X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/c7b4d76a372777e3af76ecf9966a8ab9952e52f4..9b96185ac9d9c2b08e2d50684766035054c332b3:/test/test_youtube_lists.py diff --git a/test/test_youtube_lists.py b/test/test_youtube_lists.py index 78657b5..7a33dbf 100644 --- a/test/test_youtube_lists.py +++ b/test/test_youtube_lists.py @@ -1,109 +1,71 @@ #!/usr/bin/env python - -import sys -import unittest -import json +from __future__ import unicode_literals # Allow direct execution import os -sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) +import sys +import unittest +sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) -from youtube_dl.InfoExtractors import YoutubeUserIE, YoutubePlaylistIE, YoutubeIE, YoutubeChannelIE -from youtube_dl.utils import * -from youtube_dl.FileDownloader import FileDownloader +from test.helper import FakeYDL -PARAMETERS_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)), "parameters.json") -with io.open(PARAMETERS_FILE, encoding='utf-8') as pf: - parameters = json.load(pf) -# General configuration (from __init__, not very elegant...) -jar = compat_cookiejar.CookieJar() -cookie_processor = compat_urllib_request.HTTPCookieProcessor(jar) -proxy_handler = compat_urllib_request.ProxyHandler() -opener = compat_urllib_request.build_opener(proxy_handler, cookie_processor, YoutubeDLHandler()) -compat_urllib_request.install_opener(opener) +from youtube_dl.extractor import ( + YoutubePlaylistIE, + YoutubeIE, +) -class FakeDownloader(FileDownloader): - def __init__(self): - self.result = [] - self.params = parameters - def to_screen(self, s): - print(s) - def trouble(self, s, tb=None): - raise Exception(s) - def extract_info(self, url): - self.result.append(url) - return url class TestYoutubeLists(unittest.TestCase): - def assertIsPlaylist(self,info): + def assertIsPlaylist(self, info): """Make sure the info has '_type' set to 'playlist'""" self.assertEqual(info['_type'], 'playlist') - def test_youtube_playlist(self): - dl = FakeDownloader() + def test_youtube_playlist_noplaylist(self): + dl = FakeYDL() + dl.params['noplaylist'] = True ie = YoutubePlaylistIE(dl) - result = ie.extract('https://www.youtube.com/playlist?list=PLwiyx1dc3P2JR9N8gQaQN_BCvlSlap7re')[0] - self.assertIsPlaylist(result) - self.assertEqual(result['title'], 'ytdl test PL') - ytie_results = [YoutubeIE()._extract_id(url['url']) for url in result['entries']] - self.assertEqual(ytie_results, [ 'bV9L5Ht9LgY', 'FXxLjLQi3Fg', 'tU3Bgo5qJZE']) - - def test_issue_673(self): - dl = FakeDownloader() - ie = YoutubePlaylistIE(dl) - result = ie.extract('PLBB231211A4F62143')[0] - self.assertEqual(result['title'], 'Team Fortress 2') - self.assertTrue(len(result['entries']) > 40) + result = ie.extract('https://www.youtube.com/watch?v=FXxLjLQi3Fg&list=PLwiyx1dc3P2JR9N8gQaQN_BCvlSlap7re') + self.assertEqual(result['_type'], 'url') + self.assertEqual(YoutubeIE().extract_id(result['url']), 'FXxLjLQi3Fg') - def test_youtube_playlist_long(self): - dl = FakeDownloader() + def test_youtube_course(self): + dl = FakeYDL() ie = YoutubePlaylistIE(dl) - result = ie.extract('https://www.youtube.com/playlist?list=UUBABnxM4Ar9ten8Mdjj1j0Q')[0] - self.assertIsPlaylist(result) - self.assertTrue(len(result['entries']) >= 799) + # TODO find a > 100 (paginating?) videos course + result = ie.extract('https://www.youtube.com/course?list=ECUl4u3cNGP61MdtwGTqZA0MreSaDybji8') + entries = list(result['entries']) + self.assertEqual(YoutubeIE().extract_id(entries[0]['url']), 'j9WZyLZCBzs') + self.assertEqual(len(entries), 25) + self.assertEqual(YoutubeIE().extract_id(entries[-1]['url']), 'rYefUsYuEp0') - def test_youtube_playlist_with_deleted(self): - #651 - dl = FakeDownloader() - ie = YoutubePlaylistIE(dl) - result = ie.extract('https://www.youtube.com/playlist?list=PLwP_SiAcdui0KVebT0mU9Apz359a4ubsC')[0] - ytie_results = [YoutubeIE()._extract_id(url['url']) for url in result['entries']] - self.assertFalse('pElCt5oNDuI' in ytie_results) - self.assertFalse('KdPEApIVdWM' in ytie_results) - - def test_youtube_playlist_empty(self): - dl = FakeDownloader() + def test_youtube_mix(self): + dl = FakeYDL() ie = YoutubePlaylistIE(dl) - result = ie.extract('https://www.youtube.com/playlist?list=PLtPgu7CB4gbZDA7i_euNxn75ISqxwZPYx')[0] - self.assertIsPlaylist(result) - self.assertEqual(len(result['entries']), 0) + result = ie.extract('https://www.youtube.com/watch?v=W01L70IGBgE&index=2&list=RDOQpdSVF_k_w') + entries = result['entries'] + self.assertTrue(len(entries) >= 50) + original_video = entries[0] + self.assertEqual(original_video['id'], 'OQpdSVF_k_w') - def test_youtube_course(self): - dl = FakeDownloader() + def test_youtube_toptracks(self): + print('Skipping: The playlist page gives error 500') + return + dl = FakeYDL() ie = YoutubePlaylistIE(dl) - # TODO find a > 100 (paginating?) videos course - result = ie.extract('https://www.youtube.com/course?list=ECUl4u3cNGP61MdtwGTqZA0MreSaDybji8')[0] + result = ie.extract('https://www.youtube.com/playlist?list=MCUS') entries = result['entries'] - self.assertEqual(YoutubeIE()._extract_id(entries[0]['url']), 'j9WZyLZCBzs') - self.assertEqual(len(entries), 25) - self.assertEqual(YoutubeIE()._extract_id(entries[-1]['url']), 'rYefUsYuEp0') + self.assertEqual(len(entries), 100) - def test_youtube_channel(self): - dl = FakeDownloader() - ie = YoutubeChannelIE(dl) - #test paginated channel - result = ie.extract('https://www.youtube.com/channel/UCKfVa3S1e4PHvxWcwyMMg8w')[0] - self.assertTrue(len(result['entries']) > 90) - #test autogenerated channel - result = ie.extract('https://www.youtube.com/channel/HCtnHdj3df7iM/videos')[0] - self.assertTrue(len(result['entries']) >= 18) + def test_youtube_flat_playlist_titles(self): + dl = FakeYDL() + dl.params['extract_flat'] = True + ie = YoutubePlaylistIE(dl) + result = ie.extract('https://www.youtube.com/playlist?list=PLwiyx1dc3P2JR9N8gQaQN_BCvlSlap7re') + self.assertIsPlaylist(result) + for entry in result['entries']: + self.assertTrue(entry.get('title')) - def test_youtube_user(self): - dl = FakeDownloader() - ie = YoutubeUserIE(dl) - result = ie.extract('https://www.youtube.com/user/TheLinuxFoundation')[0] - self.assertTrue(len(result['entries']) >= 320) if __name__ == '__main__': unittest.main()