Merge pull request #3691 from naglis/moevideo

[moevideo] Add new extractor
pull/8/head
Sergey M. 10 years ago
commit 759c6293bd

@ -200,6 +200,7 @@ from .mitele import MiTeleIE
from .mixcloud import MixcloudIE
from .mlb import MLBIE
from .mpora import MporaIE
from .moevideo import MoeVideoIE
from .mofosex import MofosexIE
from .mojvideo import MojvideoIE
from .mooshare import MooshareIE

@ -0,0 +1,112 @@
# coding: utf-8
from __future__ import unicode_literals
import json
import re
from .common import InfoExtractor
from ..utils import (
ExtractorError,
compat_urllib_parse,
compat_urllib_request,
int_or_none,
)
class MoeVideoIE(InfoExtractor):
IE_DESC = 'moevideo.net and playreplay.net'
_VALID_URL = r'''(?x)
https?://(?P<host>(?:www\.)?
(?:moevideo\.net|playreplay\.net))/
(?:video|framevideo)/(?P<id>[0-9]+\.[0-9A-Za-z]+)'''
_API_URL = 'http://api.letitbit.net/'
_API_KEY = 'tVL0gjqo5'
_TESTS = [
{
'url': 'http://moevideo.net/video/00297.0036103fe3d513ef27915216fd29',
'md5': '129f5ae1f6585d0e9bb4f38e774ffb3a',
'info_dict': {
'id': '00297.0036103fe3d513ef27915216fd29',
'ext': 'flv',
'title': 'Sink cut out machine',
'description': 'md5:f29ff97b663aefa760bf7ca63c8ca8a8',
'thumbnail': 're:^https?://.*\.jpg$',
'width': 540,
'height': 360,
'duration': 179,
'filesize_approx': 17822500,
}
},
{
'url': 'http://playreplay.net/video/77107.7f325710a627383d40540d8e991a',
'md5': '74f0a014d5b661f0f0e2361300d1620e',
'info_dict': {
'id': '77107.7f325710a627383d40540d8e991a',
'ext': 'flv',
'title': 'Operacion Condor.',
'description': 'md5:7e68cb2fcda66833d5081c542491a9a3',
'thumbnail': 're:^https?://.*\.jpg$',
'width': 480,
'height': 296,
'duration': 6027,
'filesize_approx': 588257923,
}
},
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
webpage = self._download_webpage(
'http://%s/video/%s' % (mobj.group('host'), video_id),
video_id, 'Downloading webpage')
title = self._og_search_title(webpage)
thumbnail = self._og_search_thumbnail(webpage)
description = self._og_search_description(webpage)
r = [
self._API_KEY,
[
'preview/flv_link',
{
'uid': video_id,
},
],
]
r_json = json.dumps(r)
post = compat_urllib_parse.urlencode({'r': r_json})
req = compat_urllib_request.Request(self._API_URL, post)
req.add_header('Content-type', 'application/x-www-form-urlencoded')
response = self._download_json(req, video_id)
if response['status'] != 'OK':
raise ExtractorError(
'%s returned error: %s' % (self.IE_NAME, response['data']),
expected=True
)
item = response['data'][0]
video_url = item['link']
duration = int_or_none(item['length'])
width = int_or_none(item['width'])
height = int_or_none(item['height'])
filesize = int_or_none(item['convert_size'])
formats = [{
'format_id': 'sd',
'http_headers': {'Range': 'bytes=0-'}, # Required to download
'url': video_url,
'width': width,
'height': height,
'filesize_approx': filesize,
}]
return {
'id': video_id,
'title': title,
'thumbnail': thumbnail,
'description': description,
'duration': duration,
'formats': formats,
}
Loading…
Cancel
Save