Merge branch 'lenaten-walla'
This commit is contained in:
commit
b3826f6c8d
|
@ -15,6 +15,7 @@ from youtube_dl.extractor import (
|
|||
DailymotionIE,
|
||||
TEDIE,
|
||||
VimeoIE,
|
||||
WallaIE,
|
||||
)
|
||||
|
||||
|
||||
|
@ -279,5 +280,32 @@ class TestVimeoSubtitles(BaseTestSubtitles):
|
|||
self.assertTrue(subtitles.get(lang) is not None, u'Subtitles for \'%s\' not extracted' % lang)
|
||||
|
||||
|
||||
class TestWallsSubtitles(BaseTestSubtitles):
|
||||
url = 'http://vod.walla.co.il/movie/2705958/the-yes-men'
|
||||
IE = WallaIE
|
||||
|
||||
def test_list_subtitles(self):
|
||||
self.DL.expect_warning(u'Automatic Captions not supported by this server')
|
||||
self.DL.params['listsubtitles'] = True
|
||||
info_dict = self.getInfoDict()
|
||||
self.assertEqual(info_dict, None)
|
||||
|
||||
def test_allsubtitles(self):
|
||||
self.DL.expect_warning(u'Automatic Captions not supported by this server')
|
||||
self.DL.params['writesubtitles'] = True
|
||||
self.DL.params['allsubtitles'] = True
|
||||
subtitles = self.getSubtitles()
|
||||
self.assertEqual(set(subtitles.keys()), set(['heb']))
|
||||
self.assertEqual(md5(subtitles['heb']), 'e758c5d7cb982f6bef14f377ec7a3920')
|
||||
|
||||
def test_nosubtitles(self):
|
||||
self.DL.expect_warning(u'video doesn\'t have subtitles')
|
||||
self.url = 'http://vod.walla.co.il/movie/2642630/one-direction-all-for-one'
|
||||
self.DL.params['writesubtitles'] = True
|
||||
self.DL.params['allsubtitles'] = True
|
||||
subtitles = self.getSubtitles()
|
||||
self.assertEqual(len(subtitles), 0)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
unittest.main()
|
||||
|
|
|
@ -440,6 +440,7 @@ from .vporn import VpornIE
|
|||
from .vube import VubeIE
|
||||
from .vuclip import VuClipIE
|
||||
from .vulture import VultureIE
|
||||
from .walla import WallaIE
|
||||
from .washingtonpost import WashingtonPostIE
|
||||
from .wat import WatIE
|
||||
from .wayofthemaster import WayOfTheMasterIE
|
||||
|
|
|
@ -0,0 +1,89 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .subtitles import SubtitlesInfoExtractor
|
||||
from ..utils import (
|
||||
xpath_text,
|
||||
int_or_none,
|
||||
)
|
||||
|
||||
|
||||
class WallaIE(SubtitlesInfoExtractor):
|
||||
_VALID_URL = r'http://vod\.walla\.co\.il/[^/]+/(?P<id>\d+)/(?P<display_id>.+)'
|
||||
_TEST = {
|
||||
'url': 'http://vod.walla.co.il/movie/2642630/one-direction-all-for-one',
|
||||
'info_dict': {
|
||||
'id': '2642630',
|
||||
'display_id': 'one-direction-all-for-one',
|
||||
'ext': 'flv',
|
||||
'title': 'וואן דיירקשן: ההיסטריה',
|
||||
'description': 'md5:de9e2512a92442574cdb0913c49bc4d8',
|
||||
'thumbnail': 're:^https?://.*\.jpg',
|
||||
'duration': 3600,
|
||||
},
|
||||
'params': {
|
||||
# rtmp download
|
||||
'skip_download': True,
|
||||
}
|
||||
}
|
||||
|
||||
_SUBTITLE_LANGS = {
|
||||
'עברית': 'heb',
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
mobj = re.match(self._VALID_URL, url)
|
||||
video_id = mobj.group('id')
|
||||
display_id = mobj.group('display_id')
|
||||
|
||||
video = self._download_xml(
|
||||
'http://video2.walla.co.il/?w=null/null/%s/@@/video/flv_pl' % video_id,
|
||||
display_id)
|
||||
|
||||
item = video.find('./items/item')
|
||||
|
||||
title = xpath_text(item, './title', 'title')
|
||||
description = xpath_text(item, './synopsis', 'description')
|
||||
thumbnail = xpath_text(item, './preview_pic', 'thumbnail')
|
||||
duration = int_or_none(xpath_text(item, './duration', 'duration'))
|
||||
|
||||
subtitles = {}
|
||||
for subtitle in item.findall('./subtitles/subtitle'):
|
||||
lang = xpath_text(subtitle, './title')
|
||||
subtitles[self._SUBTITLE_LANGS.get(lang, lang)] = xpath_text(subtitle, './src')
|
||||
|
||||
if self._downloader.params.get('listsubtitles', False):
|
||||
self._list_available_subtitles(video_id, subtitles)
|
||||
return
|
||||
|
||||
subtitles = self.extract_subtitles(video_id, subtitles)
|
||||
|
||||
formats = []
|
||||
for quality in item.findall('./qualities/quality'):
|
||||
format_id = xpath_text(quality, './title')
|
||||
fmt = {
|
||||
'url': 'rtmp://wafla.walla.co.il/vod',
|
||||
'play_path': xpath_text(quality, './src'),
|
||||
'player_url': 'http://isc.walla.co.il/w9/swf/video_swf/vod/WallaMediaPlayerAvod.swf',
|
||||
'page_url': url,
|
||||
'ext': 'flv',
|
||||
'format_id': xpath_text(quality, './title'),
|
||||
}
|
||||
m = re.search(r'^(?P<height>\d+)[Pp]', format_id)
|
||||
if m:
|
||||
fmt['height'] = int(m.group('height'))
|
||||
formats.append(fmt)
|
||||
self._sort_formats(formats)
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'display_id': display_id,
|
||||
'title': title,
|
||||
'description': description,
|
||||
'thumbnail': thumbnail,
|
||||
'duration': duration,
|
||||
'formats': formats,
|
||||
'subtitles': subtitles,
|
||||
}
|
Loading…
Reference in New Issue