[sharesix] Add new extractor

This commit is contained in:
Naglis Jonaitis 2014-09-06 18:59:15 +03:00
parent e154762c74
commit 8e20f81c5b
2 changed files with 92 additions and 0 deletions

View file

@ -296,6 +296,7 @@ from .scivee import SciVeeIE
from .screencast import ScreencastIE
from .servingsys import ServingSysIE
from .shared import SharedIE
from .sharesix import ShareSixIE
from .sina import SinaIE
from .slideshare import SlideshareIE
from .slutload import SlutloadIE

View file

@ -0,0 +1,91 @@
# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import (
compat_urllib_parse,
compat_urllib_request,
parse_duration,
)
class ShareSixIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?sharesix\.com/(?:f/)?(?P<id>[0-9a-zA-Z]+)'
_TESTS = [
{
'url': 'http://sharesix.com/f/OXjQ7Y6',
'md5': '9e8e95d8823942815a7d7c773110cc93',
'info_dict': {
'id': 'OXjQ7Y6',
'ext': 'mp4',
'title': 'big_buck_bunny_480p_surround-fix.avi',
'duration': 596,
'width': 854,
'height': 480,
},
},
{
'url': 'http://sharesix.com/lfrwoxp35zdd',
'md5': 'dd19f1435b7cec2d7912c64beeee8185',
'info_dict': {
'id': 'lfrwoxp35zdd',
'ext': 'flv',
'title': 'WhiteBoard___a_Mac_vs_PC_Parody_Cartoon.mp4.flv',
'duration': 65,
'width': 1280,
'height': 720,
},
}
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
fields = {
'method_free': 'Free'
}
post = compat_urllib_parse.urlencode(fields)
req = compat_urllib_request.Request(url, post)
req.add_header('Content-type', 'application/x-www-form-urlencoded')
webpage = self._download_webpage(req, video_id,
'Downloading video page')
video_url = self._search_regex(
r"var\slnk1\s=\s'([^']+)'", webpage, 'video URL')
title = self._html_search_regex(
r'(?s)<dt>Filename:</dt>.+?<dd>(.+?)</dd>', webpage, 'title')
duration = parse_duration(
self._search_regex(
r'(?s)<dt>Length:</dt>.+?<dd>(.+?)</dd>',
webpage,
'duration',
fatal=False
)
)
m = re.search(
r'''(?xs)<dt>Width\sx\sHeight</dt>.+?
<dd>(?P<width>\d+)\sx\s(?P<height>\d+)</dd>''',
webpage
)
width = height = None
if m:
width, height = int(m.group('width')), int(m.group('height'))
formats = [{
'format_id': 'sd',
'url': video_url,
'width': width,
'height': height,
}]
return {
'id': video_id,
'title': title,
'duration': duration,
'formats': formats,
}