X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/03342304420e5daeb428ffdcc7bbd2bbfecfa61a..bb8f7fa1e66fb0b783b3fdc24d75ed2ca0694e80:/test/test_all_urls.py diff --git a/test/test_all_urls.py b/test/test_all_urls.py index f5af184..cd1cd4b 100644 --- a/test/test_all_urls.py +++ b/test/test_all_urls.py @@ -6,6 +6,7 @@ from __future__ import unicode_literals import os import sys import unittest +import collections sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) @@ -100,8 +101,6 @@ class TestAllURLsMatching(unittest.TestCase): self.assertMatch(':ytsubs', ['youtube:subscriptions']) self.assertMatch(':ytsubscriptions', ['youtube:subscriptions']) self.assertMatch(':ythistory', ['youtube:history']) - self.assertMatch(':thedailyshow', ['ComedyCentralShows']) - self.assertMatch(':tds', ['ComedyCentralShows']) def test_vimeo_matching(self): self.assertMatch('https://vimeo.com/channels/tributes', ['vimeo:channel']) @@ -130,6 +129,15 @@ class TestAllURLsMatching(unittest.TestCase): 'https://screen.yahoo.com/smartwatches-latest-wearable-gadgets-163745379-cbs.html', ['Yahoo']) + def test_no_duplicated_ie_names(self): + name_accu = collections.defaultdict(list) + for ie in self.ies: + name_accu[ie.IE_NAME.lower()].append(type(ie).__name__) + for (ie_name, ie_list) in name_accu.items(): + self.assertEqual( + len(ie_list), 1, + 'Multiple extractors with the same IE_NAME "%s" (%s)' % (ie_name, ', '.join(ie_list))) + if __name__ == '__main__': unittest.main()