X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/87a0165ca7e39af4dacb7ec637063b2cd35ae40b..00368b4c3a5d4e909e1b7ecfc4030bf28da020f3:/test/test_download.py?ds=inline diff --git a/test/test_download.py b/test/test_download.py index 4639529..209f5f6 100644 --- a/test/test_download.py +++ b/test/test_download.py @@ -65,15 +65,31 @@ defs = gettestcases() class TestDownload(unittest.TestCase): + # Parallel testing in nosetests. See + # http://nose.readthedocs.org/en/latest/doc_tests/test_multiprocess/multiprocess.html + _multiprocess_shared_ = True + maxDiff = None + def __str__(self): + """Identify each test with the `add_ie` attribute, if available.""" + + def strclass(cls): + """From 2.7's unittest; 2.6 had _strclass so we can't import it.""" + return '%s.%s' % (cls.__module__, cls.__name__) + + add_ie = getattr(self, self._testMethodName).add_ie + return '%s (%s)%s:' % (self._testMethodName, + strclass(self.__class__), + ' [%s]' % add_ie if add_ie else '') + def setUp(self): self.defs = defs # Dynamically generate tests -def generator(test_case): +def generator(test_case, tname): def test_template(self): ie = youtube_dl.extractor.get_info_extractor(test_case['name']) @@ -102,6 +118,7 @@ def generator(test_case): return params = get_params(test_case.get('params', {})) + params['outtmpl'] = tname + '_' + params['outtmpl'] if is_playlist and 'playlist' not in test_case: params.setdefault('extract_flat', 'in_playlist') params.setdefault('skip_download', True) @@ -134,7 +151,7 @@ def generator(test_case): try_num = 1 while True: try: - # We're not using .download here sine that is just a shim + # We're not using .download here since that is just a shim # for outside error handling, and returns the exit code # instead of the result dict. res_dict = ydl.extract_info( @@ -146,7 +163,7 @@ def generator(test_case): raise if try_num == RETRIES: - report_warning('Failed due to network errors, skipping...') + report_warning('%s failed due to network errors, skipping...' % tname) return print('Retrying: {0} failed tries\n\n##########\n\n'.format(try_num)) @@ -182,7 +199,16 @@ def generator(test_case): self.assertEqual( test_case['playlist_duration_sum'], got_duration) - for tc in test_cases: + # Generalize both playlists and single videos to unified format for + # simplicity + if 'entries' not in res_dict: + res_dict['entries'] = [res_dict] + + for tc_num, tc in enumerate(test_cases): + tc_res_dict = res_dict['entries'][tc_num] + # First, check test cases' data against extracted data alone + expect_info_dict(self, tc_res_dict, tc.get('info_dict', {})) + # Now, check downloaded file consistency tc_filename = get_tc_filename(tc) if not test_case.get('params', {}).get('skip_download', False): self.assertTrue(os.path.exists(tc_filename), msg='Missing file ' + tc_filename) @@ -199,14 +225,15 @@ def generator(test_case): format_bytes(got_fsize))) if 'md5' in tc: md5_for_file = _file_md5(tc_filename) - self.assertEqual(md5_for_file, tc['md5']) + self.assertEqual(tc['md5'], md5_for_file) + # Finally, check test cases' data again but this time against + # extracted data from info JSON file written during processing info_json_fn = os.path.splitext(tc_filename)[0] + '.info.json' self.assertTrue( os.path.exists(info_json_fn), 'Missing info file %s' % info_json_fn) with io.open(info_json_fn, encoding='utf-8') as infof: info_dict = json.load(infof) - expect_info_dict(self, info_dict, tc.get('info_dict', {})) finally: try_rm_tcs_files() @@ -221,13 +248,15 @@ def generator(test_case): # And add them to TestDownload for n, test_case in enumerate(defs): - test_method = generator(test_case) tname = 'test_' + str(test_case['name']) i = 1 while hasattr(TestDownload, tname): tname = 'test_%s_%d' % (test_case['name'], i) i += 1 + test_method = generator(test_case, tname) test_method.__name__ = str(tname) + ie_list = test_case.get('add_ie') + test_method.add_ie = ie_list and ','.join(ie_list) setattr(TestDownload, test_method.__name__, test_method) del test_method