|
|
@ -2,7 +2,14 @@
|
|
|
|
from __future__ import unicode_literals
|
|
|
|
from __future__ import unicode_literals
|
|
|
|
|
|
|
|
|
|
|
|
from .common import InfoExtractor
|
|
|
|
from .common import InfoExtractor
|
|
|
|
from ..utils import int_or_none
|
|
|
|
from ..compat import (
|
|
|
|
|
|
|
|
compat_chr,
|
|
|
|
|
|
|
|
compat_ord,
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
from ..utils import (
|
|
|
|
|
|
|
|
int_or_none,
|
|
|
|
|
|
|
|
parse_filesize,
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class XMinusIE(InfoExtractor):
|
|
|
|
class XMinusIE(InfoExtractor):
|
|
|
@ -15,39 +22,46 @@ class XMinusIE(InfoExtractor):
|
|
|
|
'ext': 'mp3',
|
|
|
|
'ext': 'mp3',
|
|
|
|
'title': 'Леонид Агутин-Песенка шофера',
|
|
|
|
'title': 'Леонид Агутин-Песенка шофера',
|
|
|
|
'duration': 156,
|
|
|
|
'duration': 156,
|
|
|
|
|
|
|
|
'tbr': 320,
|
|
|
|
|
|
|
|
'filesize_approx': 5900000,
|
|
|
|
|
|
|
|
'view_count': int,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
def _real_extract(self, url):
|
|
|
|
def _real_extract(self, url):
|
|
|
|
video_id = self._match_id(url)
|
|
|
|
video_id = self._match_id(url)
|
|
|
|
|
|
|
|
|
|
|
|
# TODO more code goes here, for example ...
|
|
|
|
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
|
|
|
|
|
|
|
|
artist = self._html_search_regex(
|
|
|
|
artist = self._html_search_regex(
|
|
|
|
r'minus_track.artist="(.+?)"', webpage, 'artist')
|
|
|
|
r'minus_track\.artist="(.+?)"', webpage, 'artist')
|
|
|
|
title = artist + '-' + self._html_search_regex(
|
|
|
|
title = artist + '-' + self._html_search_regex(
|
|
|
|
r'minus_track.title="(.+?)"', webpage, 'title')
|
|
|
|
r'minus_track\.title="(.+?)"', webpage, 'title')
|
|
|
|
duration = int_or_none(self._html_search_regex(
|
|
|
|
duration = int_or_none(self._html_search_regex(
|
|
|
|
r'minus_track.dur_sec=\'([0-9]+?)\'', webpage, 'duration'))
|
|
|
|
r'minus_track\.dur_sec=\'([0-9]*?)\'',
|
|
|
|
|
|
|
|
webpage, 'duration', fatal=False))
|
|
|
|
|
|
|
|
filesize_approx = parse_filesize(self._html_search_regex(
|
|
|
|
|
|
|
|
r'<div class="filesize[^"]*"></div>\s*([0-9.]+\s*[a-zA-Z][bB])',
|
|
|
|
|
|
|
|
webpage, 'approximate filesize', fatal=False))
|
|
|
|
|
|
|
|
tbr = int_or_none(self._html_search_regex(
|
|
|
|
|
|
|
|
r'<div class="quality[^"]*"></div>\s*([0-9]+)\s*kbps',
|
|
|
|
|
|
|
|
webpage, 'bitrate', fatal=False))
|
|
|
|
|
|
|
|
view_count = int_or_none(self._html_search_regex(
|
|
|
|
|
|
|
|
r'<div class="quality.*?► ([0-9]+)',
|
|
|
|
|
|
|
|
webpage, 'view count', fatal=False))
|
|
|
|
|
|
|
|
|
|
|
|
enc_token = self._html_search_regex(
|
|
|
|
enc_token = self._html_search_regex(
|
|
|
|
r'data-mt="(.*?)"', webpage, 'enc_token')
|
|
|
|
r'data-mt="(.*?)"', webpage, 'enc_token')
|
|
|
|
token = self._decode_token(enc_token)
|
|
|
|
token = ''.join(
|
|
|
|
url = 'http://x-minus.org/dwlf/{}/{}.mp3'.format(video_id, token)
|
|
|
|
c if pos == 3 else compat_chr(compat_ord(c) - 1)
|
|
|
|
|
|
|
|
for pos, c in enumerate(reversed(enc_token)))
|
|
|
|
|
|
|
|
video_url = 'http://x-minus.org/dwlf/%s/%s.mp3' % (video_id, token)
|
|
|
|
|
|
|
|
|
|
|
|
return {
|
|
|
|
return {
|
|
|
|
'id': video_id,
|
|
|
|
'id': video_id,
|
|
|
|
'title': title,
|
|
|
|
'title': title,
|
|
|
|
'url': url,
|
|
|
|
'url': video_url,
|
|
|
|
'duration': duration,
|
|
|
|
'duration': duration,
|
|
|
|
|
|
|
|
'filesize_approx': filesize_approx,
|
|
|
|
|
|
|
|
'tbr': tbr,
|
|
|
|
|
|
|
|
'view_count': view_count,
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
def _decode_token(self, enc_token):
|
|
|
|
|
|
|
|
token = ''
|
|
|
|
|
|
|
|
pos = 0
|
|
|
|
|
|
|
|
for c in reversed(enc_token):
|
|
|
|
|
|
|
|
if pos != 3:
|
|
|
|
|
|
|
|
token += chr(ord(c) - 1)
|
|
|
|
|
|
|
|
else:
|
|
|
|
|
|
|
|
token += c
|
|
|
|
|
|
|
|
pos += 1
|
|
|
|
|
|
|
|
return token
|
|
|
|
|
|
|
|