mirror of
https://codeberg.org/polarisfm/youtube-dl
synced 2024-12-27 00:37:55 +01:00
Merge branch 'MLB' of https://github.com/chaochichen/youtube-dl into chaochichen-MLB
This commit is contained in:
commit
1aa42fedee
@ -170,6 +170,7 @@ from .metacafe import MetacafeIE
|
||||
from .metacritic import MetacriticIE
|
||||
from .mit import TechTVMITIE, MITIE, OCWMITIE
|
||||
from .mixcloud import MixcloudIE
|
||||
from .mlb import MlbIE
|
||||
from .mpora import MporaIE
|
||||
from .mofosex import MofosexIE
|
||||
from .mooshare import MooshareIE
|
||||
|
74
youtube_dl/extractor/mlb.py
Normal file
74
youtube_dl/extractor/mlb.py
Normal file
@ -0,0 +1,74 @@
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
|
||||
|
||||
class MlbIE(InfoExtractor):
|
||||
_VALID_URL = r'http?://m\.mlb\.com/video/topic/[0-9]+/v(?P<id>n?\d+)/.*$'
|
||||
_TEST = {
|
||||
'url': 'http://m.mlb.com/video/topic/81536970/v34496663/mianym-stanton-practices-for-the-home-run-derby',
|
||||
'md5': u'd9c022c10d21f849f49c05ae12a8a7e9',
|
||||
'info_dict': {
|
||||
'id': '34496663',
|
||||
'ext': 'mp4',
|
||||
'format': 'mp4',
|
||||
'description': "7/11/14: Giancarlo Stanton practices for the Home Run Derby prior to the game against the Mets",
|
||||
'title': "Stanton prepares for Derby",
|
||||
},
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
mobj = re.match(self._VALID_URL, url)
|
||||
video_id = mobj.group('id')
|
||||
|
||||
webpage = self._download_webpage(url, video_id)
|
||||
|
||||
title = self._og_search_title(webpage, default=video_id)
|
||||
description = self._html_search_regex(r'<meta name="description" (?:content|value)="(.*?)"/>', webpage, 'description', fatal=False)
|
||||
thumbnail = self._html_search_regex(r'<meta itemprop="image" (?:content|value)="(.*?)" />', webpage, 'image', fatal=False)
|
||||
|
||||
# use the video_id to find the Media detail XML
|
||||
id_len = len(video_id)
|
||||
_mediadetail_url = 'http://m.mlb.com/gen/multimedia/detail/'+video_id[id_len-3]+'/'+video_id[id_len-2]+'/'+video_id[id_len-1]+'/'+video_id+'.xml'
|
||||
|
||||
mediadetails = self._download_xml(_mediadetail_url, video_id, "Downloading media detail...")
|
||||
has1500K = 0
|
||||
has1200K = 0
|
||||
has600K = 0
|
||||
# loop through the list of url's and only get the highest quality MP4 content
|
||||
for element in mediadetails.findall('url'):
|
||||
scenario = element.attrib['playback_scenario']
|
||||
if scenario.startswith(u'FLASH'):
|
||||
if scenario.startswith(u'FLASH_1800K'):
|
||||
video_url = element.text
|
||||
# 1800K is the current highest quality video on MLB.com
|
||||
break
|
||||
else:
|
||||
if scenario.startswith(u'FLASH_1500K'):
|
||||
video_url = element.text
|
||||
has1500K = 1
|
||||
else:
|
||||
if (scenario.startswith(u'FLASH_1200K') and not has1500K):
|
||||
video_url = element.text
|
||||
has1200K = 1
|
||||
else:
|
||||
if (scenario.startswith(u'FLASH_600K') and not has1200K):
|
||||
video_url = element.text
|
||||
has600K = 1
|
||||
else:
|
||||
if (scenario.startswith(u'FLASH_300K') and not has600K):
|
||||
video_url = element.text
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'url': video_url,
|
||||
'extractor': 'mlb',
|
||||
'webpage_url': url,
|
||||
'title': title,
|
||||
'ext': 'mp4',
|
||||
'format': 'mp4',
|
||||
'description': description,
|
||||
'thumbnail': thumbnail,
|
||||
}
|
Loading…
Reference in New Issue
Block a user