Remove unused imports
[youtube-dl] / test / test_all_urls.py
index 5b6d18a8296da5d9584cf585ef0cd495eef6a2b2..0ff47cf1ead4a2c89aa24a83023a1d29cde31717 100644 (file)
@@ -15,7 +15,6 @@ from youtube_dl.extractor import (
     FacebookIE,
     gen_extractors,
     JustinTVIE,
-    PBSIE,
     YoutubeIE,
 )
 
@@ -49,6 +48,7 @@ class TestAllURLsMatching(unittest.TestCase):
         self.assertMatch('http://youtu.be/BaW_jenozKc', ['youtube'])
         self.assertMatch('http://www.youtube.com/v/BaW_jenozKc', ['youtube'])
         self.assertMatch('https://youtube.googleapis.com/v/BaW_jenozKc', ['youtube'])
+        self.assertMatch('http://www.cleanvideosearch.com/media/action/yt/watch?videoId=8v_4O44sfjM', ['youtube'])
 
     def test_youtube_channel_matching(self):
         assertChannel = lambda url: self.assertMatch(url, ['youtube:channel'])
@@ -68,28 +68,25 @@ class TestAllURLsMatching(unittest.TestCase):
     def test_youtube_show_matching(self):
         self.assertMatch('http://www.youtube.com/show/airdisasters', ['youtube:show'])
 
-    def test_youtube_truncated(self):
-        self.assertMatch('http://www.youtube.com/watch?', ['youtube:truncated_url'])
-
     def test_youtube_search_matching(self):
         self.assertMatch('http://www.youtube.com/results?search_query=making+mustard', ['youtube:search_url'])
         self.assertMatch('https://www.youtube.com/results?baz=bar&search_query=youtube-dl+test+video&filters=video&lclk=video', ['youtube:search_url'])
 
     def test_justin_tv_channelid_matching(self):
-        self.assertTrue(JustinTVIE.suitable(u"justin.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"twitch.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"www.justin.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"www.twitch.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"http://www.justin.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"http://www.twitch.tv/vanillatv"))
-        self.assertTrue(JustinTVIE.suitable(u"http://www.justin.tv/vanillatv/"))
-        self.assertTrue(JustinTVIE.suitable(u"http://www.twitch.tv/vanillatv/"))
+        self.assertTrue(JustinTVIE.suitable('justin.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('twitch.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('www.justin.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('www.twitch.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('http://www.justin.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('http://www.twitch.tv/vanillatv'))
+        self.assertTrue(JustinTVIE.suitable('http://www.justin.tv/vanillatv/'))
+        self.assertTrue(JustinTVIE.suitable('http://www.twitch.tv/vanillatv/'))
 
     def test_justintv_videoid_matching(self):
-        self.assertTrue(JustinTVIE.suitable(u"http://www.twitch.tv/vanillatv/b/328087483"))
+        self.assertTrue(JustinTVIE.suitable('http://www.twitch.tv/vanillatv/b/328087483'))
 
     def test_justin_tv_chapterid_matching(self):
-        self.assertTrue(JustinTVIE.suitable(u"http://www.twitch.tv/tsm_theoddone/c/2349361"))
+        self.assertTrue(JustinTVIE.suitable('http://www.twitch.tv/tsm_theoddone/c/2349361'))
 
     def test_youtube_extract(self):
         assertExtractId = lambda url, id: self.assertEqual(YoutubeIE.extract_id(url), id)
@@ -105,7 +102,7 @@ class TestAllURLsMatching(unittest.TestCase):
 
     def test_no_duplicates(self):
         ies = gen_extractors()
-        for tc in gettestcases():
+        for tc in gettestcases(include_onlymatching=True):
             url = tc['url']
             for ie in ies:
                 if type(ie).__name__ in ('GenericIE', tc['name'] + 'IE'):
@@ -144,7 +141,37 @@ class TestAllURLsMatching(unittest.TestCase):
         self.assertMatch('http://video.pbs.org/widget/partnerplayer/980042464/', ['PBS'])
 
     def test_ComedyCentralShows(self):
-        self.assertMatch('http://thedailyshow.cc.com/extended-interviews/xm3fnq/andrew-napolitano-extended-interview', ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thedailyshow.cc.com/extended-interviews/xm3fnq/andrew-napolitano-extended-interview',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thecolbertreport.cc.com/videos/29w6fx/-realhumanpraise-for-fox-news',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thecolbertreport.cc.com/videos/gh6urb/neil-degrasse-tyson-pt--1?xrs=eml_col_031114',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thedailyshow.cc.com/guests/michael-lewis/3efna8/exclusive---michael-lewis-extended-interview-pt--3',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thedailyshow.cc.com/episodes/sy7yv0/april-8--2014---denis-leary',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thecolbertreport.cc.com/episodes/8ase07/april-8--2014---jane-goodall',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thedailyshow.cc.com/video-playlists/npde3s/the-daily-show-19088-highlights',
+            ['ComedyCentralShows'])
+        self.assertMatch(
+            'http://thedailyshow.cc.com/special-editions/2l8fdb/special-edition---a-look-back-at-food',
+            ['ComedyCentralShows'])
+
+    def test_yahoo_https(self):
+        # https://github.com/rg3/youtube-dl/issues/2701
+        self.assertMatch(
+            'https://screen.yahoo.com/smartwatches-latest-wearable-gadgets-163745379-cbs.html',
+            ['Yahoo'])
+
 
 if __name__ == '__main__':
     unittest.main()