mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-12-20 12:04:03 +00:00
[miomio] Simplify and fix python 2.6 issue
This commit is contained in:
parent
d08225edf4
commit
f5b669113f
@ -1,74 +1,93 @@
|
|||||||
# coding: utf-8
|
# coding: utf-8
|
||||||
from __future__ import unicode_literals
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
import random
|
||||||
|
|
||||||
from .common import InfoExtractor
|
from .common import InfoExtractor
|
||||||
|
from ..utils import (
|
||||||
|
xpath_text,
|
||||||
|
int_or_none,
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
class MiomioTvIE(InfoExtractor):
|
class MiomioTvIE(InfoExtractor):
|
||||||
IE_NAME = 'miomio.tv'
|
IE_NAME = 'miomio.tv'
|
||||||
_VALID_URL = r'https?://(?:www\.)?miomio\.tv/watch/cc(?P<id>[0-9]+)'
|
_VALID_URL = r'https?://(?:www\.)?miomio\.tv/watch/cc(?P<id>[0-9]+)'
|
||||||
_TEST = {
|
_TESTS = [{
|
||||||
'url': 'http://www.miomio.tv/watch/cc179734/',
|
'url': 'http://www.miomio.tv/watch/cc179734/',
|
||||||
'md5': '48de02137d0739c15b440a224ad364b9',
|
'md5': '48de02137d0739c15b440a224ad364b9',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '179734',
|
'id': '179734',
|
||||||
'title': u'\u624b\u7ed8\u52a8\u6f2b\u9b3c\u6ce3\u4f46\u4e01\u5168\u7a0b\u753b\u6cd5',
|
'ext': 'flv',
|
||||||
'ext': 'flv'
|
'title': '手绘动漫鬼泣但丁全程画法',
|
||||||
}
|
'duration': 354,
|
||||||
}
|
},
|
||||||
|
}, {
|
||||||
|
'url': 'http://www.miomio.tv/watch/cc184024/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '43729',
|
||||||
|
'title': '《动漫同人插画绘制》',
|
||||||
|
},
|
||||||
|
'playlist_mincount': 86,
|
||||||
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
webpage = self._download_webpage(url, video_id)
|
webpage = self._download_webpage(url, video_id)
|
||||||
|
|
||||||
title = self._html_search_regex(r'<meta\s+name="description"\s+content="\s*([^"]*)\s*"', webpage, 'title')
|
title = self._html_search_meta(
|
||||||
ref_path = self._search_regex(r'src="(/mioplayer/.*?)"', webpage, 'ref_path')
|
'description', webpage, 'title', fatal=True)
|
||||||
referer = 'http://www.miomio.tv{0}'.format(ref_path)
|
|
||||||
xml_config = self._search_regex(r'flashvars="type=sina&(.*?)&cid=', webpage, 'xml config')
|
mioplayer_path = self._search_regex(
|
||||||
|
r'src="(/mioplayer/[^"]+)"', webpage, 'ref_path')
|
||||||
|
|
||||||
|
xml_config = self._search_regex(
|
||||||
|
r'flashvars="type=sina&(.+?)&',
|
||||||
|
webpage, 'xml config')
|
||||||
|
|
||||||
# skipping the following page causes lags and eventually connection drop-outs
|
# skipping the following page causes lags and eventually connection drop-outs
|
||||||
# id is normally a rotating three digit value but a fixed value always appears to work
|
self._request_webpage(
|
||||||
self._request_webpage("http://www.miomio.tv/mioplayer/mioplayerconfigfiles/xml.php?id={0}&r=cc{1}".format(id, 945), video_id)
|
'http://www.miomio.tv/mioplayer/mioplayerconfigfiles/xml.php?id=%s&r=%s' % (id, random.randint(100, 999)),
|
||||||
|
video_id)
|
||||||
|
|
||||||
# the following xml contains the actual configuration information on the video file(s)
|
# the following xml contains the actual configuration information on the video file(s)
|
||||||
xml_url = 'http://www.miomio.tv/mioplayer/mioplayerconfigfiles/sina.php?{0}'.format(xml_config)
|
vid_config = self._download_xml(
|
||||||
vidconfig = self._download_xml(xml_url, video_id)
|
'http://www.miomio.tv/mioplayer/mioplayerconfigfiles/sina.php?{0}'.format(xml_config),
|
||||||
|
video_id)
|
||||||
file_els = vidconfig.findall('.//durl')
|
|
||||||
|
|
||||||
entries = []
|
|
||||||
|
|
||||||
for file_el in file_els:
|
|
||||||
segment_id = file_el.find('order').text.strip()
|
|
||||||
segment_title = '_'.join([title, segment_id])
|
|
||||||
segment_duration = file_el.find('length').text.strip()
|
|
||||||
segment_url = file_el.find('url').text.strip()
|
|
||||||
|
|
||||||
entries.append({
|
|
||||||
'id': segment_id,
|
|
||||||
'title': segment_title,
|
|
||||||
'duration': segment_duration,
|
|
||||||
'url': segment_url
|
|
||||||
})
|
|
||||||
|
|
||||||
http_headers = {
|
http_headers = {
|
||||||
'Referer': referer,
|
'Referer': 'http://www.miomio.tv%s' % mioplayer_path,
|
||||||
'Accept-Encoding': 'gzip, deflate',
|
|
||||||
'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8'
|
|
||||||
}
|
}
|
||||||
|
|
||||||
|
entries = []
|
||||||
|
for f in vid_config.findall('./durl'):
|
||||||
|
segment_url = xpath_text(f, 'url', 'video url')
|
||||||
|
if not segment_url:
|
||||||
|
continue
|
||||||
|
order = xpath_text(f, 'order', 'order')
|
||||||
|
segment_id = video_id
|
||||||
|
segment_title = title
|
||||||
|
if order:
|
||||||
|
segment_id += '-%s' % order
|
||||||
|
segment_title += ' part %s' % order
|
||||||
|
entries.append({
|
||||||
|
'id': segment_id,
|
||||||
|
'url': segment_url,
|
||||||
|
'title': segment_title,
|
||||||
|
'duration': int_or_none(xpath_text(f, 'length', 'duration'), 1000),
|
||||||
|
'http_headers': http_headers,
|
||||||
|
})
|
||||||
|
|
||||||
if len(entries) == 1:
|
if len(entries) == 1:
|
||||||
return {
|
segment = entries[0]
|
||||||
'id': video_id,
|
segment['id'] = video_id
|
||||||
'title': title,
|
segment['title'] = title
|
||||||
'url': entries[0]['url'],
|
return segment
|
||||||
'http_headers': http_headers
|
|
||||||
}
|
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'_type': 'multi_video',
|
'_type': 'multi_video',
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'title': title,
|
|
||||||
'entries': entries,
|
'entries': entries,
|
||||||
'http_headers': http_headers
|
'title': title,
|
||||||
|
'http_headers': http_headers,
|
||||||
}
|
}
|
||||||
|
Loading…
Reference in New Issue
Block a user