mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-30 10:22:55 +01:00
Compare commits
2 Commits
99de2f38d3
...
f9f9699f2f
Author | SHA1 | Date | |
---|---|---|---|
|
f9f9699f2f | ||
|
a3cf22e590 |
@ -1,16 +1,25 @@
|
||||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import random
|
||||
import re
|
||||
import string
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
ExtractorError,
|
||||
int_or_none,
|
||||
mimetype2ext,
|
||||
parse_codecs,
|
||||
update_url_query,
|
||||
xpath_element,
|
||||
xpath_text,
|
||||
)
|
||||
from ..compat import (
|
||||
compat_b64decode,
|
||||
compat_ord,
|
||||
compat_struct_pack,
|
||||
)
|
||||
|
||||
|
||||
class VideaIE(InfoExtractor):
|
||||
@ -19,7 +28,7 @@ class VideaIE(InfoExtractor):
|
||||
videa(?:kid)?\.hu/
|
||||
(?:
|
||||
videok/(?:[^/]+/)*[^?#&]+-|
|
||||
player\?.*?\bv=|
|
||||
(?:videojs_)?player\?.*?\bv=|
|
||||
player/v/
|
||||
)
|
||||
(?P<id>[^?#&]+)
|
||||
@ -53,6 +62,7 @@ class VideaIE(InfoExtractor):
|
||||
'url': 'https://videakid.hu/player/v/8YfIAjxwWGwT8HVQ?autoplay=1',
|
||||
'only_matching': True,
|
||||
}]
|
||||
_STATIC_SECRET = 'xHb0ZvME5q8CBcoQi6AngerDu3FGO9fkUlwPmLVY_RTzj2hJIS4NasXWKy1td7p'
|
||||
|
||||
@staticmethod
|
||||
def _extract_urls(webpage):
|
||||
@ -60,26 +70,84 @@ class VideaIE(InfoExtractor):
|
||||
r'<iframe[^>]+src=(["\'])(?P<url>(?:https?:)?//videa\.hu/player\?.*?\bv=.+?)\1',
|
||||
webpage)]
|
||||
|
||||
@staticmethod
|
||||
def rc4(cipher_text, key):
|
||||
res = b''
|
||||
|
||||
key_len = len(key)
|
||||
S = list(range(256))
|
||||
|
||||
j = 0
|
||||
for i in range(256):
|
||||
j = (j + S[i] + ord(key[i % key_len])) % 256
|
||||
S[i], S[j] = S[j], S[i]
|
||||
|
||||
i = 0
|
||||
j = 0
|
||||
for m in range(len(cipher_text)):
|
||||
i = (i + 1) % 256
|
||||
j = (j + S[i]) % 256
|
||||
S[i], S[j] = S[j], S[i]
|
||||
k = S[(S[i] + S[j]) % 256]
|
||||
res += compat_struct_pack('B', k ^ compat_ord(cipher_text[m]))
|
||||
|
||||
return res.decode()
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
query = {'v': video_id}
|
||||
player_page = self._download_webpage(
|
||||
'https://videa.hu/player', video_id, query=query)
|
||||
|
||||
info = self._download_xml(
|
||||
'http://videa.hu/videaplayer_get_xml.php', video_id,
|
||||
query={'v': video_id})
|
||||
nonce = self._search_regex(
|
||||
r'_xt\s*=\s*"([^"]+)"', player_page, 'nonce')
|
||||
l = nonce[:32]
|
||||
s = nonce[32:]
|
||||
result = ''
|
||||
for i in range(0, 32):
|
||||
result += s[i - (self._STATIC_SECRET.index(l[i]) - 31)]
|
||||
|
||||
video = xpath_element(info, './/video', 'video', fatal=True)
|
||||
sources = xpath_element(info, './/video_sources', 'sources', fatal=True)
|
||||
random_seed = ''.join(random.choice(string.ascii_letters + string.digits) for _ in range(8))
|
||||
query['_s'] = random_seed
|
||||
query['_t'] = result[:16]
|
||||
|
||||
b64_info, handle = self._download_webpage_handle(
|
||||
'http://videa.hu/videaplayer_get_xml.php', video_id, query=query)
|
||||
if b64_info.startswith('<?xml'):
|
||||
info = self._parse_xml(b64_info, video_id)
|
||||
else:
|
||||
key = result[16:] + random_seed + handle.headers['x-videa-xs']
|
||||
info = self._parse_xml(self.rc4(
|
||||
compat_b64decode(b64_info), key), video_id)
|
||||
|
||||
video = xpath_element(info, './video', 'video')
|
||||
if not video:
|
||||
raise ExtractorError(xpath_element(
|
||||
info, './error', fatal=True), expected=True)
|
||||
sources = xpath_element(
|
||||
info, './video_sources', 'sources', fatal=True)
|
||||
hash_values = xpath_element(
|
||||
info, './hash_values', 'hash values', fatal=True)
|
||||
|
||||
title = xpath_text(video, './title', fatal=True)
|
||||
|
||||
formats = []
|
||||
for source in sources.findall('./video_source'):
|
||||
source_url = source.text
|
||||
if not source_url:
|
||||
source_name = source.get('name')
|
||||
source_exp = source.get('exp')
|
||||
if not (source_url and source_name and source_exp):
|
||||
continue
|
||||
hash_value = xpath_text(hash_values, 'hash_value_' + source_name)
|
||||
if not hash_value:
|
||||
continue
|
||||
source_url = update_url_query(source_url, {
|
||||
'md5': hash_value,
|
||||
'expires': source_exp,
|
||||
})
|
||||
f = parse_codecs(source.get('codecs'))
|
||||
f.update({
|
||||
'url': source_url,
|
||||
'url': self._proto_relative_url(source_url),
|
||||
'ext': mimetype2ext(source.get('mimetype')) or 'mp4',
|
||||
'format_id': source.get('name'),
|
||||
'width': int_or_none(source.get('width')),
|
||||
@ -88,8 +156,7 @@ class VideaIE(InfoExtractor):
|
||||
formats.append(f)
|
||||
self._sort_formats(formats)
|
||||
|
||||
thumbnail = xpath_text(video, './poster_src')
|
||||
duration = int_or_none(xpath_text(video, './duration'))
|
||||
thumbnail = self._proto_relative_url(xpath_text(video, './poster_src'))
|
||||
|
||||
age_limit = None
|
||||
is_adult = xpath_text(video, './is_adult_content', default=None)
|
||||
@ -100,7 +167,7 @@ class VideaIE(InfoExtractor):
|
||||
'id': video_id,
|
||||
'title': title,
|
||||
'thumbnail': thumbnail,
|
||||
'duration': duration,
|
||||
'duration': int_or_none(xpath_text(video, './duration')),
|
||||
'age_limit': age_limit,
|
||||
'formats': formats,
|
||||
}
|
||||
|
Loading…
Reference in New Issue
Block a user