1
0
mirror of https://github.com/l1ving/youtube-dl synced 2024-11-20 20:12:52 +08:00

Adapt test_download to support playlists, and remove race conditions

This commit is contained in:
Philipp Hagemeister 2013-01-01 19:30:29 +01:00
parent 6985325e01
commit 5c892b0ba9
3 changed files with 54 additions and 29 deletions

View File

@ -35,6 +35,6 @@
"username": null,
"verbose": true,
"writedescription": false,
"writeinfojson": false,
"writeinfojson": true,
"writesubtitles": false
}

View File

@ -1,5 +1,6 @@
#!/usr/bin/env python
import errno
import hashlib
import io
import os
@ -26,6 +27,14 @@ proxy_handler = compat_urllib_request.ProxyHandler()
opener = compat_urllib_request.build_opener(proxy_handler, cookie_processor, YoutubeDLHandler())
compat_urllib_request.install_opener(opener)
def _try_rm(filename):
""" Remove a file if it exists """
try:
os.remove(filename)
except OSError as ose:
if ose.errno != errno.ENOENT:
raise
class FileDownloader(youtube_dl.FileDownloader):
def __init__(self, *args, **kwargs):
self.to_stderr = self.to_screen
@ -50,15 +59,6 @@ class TestDownload(unittest.TestCase):
self.parameters = parameters
self.defs = defs
# Clear old files
self.tearDown()
def tearDown(self):
for fn in [ test.get('file', False) for test in self.defs ]:
if fn and os.path.exists(fn):
os.remove(fn)
### Dynamically generate tests
def generator(test_case):
@ -67,7 +67,7 @@ def generator(test_case):
if not ie._WORKING:
print('Skipping: IE marked as not _WORKING')
return
if not test_case['file']:
if 'playlist' not in test_case and not test_case['file']:
print('Skipping: No output file specified')
return
if 'skip' in test_case:
@ -82,19 +82,32 @@ def generator(test_case):
fd.add_info_extractor(ie())
for ien in test_case.get('add_ie', []):
fd.add_info_extractor(getattr(youtube_dl.InfoExtractors, ien + 'IE')())
test_cases = test_case.get('playlist', [test_case])
for tc in test_cases:
_try_rm(tc['file'])
_try_rm(tc['file'] + '.info.json')
try:
fd.download([test_case['url']])
self.assertTrue(os.path.exists(test_case['file']))
if 'md5' in test_case:
md5_for_file = _file_md5(test_case['file'])
self.assertEqual(md5_for_file, test_case['md5'])
info_dict = fd.processed_info_dicts[0]
for (info_field, value) in test_case.get('info_dict', {}).items():
for tc in test_cases:
self.assertTrue(os.path.exists(tc['file']))
self.assertTrue(os.path.exists(tc['file'] + '.info.json'))
if 'md5' in tc:
md5_for_file = _file_md5(tc['file'])
self.assertEqual(md5_for_file, tc['md5'])
with io.open(tc['file'] + '.info.json', encoding='utf-8') as infof:
info_dict = json.load(infof)
for (info_field, value) in tc.get('info_dict', {}).items():
if value.startswith('md5:'):
md5_info_value = hashlib.md5(info_dict.get(info_field, '')).hexdigest()
self.assertEqual(value[3:], md5_info_value)
else:
self.assertEqual(value, info_dict.get(info_field))
finally:
for tc in test_cases:
_try_rm(tc['file'])
_try_rm(tc['file'] + '.info.json')
return test_template

View File

@ -124,16 +124,28 @@
{
"name": "Steam",
"url": "http://store.steampowered.com/video/105600/",
"playlist": [
{
"file": "81300.flv",
"md5": "f870007cee7065d7c76b88f0a45ecc07",
"info_dict": {
"title": "Terraria 1.1 Trailer"
}
},
{
"file": "80859.flv",
"md5": "61aaf31a5c5c3041afb58fb83cbb5751",
"info_dict": {
"title": "Terraria Trailer"
}
}
]
},
{
"name": "Ustream",
"url": "http://www.ustream.tv/recorded/20274954",
"files": [["20274954.flv", "088f151799e8f572f84eb62f17d73e5c" ]],
"file": "20274954.flv",
"md5": "088f151799e8f572f84eb62f17d73e5c",
"info_dict": {
"title": "Young Americans for Liberty February 7, 2012 2:28 AM"
}