mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-27 17:26:51 +01:00
[testurl] Add extractor
This is a pseudo extractor that can be used to quickly look up test URLs, or test without the test harness.
This commit is contained in:
parent
f6f01ea17b
commit
a9c7198a0b
@ -216,6 +216,7 @@ from .sztvhu import SztvHuIE
|
||||
from .teamcoco import TeamcocoIE
|
||||
from .techtalks import TechTalksIE
|
||||
from .ted import TEDIE
|
||||
from .testurl import TestURLIE
|
||||
from .tf1 import TF1IE
|
||||
from .theplatform import ThePlatformIE
|
||||
from .thisav import ThisAVIE
|
||||
|
66
youtube_dl/extractor/testurl.py
Normal file
66
youtube_dl/extractor/testurl.py
Normal file
@ -0,0 +1,66 @@
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import ExtractorError
|
||||
|
||||
|
||||
class TestURLIE(InfoExtractor):
|
||||
""" Allows adressing of the test cases as test:yout.*be_1 """
|
||||
|
||||
IE_DESC = False # Do not list
|
||||
_VALID_URL = r'test(?:url)?:(?P<id>(?P<extractor>.+?)(?:_(?P<num>[0-9]+))?)$'
|
||||
|
||||
def _real_extract(self, url):
|
||||
from ..extractor import gen_extractors
|
||||
|
||||
mobj = re.match(self._VALID_URL, url)
|
||||
video_id = mobj.group('id')
|
||||
extractor_id = mobj.group('extractor')
|
||||
all_extractors = gen_extractors()
|
||||
|
||||
rex = re.compile(extractor_id, flags=re.IGNORECASE)
|
||||
matching_extractors = [
|
||||
e for e in all_extractors if rex.search(e.IE_NAME)]
|
||||
|
||||
if len(matching_extractors) == 0:
|
||||
raise ExtractorError(
|
||||
'No extractors matching %r found' % extractor_id,
|
||||
expected=True)
|
||||
elif len(matching_extractors) > 1:
|
||||
# Is it obvious which one to pick?
|
||||
try:
|
||||
extractor = next(
|
||||
ie for ie in matching_extractors
|
||||
if ie.IE_NAME.lower() == extractor_id.lower())
|
||||
except StopIteration:
|
||||
raise ExtractorError(
|
||||
('Found multiple matching extractors: %s' %
|
||||
' '.join(ie.IE_NAME for ie in matching_extractors)),
|
||||
expected=True)
|
||||
|
||||
num_str = mobj.group('num')
|
||||
num = int(num_str) if num_str else 0
|
||||
|
||||
testcases = []
|
||||
t = getattr(extractor, '_TEST', None)
|
||||
if t:
|
||||
testcases.append(t)
|
||||
testcases.extend(getattr(extractor, '_TESTS', []))
|
||||
|
||||
try:
|
||||
tc = testcases[num]
|
||||
except IndexError:
|
||||
raise ExtractorError(
|
||||
('Test case %d not found, got only %d tests' %
|
||||
(num, len(testcases))),
|
||||
expected=True)
|
||||
|
||||
self.to_screen('Test URL: %s' % tc['url'])
|
||||
|
||||
return {
|
||||
'_type': 'url',
|
||||
'url': tc['url'],
|
||||
'id': video_id,
|
||||
}
|
Loading…
Reference in New Issue
Block a user