mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-12-15 05:37:36 +01:00
[openload] Fix extraction (closes #9706)
This commit is contained in:
parent
d3f8e038fe
commit
c1decda58c
@ -1,15 +1,14 @@
|
|||||||
# coding: utf-8
|
# coding: utf-8
|
||||||
from __future__ import unicode_literals
|
from __future__ import unicode_literals, division
|
||||||
|
|
||||||
import re
|
import math
|
||||||
|
|
||||||
from .common import InfoExtractor
|
from .common import InfoExtractor
|
||||||
from ..compat import compat_chr
|
from ..compat import compat_chr
|
||||||
from ..utils import (
|
from ..utils import (
|
||||||
|
decode_png,
|
||||||
determine_ext,
|
determine_ext,
|
||||||
encode_base_n,
|
|
||||||
ExtractorError,
|
ExtractorError,
|
||||||
mimetype2ext,
|
|
||||||
)
|
)
|
||||||
|
|
||||||
|
|
||||||
@ -41,60 +40,6 @@ class OpenloadIE(InfoExtractor):
|
|||||||
'only_matching': True,
|
'only_matching': True,
|
||||||
}]
|
}]
|
||||||
|
|
||||||
@staticmethod
|
|
||||||
def openload_level2_debase(m):
|
|
||||||
radix, num = int(m.group(1)) + 27, int(m.group(2))
|
|
||||||
return '"' + encode_base_n(num, radix) + '"'
|
|
||||||
|
|
||||||
@classmethod
|
|
||||||
def openload_level2(cls, txt):
|
|
||||||
# The function name is ǃ \u01c3
|
|
||||||
# Using escaped unicode literals does not work in Python 3.2
|
|
||||||
return re.sub(r'ǃ\((\d+),(\d+)\)', cls.openload_level2_debase, txt, re.UNICODE).replace('"+"', '')
|
|
||||||
|
|
||||||
# Openload uses a variant of aadecode
|
|
||||||
# openload_decode and related functions are originally written by
|
|
||||||
# vitas@matfyz.cz and released with public domain
|
|
||||||
# See https://github.com/rg3/youtube-dl/issues/8489
|
|
||||||
@classmethod
|
|
||||||
def openload_decode(cls, txt):
|
|
||||||
symbol_table = [
|
|
||||||
('_', '(゚Д゚) [゚Θ゚]'),
|
|
||||||
('a', '(゚Д゚) [゚ω゚ノ]'),
|
|
||||||
('b', '(゚Д゚) [゚Θ゚ノ]'),
|
|
||||||
('c', '(゚Д゚) [\'c\']'),
|
|
||||||
('d', '(゚Д゚) [゚ー゚ノ]'),
|
|
||||||
('e', '(゚Д゚) [゚Д゚ノ]'),
|
|
||||||
('f', '(゚Д゚) [1]'),
|
|
||||||
|
|
||||||
('o', '(゚Д゚) [\'o\']'),
|
|
||||||
('u', '(o゚ー゚o)'),
|
|
||||||
('c', '(゚Д゚) [\'c\']'),
|
|
||||||
|
|
||||||
('7', '((゚ー゚) + (o^_^o))'),
|
|
||||||
('6', '((o^_^o) +(o^_^o) +(c^_^o))'),
|
|
||||||
('5', '((゚ー゚) + (゚Θ゚))'),
|
|
||||||
('4', '(-~3)'),
|
|
||||||
('3', '(-~-~1)'),
|
|
||||||
('2', '(-~1)'),
|
|
||||||
('1', '(-~0)'),
|
|
||||||
('0', '((c^_^o)-(c^_^o))'),
|
|
||||||
]
|
|
||||||
delim = '(゚Д゚)[゚ε゚]+'
|
|
||||||
ret = ''
|
|
||||||
for aachar in txt.split(delim):
|
|
||||||
for val, pat in symbol_table:
|
|
||||||
aachar = aachar.replace(pat, val)
|
|
||||||
aachar = aachar.replace('+ ', '')
|
|
||||||
m = re.match(r'^\d+', aachar)
|
|
||||||
if m:
|
|
||||||
ret += compat_chr(int(m.group(0), 8))
|
|
||||||
else:
|
|
||||||
m = re.match(r'^u([\da-f]+)', aachar)
|
|
||||||
if m:
|
|
||||||
ret += compat_chr(int(m.group(1), 16))
|
|
||||||
return cls.openload_level2(ret)
|
|
||||||
|
|
||||||
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)
|
||||||
@ -102,29 +47,77 @@ class OpenloadIE(InfoExtractor):
|
|||||||
if 'File not found' in webpage:
|
if 'File not found' in webpage:
|
||||||
raise ExtractorError('File not found', expected=True)
|
raise ExtractorError('File not found', expected=True)
|
||||||
|
|
||||||
code = self._search_regex(
|
# The following extraction logic is proposed by @Belderak and @gdkchan
|
||||||
r'</video>\s*</div>\s*<script[^>]+>[^>]+</script>\s*<script[^>]+>([^<]+)</script>',
|
# and declared to be used freely in youtube-dl
|
||||||
webpage, 'JS code')
|
# See https://github.com/rg3/youtube-dl/issues/9706
|
||||||
|
|
||||||
decoded = self.openload_decode(code)
|
numbers_js = self._download_webpage(
|
||||||
|
'https://openload.co/assets/js/obfuscator/n.js', video_id,
|
||||||
|
note='Downloading signature numbers')
|
||||||
|
signums = self._search_regex(
|
||||||
|
r'window\.signatureNumbers\s*=\s*[\'"](?P<data>[a-z]+)[\'"]',
|
||||||
|
numbers_js, 'signature numbers', group='data')
|
||||||
|
|
||||||
video_url = self._search_regex(
|
linkimg_uri = self._search_regex(
|
||||||
r'return\s+"(https?://[^"]+)"', decoded, 'video URL')
|
r'<img[^>]+id="linkimg"[^>]+src="([^"]+)"', webpage, 'link image')
|
||||||
|
linkimg = self._request_webpage(
|
||||||
|
linkimg_uri, video_id, note=False).read()
|
||||||
|
|
||||||
|
width, height, pixels = decode_png(linkimg)
|
||||||
|
|
||||||
|
output = ''
|
||||||
|
for y in range(height):
|
||||||
|
for x in range(width):
|
||||||
|
r, g, b = pixels[y][3 * x:3 * x + 3]
|
||||||
|
if r == 0 and g == 0 and b == 0:
|
||||||
|
break
|
||||||
|
else:
|
||||||
|
output += compat_chr(r)
|
||||||
|
output += compat_chr(g)
|
||||||
|
output += compat_chr(b)
|
||||||
|
|
||||||
|
img_str_length = len(output) // 200
|
||||||
|
img_str = [[0 for x in range(img_str_length)] for y in range(10)]
|
||||||
|
|
||||||
|
sig_str_length = len(signums) // 260
|
||||||
|
sig_str = [[0 for x in range(sig_str_length)] for y in range(10)]
|
||||||
|
|
||||||
|
for i in range(10):
|
||||||
|
for j in range(img_str_length):
|
||||||
|
begin = i * img_str_length * 20 + j * 20
|
||||||
|
img_str[i][j] = output[begin:begin + 20]
|
||||||
|
for j in range(sig_str_length):
|
||||||
|
begin = i * sig_str_length * 26 + j * 26
|
||||||
|
sig_str[i][j] = signums[begin:begin + 26]
|
||||||
|
|
||||||
|
parts = []
|
||||||
|
# TODO: find better names for str_, chr_ and sum_
|
||||||
|
str_ = ''
|
||||||
|
for i in [2, 3, 5, 7]:
|
||||||
|
str_ = ''
|
||||||
|
sum_ = float(99)
|
||||||
|
for j in range(len(sig_str[i])):
|
||||||
|
for chr_idx in range(len(img_str[i][j])):
|
||||||
|
if sum_ > float(122):
|
||||||
|
sum_ = float(98)
|
||||||
|
chr_ = compat_chr(int(math.floor(sum_)))
|
||||||
|
if sig_str[i][j][chr_idx] == chr_ and j >= len(str_):
|
||||||
|
sum_ += float(2.5)
|
||||||
|
str_ += img_str[i][j][chr_idx]
|
||||||
|
parts.append(str_.replace(',', ''))
|
||||||
|
|
||||||
|
video_url = 'https://openload.co/stream/%s~%s~%s~%s' % (parts[3], parts[1], parts[2], parts[0])
|
||||||
|
|
||||||
title = self._og_search_title(webpage, default=None) or self._search_regex(
|
title = self._og_search_title(webpage, default=None) or self._search_regex(
|
||||||
r'<span[^>]+class=["\']title["\'][^>]*>([^<]+)', webpage,
|
r'<span[^>]+class=["\']title["\'][^>]*>([^<]+)', webpage,
|
||||||
'title', default=None) or self._html_search_meta(
|
'title', default=None) or self._html_search_meta(
|
||||||
'description', webpage, 'title', fatal=True)
|
'description', webpage, 'title', fatal=True)
|
||||||
|
|
||||||
ext = mimetype2ext(self._search_regex(
|
|
||||||
r'window\.vt\s*=\s*(["\'])(?P<mimetype>.+?)\1', decoded,
|
|
||||||
'mimetype', default=None, group='mimetype')) or determine_ext(
|
|
||||||
video_url, 'mp4')
|
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'title': title,
|
'title': title,
|
||||||
'ext': ext,
|
|
||||||
'thumbnail': self._og_search_thumbnail(webpage, default=None),
|
'thumbnail': self._og_search_thumbnail(webpage, default=None),
|
||||||
'url': video_url,
|
'url': video_url,
|
||||||
|
# Seems all videos have extensions in their titles
|
||||||
|
'ext': determine_ext(title),
|
||||||
}
|
}
|
||||||
|
Loading…
Reference in New Issue
Block a user