- #!/usr/bin/env python
-
- from __future__ import unicode_literals
-
- # Allow direct execution
- import os
- import sys
- import unittest
- sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-
- from test.helper import (
- assertGreaterEqual,
- expect_warnings,
- get_params,
- gettestcases,
- expect_info_dict,
- try_rm,
- report_warning,
- )
-
-
- import hashlib
- import io
- import json
- import socket
-
- import youtube_dl.YoutubeDL
- from youtube_dl.compat import (
- compat_http_client,
- compat_urllib_error,
- compat_HTTPError,
- )
- from youtube_dl.utils import (
- DownloadError,
- ExtractorError,
- format_bytes,
- UnavailableVideoError,
- )
- from youtube_dl.extractor import get_info_extractor
-
- RETRIES = 3
-
-
- class YoutubeDL(youtube_dl.YoutubeDL):
- def __init__(self, *args, **kwargs):
- self.to_stderr = self.to_screen
- self.processed_info_dicts = []
- super(YoutubeDL, self).__init__(*args, **kwargs)
-
- def report_warning(self, message):
- # Don't accept warnings during tests
- raise ExtractorError(message)
-
- def process_info(self, info_dict):
- self.processed_info_dicts.append(info_dict)
- return super(YoutubeDL, self).process_info(info_dict)
-
-
- def _file_md5(fn):
- with open(fn, 'rb') as f:
- return hashlib.md5(f.read()).hexdigest()
-
- defs = gettestcases()
-
-
- class TestDownload(unittest.TestCase):
- maxDiff = None
-
- def setUp(self):
- self.defs = defs
-
- # Dynamically generate tests
-
-
- def generator(test_case):
-
- def test_template(self):
- ie = youtube_dl.extractor.get_info_extractor(test_case['name'])
- other_ies = [get_info_extractor(ie_key) for ie_key in test_case.get('add_ie', [])]
- is_playlist = any(k.startswith('playlist') for k in test_case)
- test_cases = test_case.get(
- 'playlist', [] if is_playlist else [test_case])
-
- def print_skipping(reason):
- print('Skipping %s: %s' % (test_case['name'], reason))
- if not ie.working():
- print_skipping('IE marked as not _WORKING')
- return
-
- for tc in test_cases:
- info_dict = tc.get('info_dict', {})
- if not (info_dict.get('id') and info_dict.get('ext')):
- raise Exception('Test definition incorrect. The output file cannot be known. Are both \'id\' and \'ext\' keys present?')
-
- if 'skip' in test_case:
- print_skipping(test_case['skip'])
- return
- for other_ie in other_ies:
- if not other_ie.working():
- print_skipping('test depends on %sIE, marked as not WORKING' % other_ie.ie_key())
- return
-
- params = get_params(test_case.get('params', {}))
- if is_playlist and 'playlist' not in test_case:
- params.setdefault('extract_flat', 'in_playlist')
- params.setdefault('skip_download', True)
-
- ydl = YoutubeDL(params, auto_init=False)
- ydl.add_default_info_extractors()
- finished_hook_called = set()
-
- def _hook(status):
- if status['status'] == 'finished':
- finished_hook_called.add(status['filename'])
- ydl.add_progress_hook(_hook)
- expect_warnings(ydl, test_case.get('expected_warnings', []))
-
- def get_tc_filename(tc):
- return ydl.prepare_filename(tc.get('info_dict', {}))
-
- res_dict = None
-
- def try_rm_tcs_files(tcs=None):
- if tcs is None:
- tcs = test_cases
- for tc in tcs:
- tc_filename = get_tc_filename(tc)
- try_rm(tc_filename)
- try_rm(tc_filename + '.part')
- try_rm(os.path.splitext(tc_filename)[0] + '.info.json')
- try_rm_tcs_files()
- try:
- try_num = 1
- while True:
- try:
- # We're not using .download here sine that is just a shim
- # for outside error handling, and returns the exit code
- # instead of the result dict.
- res_dict = ydl.extract_info(
- test_case['url'],
- force_generic_extractor=params.get('force_generic_extractor', False))
- except (DownloadError, ExtractorError) as err:
- # Check if the exception is not a network related one
- if not err.exc_info[0] in (compat_urllib_error.URLError, socket.timeout, UnavailableVideoError, compat_http_client.BadStatusLine) or (err.exc_info[0] == compat_HTTPError and err.exc_info[1].code == 503):
- raise
-
- if try_num == RETRIES:
- report_warning('Failed due to network errors, skipping...')
- return
-
- print('Retrying: {0} failed tries\n\n##########\n\n'.format(try_num))
-
- try_num += 1
- else:
- break
-
- if is_playlist:
- self.assertTrue(res_dict['_type'] in ['playlist', 'multi_video'])
- self.assertTrue('entries' in res_dict)
- expect_info_dict(self, res_dict, test_case.get('info_dict', {}))
-
- if 'playlist_mincount' in test_case:
- assertGreaterEqual(
- self,
- len(res_dict['entries']),
- test_case['playlist_mincount'],
- 'Expected at least %d in playlist %s, but got only %d' % (
- test_case['playlist_mincount'], test_case['url'],
- len(res_dict['entries'])))
- if 'playlist_count' in test_case:
- self.assertEqual(
- len(res_dict['entries']),
- test_case['playlist_count'],
- 'Expected %d entries in playlist %s, but got %d.' % (
- test_case['playlist_count'],
- test_case['url'],
- len(res_dict['entries']),
- ))
- if 'playlist_duration_sum' in test_case:
- got_duration = sum(e['duration'] for e in res_dict['entries'])
- self.assertEqual(
- test_case['playlist_duration_sum'], got_duration)
-
- for tc in test_cases:
- 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)
- self.assertTrue(tc_filename in finished_hook_called)
- expected_minsize = tc.get('file_minsize', 10000)
- if expected_minsize is not None:
- if params.get('test'):
- expected_minsize = max(expected_minsize, 10000)
- got_fsize = os.path.getsize(tc_filename)
- assertGreaterEqual(
- self, got_fsize, expected_minsize,
- 'Expected %s to be at least %s, but it\'s only %s ' %
- (tc_filename, format_bytes(expected_minsize),
- format_bytes(got_fsize)))
- if 'md5' in tc:
- md5_for_file = _file_md5(tc_filename)
- self.assertEqual(md5_for_file, tc['md5'])
- 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()
- if is_playlist and res_dict is not None and res_dict.get('entries'):
- # Remove all other files that may have been extracted if the
- # extractor returns full results even with extract_flat
- res_tcs = [{'info_dict': e} for e in res_dict['entries']]
- try_rm_tcs_files(res_tcs)
-
- return test_template
-
- # 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.__name__ = str(tname)
- setattr(TestDownload, test_method.__name__, test_method)
- del test_method
-
-
- if __name__ == '__main__':
- unittest.main()
|