@ -458,9 +452,9 @@ def trim_str(*, start=None, end=None):
|
|||
return trim
|
||||
|
||||
|
||||
def unpack(func, **kwargs):
|
||||
def unpack(func):
|
||||
@functools.wraps(func)
|
||||
def inner(items):
|
||||
def inner(items, **kwargs):
|
||||
return func(*items, **kwargs)
|
||||
|
||||
return inner
|
||||
|
|
Loading…
Reference in New Issue
Block a user