aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--test/test_all_urls.py14
-rw-r--r--test/test_youtube_lists.py13
-rwxr-xr-xyoutube_dl/InfoExtractors.py65
3 files changed, 59 insertions, 33 deletions
diff --git a/test/test_all_urls.py b/test/test_all_urls.py
index 06de8e7b8..69717b3fc 100644
--- a/test/test_all_urls.py
+++ b/test/test_all_urls.py
@@ -11,12 +11,18 @@ from youtube_dl.InfoExtractors import YoutubeIE, YoutubePlaylistIE
class TestAllURLsMatching(unittest.TestCase):
def test_youtube_playlist_matching(self):
- self.assertTrue(YoutubePlaylistIE().suitable(u'ECUl4u3cNGP61MdtwGTqZA0MreSaDybji8'))
- self.assertTrue(YoutubePlaylistIE().suitable(u'PL63F0C78739B09958'))
- self.assertFalse(YoutubePlaylistIE().suitable(u'PLtS2H6bU1M'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'ECUl4u3cNGP61MdtwGTqZA0MreSaDybji8'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'UUBABnxM4Ar9ten8Mdjj1j0Q')) #585
+ self.assertTrue(YoutubePlaylistIE.suitable(u'PL63F0C78739B09958'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'https://www.youtube.com/playlist?list=UUBABnxM4Ar9ten8Mdjj1j0Q'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'https://www.youtube.com/course?list=ECUl4u3cNGP61MdtwGTqZA0MreSaDybji8'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'https://www.youtube.com/playlist?list=PLwP_SiAcdui0KVebT0mU9Apz359a4ubsC'))
+ self.assertTrue(YoutubePlaylistIE.suitable(u'https://www.youtube.com/watch?v=AV6J6_AeFEQ&playnext=1&list=PL4023E734DA416012')) #668
+ self.assertFalse(YoutubePlaylistIE.suitable(u'PLtS2H6bU1M'))
def test_youtube_matching(self):
- self.assertTrue(YoutubeIE().suitable(u'PLtS2H6bU1M'))
+ self.assertTrue(YoutubeIE.suitable(u'PLtS2H6bU1M'))
+ self.assertFalse(YoutubeIE.suitable(u'https://www.youtube.com/watch?v=AV6J6_AeFEQ&playnext=1&list=PL4023E734DA416012')) #668
def test_youtube_extract(self):
self.assertEqual(YoutubeIE()._extract_id('http://www.youtube.com/watch?&v=BaW_jenozKc'), 'BaW_jenozKc')
diff --git a/test/test_youtube_lists.py b/test/test_youtube_lists.py
index 69b0f4447..7055d5a7a 100644
--- a/test/test_youtube_lists.py
+++ b/test/test_youtube_lists.py
@@ -41,6 +41,18 @@ class TestYoutubeLists(unittest.TestCase):
self.assertEqual(map(lambda x: YoutubeIE()._extract_id(x[0]), DL.result),
[ 'bV9L5Ht9LgY', 'FXxLjLQi3Fg', 'tU3Bgo5qJZE' ])
+ #661
+ DL = FakeDownloader()
+ IE = YoutubePlaylistIE(DL)
+ IE.extract('PLMCmkNmxw6Z9eduM7BZjSEh7HiU543Ig0')
+ self.assertTrue(len(DL.result) > 20)
+
+ #673
+ DL = FakeDownloader()
+ IE = YoutubePlaylistIE(DL)
+ IE.extract('PLBB231211A4F62143')
+ self.assertTrue(len(DL.result) > 40)
+
def test_youtube_playlist_long(self):
DL = FakeDownloader()
IE = YoutubePlaylistIE(DL)
@@ -48,6 +60,7 @@ class TestYoutubeLists(unittest.TestCase):
self.assertTrue(len(DL.result) >= 799)
def test_youtube_playlist_with_deleted(self):
+ #651
DL = FakeDownloader()
IE = YoutubePlaylistIE(DL)
IE.extract('https://www.youtube.com/playlist?list=PLwP_SiAcdui0KVebT0mU9Apz359a4ubsC')
diff --git a/youtube_dl/InfoExtractors.py b/youtube_dl/InfoExtractors.py
index 021579ce0..490f30491 100755
--- a/youtube_dl/InfoExtractors.py
+++ b/youtube_dl/InfoExtractors.py
@@ -74,13 +74,15 @@ class InfoExtractor(object):
self._ready = False
self.set_downloader(downloader)
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url) is not None
+ return re.match(cls._VALID_URL, url) is not None
- def working(self):
+ @classmethod
+ def working(cls):
"""Getter method for _WORKING."""
- return self._WORKING
+ return cls._WORKING
def initialize(self):
"""Initializes an instance (authentication, etc)."""
@@ -137,7 +139,6 @@ class YoutubeIE(InfoExtractor):
(?:youtu\.be/|(?:\w+\.)?youtube(?:-nocookie)?\.com/|
tube\.majestyc\.net/) # the various hostnames, with wildcard subdomains
(?:.*?\#/)? # handle anchor (#/) redirect urls
- (?!view_play_list|my_playlists|artist|playlist) # ignore playlist URLs
(?: # the various things that can precede the ID:
(?:(?:v|embed|e)/) # v/ or embed/ or e/
|(?: # or the v= param in all its forms
@@ -189,9 +190,11 @@ class YoutubeIE(InfoExtractor):
}
IE_NAME = u'youtube'
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url, re.VERBOSE) is not None
+ if YoutubePlaylistIE.suitable(url): return False
+ return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
def report_lang(self):
"""Report attempt to set language."""
@@ -1668,17 +1671,17 @@ class YoutubePlaylistIE(InfoExtractor):
(?:\w+\.)?
youtube\.com/
(?:
- (?:course|view_play_list|my_playlists|artist|playlist)
- \? .*? (p|a|list)=
+ (?:course|view_play_list|my_playlists|artist|playlist|watch)
+ \? (?:.*?&)*? (?:p|a|list)=
| user/.*?/user/
| p/
| user/.*?#[pg]/c/
)
- (?:PL|EC)?
- |PL|EC)
- ([0-9A-Za-z-_]{10,})
- (?:/.*?/([0-9A-Za-z_-]+))?
- .*"""
+ ((?:PL|EC|UU)?[0-9A-Za-z-_]{10,})
+ .*
+ |
+ ((?:PL|EC|UU)[0-9A-Za-z-_]{10,})
+ )"""
_TEMPLATE_URL = 'https://gdata.youtube.com/feeds/api/playlists/%s?max-results=%i&start-index=%i&v=2&alt=json'
_MAX_RESULTS = 50
IE_NAME = u'youtube:playlist'
@@ -1686,9 +1689,10 @@ class YoutubePlaylistIE(InfoExtractor):
def __init__(self, downloader=None):
InfoExtractor.__init__(self, downloader)
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url, re.VERBOSE) is not None
+ return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
def report_download_page(self, playlist_id, pagenum):
"""Report attempt to download playlist page with given number."""
@@ -1701,13 +1705,8 @@ class YoutubePlaylistIE(InfoExtractor):
self._downloader.trouble(u'ERROR: invalid url: %s' % url)
return
- # Single video case
- if mobj.group(3) is not None:
- self._downloader.download([mobj.group(3)])
- return
-
# Download playlist videos from API
- playlist_id = mobj.group(2)
+ playlist_id = mobj.group(1) or mobj.group(2)
page_num = 1
videos = []
@@ -1727,7 +1726,12 @@ class YoutubePlaylistIE(InfoExtractor):
self._downloader.trouble(u'ERROR: Invalid JSON in API response: ' + compat_str(err))
return
- videos += [(entry['yt$position']['$t'], entry['content']['src']) for entry in response['feed']['entry']]
+ if not 'feed' in response or not 'entry' in response['feed']:
+ self._downloader.trouble(u'ERROR: Got a malformed response from YouTube API')
+ return
+ videos += [ (entry['yt$position']['$t'], entry['content']['src'])
+ for entry in response['feed']['entry']
+ if 'content' in entry ]
if len(response['feed']['entry']) < self._MAX_RESULTS:
break
@@ -2313,9 +2317,10 @@ class ComedyCentralIE(InfoExtractor):
'400': '384x216',
}
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url, re.VERBOSE) is not None
+ return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
def report_extraction(self, episode_id):
self._downloader.to_screen(u'[comedycentral] %s: Extracting information' % episode_id)
@@ -3628,9 +3633,10 @@ class SteamIE(InfoExtractor):
(?P<videoID>\d*)(?P<extra>\??) #For urltype == video we sometimes get the videoID
"""
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url, re.VERBOSE) is not None
+ return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
def _real_extract(self, url):
m = re.match(self._VALID_URL, url, re.VERBOSE)
@@ -4004,9 +4010,10 @@ class TEDIE(InfoExtractor):
/(?P<name>\w+) # Here goes the name and then ".html"
'''
- def suitable(self, url):
+ @classmethod
+ def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
- return re.match(self._VALID_URL, url, re.VERBOSE) is not None
+ return re.match(cls._VALID_URL, url, re.VERBOSE) is not None
def _real_extract(self, url):
m=re.match(self._VALID_URL, url, re.VERBOSE)