1
0
Fork 0
mirror of https://gitlab.com/dstftw/youtube-dl.git synced 2020-11-16 09:42:26 +00:00
youtube-dl/youtube_dl/extractor/youtube.py

1822 lines
79 KiB
Python
Raw Normal View History

# coding: utf-8
import collections
import errno
import io
2013-09-22 08:37:23 +00:00
import itertools
import json
import os.path
import re
import struct
import traceback
import zlib
from .common import InfoExtractor, SearchInfoExtractor
from .subtitles import SubtitlesInfoExtractor
from ..jsinterp import JSInterpreter
from ..utils import (
compat_chr,
compat_parse_qs,
compat_urllib_parse,
compat_urllib_request,
compat_urlparse,
compat_str,
clean_html,
get_cachedir,
get_element_by_id,
get_element_by_attribute,
ExtractorError,
int_or_none,
PagedList,
unescapeHTML,
unified_strdate,
orderedSet,
write_json_file,
uppercase_escape,
)
class YoutubeBaseInfoExtractor(InfoExtractor):
"""Provide base functions for Youtube extractors"""
_LOGIN_URL = 'https://accounts.google.com/ServiceLogin'
_LANG_URL = r'https://www.youtube.com/?hl=en&persist_hl=1&gl=US&persist_gl=1&opt_out_ackd=1'
2014-01-25 16:11:55 +00:00
_AGE_URL = 'https://www.youtube.com/verify_age?next_url=/&gl=US&hl=en'
_NETRC_MACHINE = 'youtube'
# If True it will raise an error if no login info is provided
_LOGIN_REQUIRED = False
def _set_language(self):
return bool(self._download_webpage(
self._LANG_URL, None,
note=u'Setting language', errnote='unable to set language',
fatal=False))
def _login(self):
(username, password) = self._get_login_info()
# No authentication to be performed
if username is None:
if self._LOGIN_REQUIRED:
raise ExtractorError(u'No login info available, needed for using %s.' % self.IE_NAME, expected=True)
return False
login_page = self._download_webpage(
self._LOGIN_URL, None,
note=u'Downloading login page',
errnote=u'unable to fetch login page', fatal=False)
if login_page is False:
return
2013-10-29 05:45:54 +00:00
galx = self._search_regex(r'(?s)<input.+?name="GALX".+?value="(.+?)"',
login_page, u'Login GALX parameter')
# Log in
login_form_strs = {
u'continue': u'https://www.youtube.com/signin?action_handle_signin=true&feature=sign_in_button&hl=en_US&nomobiletemp=1',
u'Email': username,
u'GALX': galx,
u'Passwd': password,
u'PersistentCookie': u'yes',
u'_utf8': u'',
u'bgresponse': u'js_disabled',
u'checkConnection': u'',
u'checkedDomains': u'youtube',
u'dnConn': u'',
u'pstMsg': u'0',
u'rmShown': u'1',
u'secTok': u'',
u'signIn': u'Sign in',
u'timeStmp': u'',
u'service': u'youtube',
u'uilel': u'3',
u'hl': u'en_US',
}
# Convert to UTF-8 *before* urlencode because Python 2.x's urlencode
# chokes on unicode
login_form = dict((k.encode('utf-8'), v.encode('utf-8')) for k,v in login_form_strs.items())
login_data = compat_urllib_parse.urlencode(login_form).encode('ascii')
req = compat_urllib_request.Request(self._LOGIN_URL, login_data)
login_results = self._download_webpage(
req, None,
note=u'Logging in', errnote=u'unable to log in', fatal=False)
if login_results is False:
return False
if re.search(r'(?i)<form[^>]* id="gaia_loginform"', login_results) is not None:
self._downloader.report_warning(u'unable to log in: bad username or password')
return False
return True
def _confirm_age(self):
age_form = {
'next_url': '/',
'action_confirm': 'Confirm',
}
req = compat_urllib_request.Request(self._AGE_URL,
compat_urllib_parse.urlencode(age_form).encode('ascii'))
self._download_webpage(
req, None,
note=u'Confirming age', errnote=u'Unable to confirm age')
return True
def _real_initialize(self):
if self._downloader is None:
return
if not self._set_language():
return
if not self._login():
return
self._confirm_age()
class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor):
IE_DESC = u'YouTube.com'
_VALID_URL = r"""(?x)^
(
(?:https?://|//)? # http(s):// or protocol-independent URL (optional)
(?:(?:(?:(?:\w+\.)?[yY][oO][uU][tT][uU][bB][eE](?:-nocookie)?\.com/|
(?:www\.)?deturl\.com/www\.youtube\.com/|
2014-02-10 00:30:47 +00:00
(?:www\.)?pwnyoutube\.com/|
(?:www\.)?yourepeat\.com/|
tube\.majestyc\.net/|
youtube\.googleapis\.com/) # the various hostnames, with wildcard subdomains
(?:.*?\#/)? # handle anchor (#/) redirect urls
(?: # the various things that can precede the ID:
(?:(?:v|embed|e)/) # v/ or embed/ or e/
|(?: # or the v= param in all its forms
(?:(?:watch|movie)(?:_popup)?(?:\.php)?/?)? # preceding watch(_popup|.php) or nothing (like /?v=xxxx)
(?:\?|\#!?) # the params delimiter ? or # or #!
(?:.*?&)? # any other preceding param (like /?s=tuff&v=xxxx)
v=
)
))
|youtu\.be/ # just youtu.be/xxxx
|https?://(?:www\.)?cleanvideosearch\.com/media/action/yt/watch\?videoId=
)
)? # all until now is optional -> you can pass the naked ID
([0-9A-Za-z_-]{11}) # here is it! the YouTube video ID
(?(1).+)? # if we found the ID, everything can follow
$"""
_NEXT_URL_RE = r'[\?&]next_url=([^&]+)'
_formats = {
'5': {'ext': 'flv', 'width': 400, 'height': 240},
'6': {'ext': 'flv', 'width': 450, 'height': 270},
'13': {'ext': '3gp'},
'17': {'ext': '3gp', 'width': 176, 'height': 144},
'18': {'ext': 'mp4', 'width': 640, 'height': 360},
'22': {'ext': 'mp4', 'width': 1280, 'height': 720},
'34': {'ext': 'flv', 'width': 640, 'height': 360},
'35': {'ext': 'flv', 'width': 854, 'height': 480},
'36': {'ext': '3gp', 'width': 320, 'height': 240},
'37': {'ext': 'mp4', 'width': 1920, 'height': 1080},
'38': {'ext': 'mp4', 'width': 4096, 'height': 3072},
'43': {'ext': 'webm', 'width': 640, 'height': 360},
'44': {'ext': 'webm', 'width': 854, 'height': 480},
'45': {'ext': 'webm', 'width': 1280, 'height': 720},
'46': {'ext': 'webm', 'width': 1920, 'height': 1080},
# 3d videos
'82': {'ext': 'mp4', 'height': 360, 'format_note': '3D', 'preference': -20},
'83': {'ext': 'mp4', 'height': 480, 'format_note': '3D', 'preference': -20},
'84': {'ext': 'mp4', 'height': 720, 'format_note': '3D', 'preference': -20},
'85': {'ext': 'mp4', 'height': 1080, 'format_note': '3D', 'preference': -20},
'100': {'ext': 'webm', 'height': 360, 'format_note': '3D', 'preference': -20},
'101': {'ext': 'webm', 'height': 480, 'format_note': '3D', 'preference': -20},
'102': {'ext': 'webm', 'height': 720, 'format_note': '3D', 'preference': -20},
2013-08-20 01:22:25 +00:00
2013-09-04 01:49:35 +00:00
# Apple HTTP Live Streaming
'92': {'ext': 'mp4', 'height': 240, 'format_note': 'HLS', 'preference': -10},
'93': {'ext': 'mp4', 'height': 360, 'format_note': 'HLS', 'preference': -10},
'94': {'ext': 'mp4', 'height': 480, 'format_note': 'HLS', 'preference': -10},
'95': {'ext': 'mp4', 'height': 720, 'format_note': 'HLS', 'preference': -10},
'96': {'ext': 'mp4', 'height': 1080, 'format_note': 'HLS', 'preference': -10},
'132': {'ext': 'mp4', 'height': 240, 'format_note': 'HLS', 'preference': -10},
'151': {'ext': 'mp4', 'height': 72, 'format_note': 'HLS', 'preference': -10},
# DASH mp4 video
'133': {'ext': 'mp4', 'height': 240, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'134': {'ext': 'mp4', 'height': 360, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'135': {'ext': 'mp4', 'height': 480, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'136': {'ext': 'mp4', 'height': 720, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'137': {'ext': 'mp4', 'height': 1080, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'138': {'ext': 'mp4', 'height': 2160, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'160': {'ext': 'mp4', 'height': 144, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'264': {'ext': 'mp4', 'height': 1440, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
2013-08-20 01:22:25 +00:00
# Dash mp4 audio
'139': {'ext': 'm4a', 'format_note': 'DASH audio', 'vcodec': 'none', 'abr': 48, 'preference': -50},
'140': {'ext': 'm4a', 'format_note': 'DASH audio', 'vcodec': 'none', 'abr': 128, 'preference': -50},
'141': {'ext': 'm4a', 'format_note': 'DASH audio', 'vcodec': 'none', 'abr': 256, 'preference': -50},
2013-08-20 01:22:25 +00:00
# Dash webm
'167': {'ext': 'webm', 'height': 360, 'width': 640, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'168': {'ext': 'webm', 'height': 480, 'width': 854, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'169': {'ext': 'webm', 'height': 720, 'width': 1280, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'170': {'ext': 'webm', 'height': 1080, 'width': 1920, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'218': {'ext': 'webm', 'height': 480, 'width': 854, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'219': {'ext': 'webm', 'height': 480, 'width': 854, 'format_note': 'DASH video', 'acodec': 'none', 'container': 'webm', 'vcodec': 'VP8', 'preference': -40},
'242': {'ext': 'webm', 'height': 240, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'243': {'ext': 'webm', 'height': 360, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'244': {'ext': 'webm', 'height': 480, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'245': {'ext': 'webm', 'height': 480, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'246': {'ext': 'webm', 'height': 480, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'247': {'ext': 'webm', 'height': 720, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
'248': {'ext': 'webm', 'height': 1080, 'format_note': 'DASH video', 'acodec': 'none', 'preference': -40},
# Dash webm audio
'171': {'ext': 'webm', 'vcodec': 'none', 'format_note': 'DASH audio', 'abr': 48, 'preference': -50},
'172': {'ext': 'webm', 'vcodec': 'none', 'format_note': 'DASH audio', 'abr': 256, 'preference': -50},
# RTMP (unnamed)
'_rtmp': {'protocol': 'rtmp'},
}
2013-08-20 01:22:25 +00:00
IE_NAME = u'youtube'
_TESTS = [
{
2013-06-27 17:55:39 +00:00
u"url": u"http://www.youtube.com/watch?v=BaW_jenozKc",
u"file": u"BaW_jenozKc.mp4",
u"info_dict": {
u"title": u"youtube-dl test video \"'/\\ä↭𝕐",
u"uploader": u"Philipp Hagemeister",
u"uploader_id": u"phihag",
u"upload_date": u"20121002",
2013-12-04 13:16:52 +00:00
u"description": u"test chars: \"'/\\ä↭𝕐\ntest URL: https://github.com/rg3/youtube-dl/issues/1892\n\nThis is a test video for youtube-dl.\n\nFor more information, contact phihag@phihag.de ."
}
2013-06-27 17:55:39 +00:00
},
{
u"url": u"http://www.youtube.com/watch?v=UxxajLWwzqY",
u"file": u"UxxajLWwzqY.mp4",
u"note": u"Test generic use_cipher_signature video (#897)",
u"info_dict": {
u"upload_date": u"20120506",
u"title": u"Icona Pop - I Love It (feat. Charli XCX) [OFFICIAL VIDEO]",
u"description": u"md5:fea86fda2d5a5784273df5c7cc994d9f",
u"uploader": u"Icona Pop",
2013-06-27 17:55:39 +00:00
u"uploader_id": u"IconaPop"
}
},
{
u"url": u"https://www.youtube.com/watch?v=07FYdnEawAQ",
u"file": u"07FYdnEawAQ.mp4",
u"note": u"Test VEVO video with age protection (#956)",
u"info_dict": {
u"upload_date": u"20130703",
u"title": u"Justin Timberlake - Tunnel Vision (Explicit)",
u"description": u"md5:64249768eec3bc4276236606ea996373",
u"uploader": u"justintimberlakeVEVO",
u"uploader_id": u"justintimberlakeVEVO"
}
},
2013-11-18 12:05:18 +00:00
{
u"url": u"//www.YouTube.com/watch?v=yZIXLfi8CZQ",
2013-11-18 12:05:18 +00:00
u"file": u"yZIXLfi8CZQ.mp4",
u"note": u"Embed-only video (#1746)",
u"info_dict": {
u"upload_date": u"20120608",
u"title": u"Principal Sexually Assaults A Teacher - Episode 117 - 8th June 2012",
u"description": u"md5:09b78bd971f1e3e289601dfba15ca4f7",
u"uploader": u"SET India",
u"uploader_id": u"setindia"
}
},
{
u"url": u"http://www.youtube.com/watch?v=a9LDPn-MO4I",
u"file": u"a9LDPn-MO4I.m4a",
u"note": u"256k DASH audio (format 141) via DASH manifest",
u"info_dict": {
u"upload_date": "20121002",
u"uploader_id": "8KVIDEO",
u"description": "No description available.",
u"uploader": "8KVIDEO",
u"title": "UHDTV TEST 8K VIDEO.mp4"
},
u"params": {
u"youtube_include_dash_manifest": True,
u"format": "141",
},
},
# DASH manifest with encrypted signature
{
u'url': u'https://www.youtube.com/watch?v=IB3lcPjvWLA',
u'info_dict': {
u'id': u'IB3lcPjvWLA',
u'ext': u'm4a',
u'title': u'Afrojack - The Spark ft. Spree Wilson',
u'description': u'md5:9717375db5a9a3992be4668bbf3bc0a8',
u'uploader': u'AfrojackVEVO',
u'uploader_id': u'AfrojackVEVO',
u'upload_date': u'20131011',
},
u"params": {
u'youtube_include_dash_manifest': True,
u'format': '141',
},
},
]
@classmethod
def suitable(cls, url):
"""Receives a URL and returns True if suitable for this IE."""
if YoutubePlaylistIE.suitable(url): return False
2013-11-18 12:05:18 +00:00
return re.match(cls._VALID_URL, url) is not None
def __init__(self, *args, **kwargs):
super(YoutubeIE, self).__init__(*args, **kwargs)
2013-09-21 13:19:48 +00:00
self._player_cache = {}
def report_video_info_webpage_download(self, video_id):
"""Report attempt to download video info webpage."""
self.to_screen(u'%s: Downloading video info webpage' % video_id)
def report_information_extraction(self, video_id):
"""Report attempt to extract video information."""
self.to_screen(u'%s: Extracting video information' % video_id)
def report_unavailable_format(self, video_id, format):
"""Report extracted video URL."""
self.to_screen(u'%s: Format %s not available' % (video_id, format))
def report_rtmp_download(self):
"""Indicate the download will use the RTMP protocol."""
self.to_screen(u'RTMP download detected')
def _extract_signature_function(self, video_id, player_url, slen):
id_m = re.match(r'.*-(?P<id>[a-zA-Z0-9_-]+)\.(?P<ext>[a-z]+)$',
2013-09-21 13:19:48 +00:00
player_url)
player_type = id_m.group('ext')
player_id = id_m.group('id')
# Read from filesystem cache
func_id = '%s_%s_%d' % (player_type, player_id, slen)
assert os.path.basename(func_id) == func_id
cache_dir = get_cachedir(self._downloader.params)
cache_enabled = cache_dir is not None
if cache_enabled:
cache_fn = os.path.join(os.path.expanduser(cache_dir),
u'youtube-sigfuncs',
func_id + '.json')
try:
with io.open(cache_fn, 'r', encoding='utf-8') as cachef:
cache_spec = json.load(cachef)
return lambda s: u''.join(s[i] for i in cache_spec)
except IOError:
pass # No cache available
2013-09-21 13:19:48 +00:00
if player_type == 'js':
code = self._download_webpage(
player_url, video_id,
2013-09-21 13:19:48 +00:00
note=u'Downloading %s player %s' % (player_type, player_id),
errnote=u'Download of %s failed' % player_url)
2013-09-21 13:19:48 +00:00
res = self._parse_sig_js(code)
elif player_type == 'swf':
urlh = self._request_webpage(
player_url, video_id,
2013-09-21 13:19:48 +00:00
note=u'Downloading %s player %s' % (player_type, player_id),
errnote=u'Download of %s failed' % player_url)
code = urlh.read()
2013-09-21 13:19:48 +00:00
res = self._parse_sig_swf(code)
else:
assert False, 'Invalid player type %r' % player_type
if cache_enabled:
try:
2013-09-22 10:18:16 +00:00
test_string = u''.join(map(compat_chr, range(slen)))
cache_res = res(test_string)
cache_spec = [ord(c) for c in cache_res]
try:
os.makedirs(os.path.dirname(cache_fn))
except OSError as ose:
if ose.errno != errno.EEXIST:
raise
write_json_file(cache_spec, cache_fn)
2013-09-22 08:37:23 +00:00
except Exception:
tb = traceback.format_exc()
self._downloader.report_warning(
u'Writing cache to %r failed: %s' % (cache_fn, tb))
2013-09-21 13:19:48 +00:00
return res
def _print_sig_code(self, func, slen):
def gen_sig_code(idxs):
def _genslice(start, end, step):
starts = u'' if start == 0 else str(start)
ends = (u':%d' % (end+step)) if end + step >= 0 else u':'
steps = u'' if step == 1 else (u':%d' % step)
return u's[%s%s%s]' % (starts, ends, steps)
step = None
2013-09-22 08:37:23 +00:00
start = '(Never used)' # Quelch pyflakes warnings - start will be
# set as soon as step is set
for i, prev in zip(idxs[1:], idxs[:-1]):
if step is not None:
if i - prev == step:
continue
yield _genslice(start, prev, step)
step = None
continue
if i - prev in [-1, 1]:
step = i - prev
start = prev
continue
else:
yield u's[%d]' % prev
if step is None:
yield u's[%d]' % i
else:
yield _genslice(start, i, step)
2013-09-22 10:18:16 +00:00
test_string = u''.join(map(compat_chr, range(slen)))
cache_res = func(test_string)
cache_spec = [ord(c) for c in cache_res]
expr_code = u' + '.join(gen_sig_code(cache_spec))
code = u'if len(s) == %d:\n return %s\n' % (slen, expr_code)
self.to_screen(u'Extracted signature function:\n' + code)
def _parse_sig_js(self, jscode):
funcname = self._search_regex(
r'signature=([a-zA-Z]+)', jscode,
u'Initial JS player signature function name')
jsi = JSInterpreter(jscode)
initial_function = jsi.extract_function(funcname)
return lambda s: initial_function([s])
def _parse_sig_swf(self, file_contents):
if file_contents[1:3] != b'WS':
raise ExtractorError(
u'Not an SWF file; header is %r' % file_contents[:3])
if file_contents[:1] == b'C':
content = zlib.decompress(file_contents[8:])
else:
raise NotImplementedError(u'Unsupported compression format %r' %
file_contents[:1])
def extract_tags(content):
pos = 0
while pos < len(content):
header16 = struct.unpack('<H', content[pos:pos+2])[0]
pos += 2
tag_code = header16 >> 6
tag_len = header16 & 0x3f
if tag_len == 0x3f:
tag_len = struct.unpack('<I', content[pos:pos+4])[0]
pos += 4
assert pos+tag_len <= len(content)
yield (tag_code, content[pos:pos+tag_len])
pos += tag_len
code_tag = next(tag
for tag_code, tag in extract_tags(content)
if tag_code == 82)
p = code_tag.index(b'\0', 4) + 1
2013-09-21 13:32:37 +00:00
code_reader = io.BytesIO(code_tag[p:])
# Parse ABC (AVM2 ByteCode)
2013-09-21 13:32:37 +00:00
def read_int(reader=None):
if reader is None:
reader = code_reader
res = 0
shift = 0
for _ in range(5):
2013-09-21 13:32:37 +00:00
buf = reader.read(1)
assert len(buf) == 1
b = struct.unpack('<B', buf)[0]
res = res | ((b & 0x7f) << shift)
if b & 0x80 == 0:
break
shift += 7
2013-09-21 13:32:37 +00:00
return res
def u30(reader=None):
res = read_int(reader)
assert res & 0xf0000000 == 0
return res
u32 = read_int
2013-09-21 13:32:37 +00:00
def s32(reader=None):
v = read_int(reader)
if v & 0x80000000 != 0:
v = - ((v ^ 0xffffffff) + 1)
2013-09-21 13:32:37 +00:00
return v
2013-09-22 08:37:23 +00:00
def read_string(reader=None):
2013-09-21 13:32:37 +00:00
if reader is None:
reader = code_reader
slen = u30(reader)
resb = reader.read(slen)
assert len(resb) == slen
return resb.decode('utf-8')
def read_bytes(count, reader=None):
if reader is None:
reader = code_reader
resb = reader.read(count)
assert len(resb) == count
return resb
def read_byte(reader=None):
resb = read_bytes(1, reader=reader)
res = struct.unpack('<B', resb)[0]
return res
# minor_version + major_version
2013-09-22 08:37:23 +00:00
read_bytes(2 + 2)
# Constant pool
2013-09-21 13:32:37 +00:00
int_count = u30()
for _c in range(1, int_count):
2013-09-22 08:37:23 +00:00
s32()
2013-09-21 13:32:37 +00:00
uint_count = u30()
for _c in range(1, uint_count):
2013-09-22 08:37:23 +00:00
u32()
2013-09-21 13:32:37 +00:00
double_count = u30()
2013-09-22 08:37:23 +00:00
read_bytes((double_count-1) * 8)
2013-09-21 13:32:37 +00:00
string_count = u30()
constant_strings = [u'']
for _c in range(1, string_count):
2013-09-22 08:37:23 +00:00
s = read_string()
constant_strings.append(s)
2013-09-21 13:32:37 +00:00
namespace_count = u30()
for _c in range(1, namespace_count):
2013-09-22 08:37:23 +00:00
read_bytes(1) # kind
u30() # name
2013-09-21 13:32:37 +00:00
ns_set_count = u30()
for _c in range(1, ns_set_count):
2013-09-21 13:32:37 +00:00
count = u30()
for _c2 in range(count):
2013-09-22 08:37:23 +00:00
u30()
2013-09-21 13:32:37 +00:00
multiname_count = u30()
MULTINAME_SIZES = {
0x07: 2, # QName
0x0d: 2, # QNameA
0x0f: 1, # RTQName
0x10: 1, # RTQNameA
0x11: 0, # RTQNameL
0x12: 0, # RTQNameLA
0x09: 2, # Multiname
0x0e: 2, # MultinameA
0x1b: 1, # MultinameL
0x1c: 1, # MultinameLA
}
multinames = [u'']
for _c in range(1, multiname_count):
2013-09-21 13:32:37 +00:00
kind = u30()
assert kind in MULTINAME_SIZES, u'Invalid multiname kind %r' % kind
if kind == 0x07:
2013-09-22 08:37:23 +00:00
u30() # namespace_idx
2013-09-21 13:32:37 +00:00
name_idx = u30()
multinames.append(constant_strings[name_idx])
else:
multinames.append('[MULTINAME kind: %d]' % kind)
for _c2 in range(MULTINAME_SIZES[kind]):
2013-09-22 08:37:23 +00:00
u30()
# Methods
2013-09-21 13:32:37 +00:00
method_count = u30()
MethodInfo = collections.namedtuple(
'MethodInfo',
['NEED_ARGUMENTS', 'NEED_REST'])
method_infos = []
for method_id in range(method_count):
2013-09-21 13:32:37 +00:00
param_count = u30()
2013-09-22 08:37:23 +00:00
u30() # return type
for _ in range(param_count):
2013-09-22 08:37:23 +00:00
u30() # param type
u30() # name index (always 0 for youtube)
2013-09-21 13:32:37 +00:00
flags = read_byte()
if flags & 0x08 != 0:
# Options present
2013-09-21 13:32:37 +00:00
option_count = u30()
for c in range(option_count):
2013-09-22 08:37:23 +00:00
u30() # val
read_bytes(1) # kind
if flags & 0x80 != 0:
# Param names present
for _ in range(param_count):
2013-09-22 08:37:23 +00:00
u30() # param name
mi = MethodInfo(flags & 0x01 != 0, flags & 0x04 != 0)
method_infos.append(mi)
# Metadata
2013-09-21 13:32:37 +00:00
metadata_count = u30()
for _c in range(metadata_count):
2013-09-22 08:37:23 +00:00
u30() # name
2013-09-21 13:32:37 +00:00
item_count = u30()
for _c2 in range(item_count):
2013-09-22 08:37:23 +00:00
u30() # key
u30() # value
2013-09-21 13:32:37 +00:00
def parse_traits_info():
trait_name_idx = u30()
kind_full = read_byte()
kind = kind_full & 0x0f
attrs = kind_full >> 4
methods = {}
if kind in [0x00, 0x06]: # Slot or Const
2013-09-22 08:37:23 +00:00
u30() # Slot id
u30() # type_name_idx
2013-09-21 13:32:37 +00:00
vindex = u30()
if vindex != 0:
2013-09-22 08:37:23 +00:00
read_byte() # vkind
elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter
2013-09-22 08:37:23 +00:00
u30() # disp_id
2013-09-21 13:32:37 +00:00
method_idx = u30()
methods[multinames[trait_name_idx]] = method_idx
elif kind == 0x04: # Class
2013-09-22 08:37:23 +00:00
u30() # slot_id
u30() # classi
elif kind == 0x05: # Function
2013-09-22 08:37:23 +00:00
u30() # slot_id
2013-09-21 13:32:37 +00:00
function_idx = u30()
methods[function_idx] = multinames[trait_name_idx]
else:
raise ExtractorError(u'Unsupported trait kind %d' % kind)
if attrs & 0x4 != 0: # Metadata present
2013-09-21 13:32:37 +00:00
metadata_count = u30()
for _c3 in range(metadata_count):
2013-09-22 08:37:23 +00:00
u30() # metadata index
2013-09-21 13:32:37 +00:00
return methods
# Classes
TARGET_CLASSNAME = u'SignatureDecipher'
searched_idx = multinames.index(TARGET_CLASSNAME)
searched_class_id = None
2013-09-21 13:32:37 +00:00
class_count = u30()
for class_id in range(class_count):
2013-09-21 13:32:37 +00:00
name_idx = u30()
if name_idx == searched_idx:
# We found the class we're looking for!
searched_class_id = class_id
2013-09-22 08:37:23 +00:00
u30() # super_name idx
2013-09-21 13:32:37 +00:00
flags = read_byte()
if flags & 0x08 != 0: # Protected namespace is present
2013-09-22 08:37:23 +00:00
u30() # protected_ns_idx
2013-09-21 13:32:37 +00:00
intrf_count = u30()
for _c2 in range(intrf_count):
2013-09-22 08:37:23 +00:00
u30()
u30() # iinit
2013-09-21 13:32:37 +00:00
trait_count = u30()
for _c2 in range(trait_count):
2013-09-22 08:37:23 +00:00
parse_traits_info()
if searched_class_id is None:
raise ExtractorError(u'Target class %r not found' %
TARGET_CLASSNAME)
method_names = {}
method_idxs = {}
for class_id in range(class_count):
2013-09-22 08:37:23 +00:00
u30() # cinit
2013-09-21 13:32:37 +00:00
trait_count = u30()
for _c2 in range(trait_count):
2013-09-21 13:32:37 +00:00
trait_methods = parse_traits_info()
if class_id == searched_class_id:
method_names.update(trait_methods.items())
method_idxs.update(dict(
(idx, name)
for name, idx in trait_methods.items()))
# Scripts
2013-09-21 13:32:37 +00:00
script_count = u30()
for _c in range(script_count):
2013-09-22 08:37:23 +00:00
u30() # init
2013-09-21 13:32:37 +00:00
trait_count = u30()
for _c2 in range(trait_count):
2013-09-22 08:37:23 +00:00
parse_traits_info()
# Method bodies
2013-09-21 13:32:37 +00:00
method_body_count = u30()
Method = collections.namedtuple('Method', ['code', 'local_count'])
methods = {}
for _c in range(method_body_count):
2013-09-21 13:32:37 +00:00
method_idx = u30()
2013-09-22 08:37:23 +00:00
u30() # max_stack
2013-09-21 13:32:37 +00:00
local_count = u30()
2013-09-22 08:37:23 +00:00
u30() # init_scope_depth
u30() # max_scope_depth
2013-09-21 13:32:37 +00:00
code_length = u30()
code = read_bytes(code_length)
if method_idx in method_idxs:
2013-09-21 13:32:37 +00:00
m = Method(code, local_count)
methods[method_idxs[method_idx]] = m
2013-09-21 13:32:37 +00:00
exception_count = u30()
for _c2 in range(exception_count):
2013-09-22 08:37:23 +00:00
u30() # from
u30() # to
u30() # target
u30() # exc_type
u30() # var_name
2013-09-21 13:32:37 +00:00
trait_count = u30()
for _c2 in range(trait_count):
2013-09-22 08:37:23 +00:00
parse_traits_info()
2013-09-21 13:32:37 +00:00
assert p + code_reader.tell() == len(code_tag)
assert len(methods) == len(method_idxs)
method_pyfunctions = {}
def extract_function(func_name):
if func_name in method_pyfunctions:
return method_pyfunctions[func_name]
if func_name not in methods:
raise ExtractorError(u'Cannot find function %r' % func_name)
m = methods[func_name]
def resfunc(args):
registers = ['(this)'] + list(args) + [None] * m.local_count
stack = []
coder = io.BytesIO(m.code)
while True:
opcode = struct.unpack('!B', coder.read(1))[0]
2013-09-21 12:48:12 +00:00
if opcode == 36: # pushbyte
v = struct.unpack('!B', coder.read(1))[0]
stack.append(v)
elif opcode == 44: # pushstring
idx = u30(coder)
stack.append(constant_strings[idx])
elif opcode == 48: # pushscope
# We don't implement the scope register, so we'll just
# ignore the popped value
stack.pop()
elif opcode == 70: # callproperty
index = u30(coder)
mname = multinames[index]
arg_count = u30(coder)
args = list(reversed(
[stack.pop() for _ in range(arg_count)]))
obj = stack.pop()
if mname == u'split':
assert len(args) == 1
assert isinstance(args[0], compat_str)
assert isinstance(obj, compat_str)
if args[0] == u'':
res = list(obj)
else:
res = obj.split(args[0])
stack.append(res)
2013-09-21 12:48:12 +00:00
elif mname == u'slice':
assert len(args) == 1
assert isinstance(args[0], int)
assert isinstance(obj, list)
res = obj[args[0]:]
stack.append(res)
elif mname == u'join':
assert len(args) == 1
assert isinstance(args[0], compat_str)
assert isinstance(obj, list)
res = args[0].join(obj)
stack.append(res)
elif mname in method_pyfunctions:
stack.append(method_pyfunctions[mname](args))
else:
raise NotImplementedError(
u'Unsupported property %r on %r'
% (mname, obj))
2013-09-21 12:48:12 +00:00
elif opcode == 72: # returnvalue
res = stack.pop()
return res
elif opcode == 79: # callpropvoid
index = u30(coder)
mname = multinames[index]
arg_count = u30(coder)
args = list(reversed(
[stack.pop() for _ in range(arg_count)]))
obj = stack.pop()
if mname == u'reverse':
assert isinstance(obj, list)
obj.reverse()
else:
raise NotImplementedError(
u'Unsupported (void) property %r on %r'
% (mname, obj))
elif opcode == 93: # findpropstrict
index = u30(coder)
mname = multinames[index]
res = extract_function(mname)
stack.append(res)
elif opcode == 97: # setproperty
index = u30(coder)
value = stack.pop()
idx = stack.pop()
obj = stack.pop()
assert isinstance(obj, list)
assert isinstance(idx, int)
obj[idx] = value
elif opcode == 98: # getlocal
index = u30(coder)
stack.append(registers[index])
elif opcode == 99: # setlocal
index = u30(coder)
value = stack.pop()
registers[index] = value
elif opcode == 102: # getproperty
index = u30(coder)
pname = multinames[index]
if pname == u'length':
obj = stack.pop()
assert isinstance(obj, list)
stack.append(len(obj))
else: # Assume attribute access
idx = stack.pop()
assert isinstance(idx, int)
obj = stack.pop()
assert isinstance(obj, list)
stack.append(obj[idx])
elif opcode == 128: # coerce
2013-09-22 08:37:23 +00:00
u30(coder)
elif opcode == 133: # coerce_s
assert isinstance(stack[-1], (type(None), compat_str))
elif opcode == 164: # modulo
value2 = stack.pop()
value1 = stack.pop()
res = value1 % value2
stack.append(res)
2013-09-21 12:48:12 +00:00
elif opcode == 208: # getlocal_0
stack.append(registers[0])
elif opcode == 209: # getlocal_1
stack.append(registers[1])
elif opcode == 210: # getlocal_2
stack.append(registers[2])
elif opcode == 211: # getlocal_3
stack.append(registers[3])
elif opcode == 214: # setlocal_2
registers[2] = stack.pop()
elif opcode == 215: # setlocal_3
registers[3] = stack.pop()
else:
raise NotImplementedError(
u'Unsupported opcode %d' % opcode)
method_pyfunctions[func_name] = resfunc
return resfunc
initial_function = extract_function(u'decipher')
return lambda s: initial_function([s])
2013-09-21 13:19:48 +00:00
def _decrypt_signature(self, s, video_id, player_url, age_gate=False):
"""Turn the encrypted s field into a working signature"""
2013-09-21 13:19:48 +00:00
if player_url is not None:
if player_url.startswith(u'//'):
player_url = u'https:' + player_url
try:
player_id = (player_url, len(s))
if player_id not in self._player_cache:
2013-09-21 13:19:48 +00:00
func = self._extract_signature_function(
video_id, player_url, len(s)
)
self._player_cache[player_id] = func
func = self._player_cache[player_id]
if self._downloader.params.get('youtube_print_sig_code'):
self._print_sig_code(func, len(s))
return func(s)
2013-09-22 08:37:23 +00:00
except Exception:
tb = traceback.format_exc()
2013-09-21 13:19:48 +00:00
self._downloader.report_warning(
u'Automatic signature extraction failed: ' + tb)
self._downloader.report_warning(
u'Warning: Falling back to static signature algorithm')
2013-09-21 13:34:29 +00:00
return self._static_decrypt_signature(
s, video_id, player_url, age_gate)
2013-09-21 13:34:29 +00:00
def _static_decrypt_signature(self, s, video_id, player_url, age_gate):
if age_gate:
# The videos with age protection use another player, so the
# algorithms can be different.
if len(s) == 86:
return s[2:63] + s[82] + s[64:82] + s[63]
if len(s) == 93:
return s[86:29:-1] + s[88] + s[28:5:-1]
elif len(s) == 92:
return s[25] + s[3:25] + s[0] + s[26:42] + s[79] + s[43:79] + s[91] + s[80:83]
2013-09-20 12:43:16 +00:00
elif len(s) == 91:
return s[84:27:-1] + s[86] + s[26:5:-1]
elif len(s) == 90:
return s[25] + s[3:25] + s[2] + s[26:40] + s[77] + s[41:77] + s[89] + s[78:81]
elif len(s) == 89:
return s[84:78:-1] + s[87] + s[77:60:-1] + s[0] + s[59:3:-1]
elif len(s) == 88:
return s[7:28] + s[87] + s[29:45] + s[55] + s[46:55] + s[2] + s[56:87] + s[28]
elif len(s) == 87:
2013-08-15 20:00:20 +00:00
return s[6:27] + s[4] + s[28:39] + s[27] + s[40:59] + s[2] + s[60:]
elif len(s) == 86:
2013-09-24 19:20:42 +00:00
return s[80:72:-1] + s[16] + s[71:39:-1] + s[72] + s[38:16:-1] + s[82] + s[15::-1]
elif len(s) == 85:
return s[3:11] + s[0] + s[12:55] + s[84] + s[56:84]
elif len(s) == 84:
return s[78:70:-1] + s[14] + s[69:37:-1] + s[70] + s[36:14:-1] + s[80] + s[:14][::-1]
elif len(s) == 83:
return s[80:63:-1] + s[0] + s[62:0:-1] + s[63]
elif len(s) == 82:
2013-10-03 22:43:01 +00:00
return s[80:37:-1] + s[7] + s[36:7:-1] + s[0] + s[6:0:-1] + s[37]
elif len(s) == 81:
return s[56] + s[79:56:-1] + s[41] + s[55:41:-1] + s[80] + s[40:34:-1] + s[0] + s[33:29:-1] + s[34] + s[28:9:-1] + s[29] + s[8:0:-1] + s[9]
elif len(s) == 80:
return s[1:19] + s[0] + s[20:68] + s[19] + s[69:80]
elif len(s) == 79:
return s[54] + s[77:54:-1] + s[39] + s[53:39:-1] + s[78] + s[38:34:-1] + s[0] + s[33:29:-1] + s[34] + s[28:9:-1] + s[29] + s[8:0:-1] + s[9]
else:
raise ExtractorError(u'Unable to decrypt signature, key length %d not supported; retrying might work' % (len(s)))
def _get_available_subtitles(self, video_id, webpage):
try:
sub_list = self._download_webpage(
2014-01-25 16:11:55 +00:00
'https://video.google.com/timedtext?hl=en&type=list&v=%s' % video_id,
video_id, note=False)
except ExtractorError as err:
self._downloader.report_warning(u'unable to download video subtitles: %s' % compat_str(err))
return {}
lang_list = re.findall(r'name="([^"]*)"[^>]+lang_code="([\w\-]+)"', sub_list)
sub_lang_list = {}
for l in lang_list:
lang = l[1]
params = compat_urllib_parse.urlencode({
'lang': lang,
'v': video_id,
'fmt': self._downloader.params.get('subtitlesformat', 'srt'),
'name': unescapeHTML(l[0]).encode('utf-8'),
})
2014-01-25 16:11:55 +00:00
url = u'https://www.youtube.com/api/timedtext?' + params
sub_lang_list[lang] = url
if not sub_lang_list:
self._downloader.report_warning(u'video doesn\'t have subtitles')
return {}
return sub_lang_list
def _get_available_automatic_caption(self, video_id, webpage):
"""We need the webpage for getting the captions url, pass it as an
argument to speed up the process."""
sub_format = self._downloader.params.get('subtitlesformat', 'srt')
self.to_screen(u'%s: Looking for automatic captions' % video_id)
mobj = re.search(r';ytplayer.config = ({.*?});', webpage)
err_msg = u'Couldn\'t find automatic captions for %s' % video_id
if mobj is None:
self._downloader.report_warning(err_msg)
return {}
player_config = json.loads(mobj.group(1))
try:
args = player_config[u'args']
caption_url = args[u'ttsurl']
timestamp = args[u'timestamp']
# We get the available subtitles
list_params = compat_urllib_parse.urlencode({
'type': 'list',
'tlangs': 1,
'asrs': 1,
})
list_url = caption_url + '&' + list_params
caption_list = self._download_xml(list_url, video_id)
original_lang_node = caption_list.find('track')
if original_lang_node is None or original_lang_node.attrib.get('kind') != 'asr' :
self._downloader.report_warning(u'Video doesn\'t have automatic captions')
return {}
original_lang = original_lang_node.attrib['lang_code']
sub_lang_list = {}
for lang_node in caption_list.findall('target'):
sub_lang = lang_node.attrib['lang_code']
params = compat_urllib_parse.urlencode({
'lang': original_lang,
'tlang': sub_lang,
'fmt': sub_format,
'ts': timestamp,
'kind': 'asr',
})
sub_lang_list[sub_lang] = caption_url + '&' + params
return sub_lang_list
# An extractor error can be raise by the download process if there are
# no automatic captions but there are subtitles
except (KeyError, ExtractorError):
self._downloader.report_warning(err_msg)
return {}
@classmethod
def extract_id(cls, url):
mobj = re.match(cls._VALID_URL, url, re.VERBOSE)
if mobj is None:
raise ExtractorError(u'Invalid URL: %s' % url)
video_id = mobj.group(2)
return video_id
def _extract_from_m3u8(self, manifest_url, video_id):
url_map = {}
def _get_urls(_manifest):
lines = _manifest.split('\n')
urls = filter(lambda l: l and not l.startswith('#'),
lines)
return urls
manifest = self._download_webpage(manifest_url, video_id, u'Downloading formats manifest')
formats_urls = _get_urls(manifest)
for format_url in formats_urls:
itag = self._search_regex(r'itag/(\d+?)/', format_url, 'itag')
url_map[itag] = format_url
return url_map
def _extract_annotations(self, video_id):
url = 'https://www.youtube.com/annotations_invideo?features=1&legacy=1&video_id=%s' % video_id
return self._download_webpage(url, video_id, note=u'Searching for annotations.', errnote=u'Unable to download video annotations.')
def _real_extract(self, url):
proto = (
u'http' if self._downloader.params.get('prefer_insecure', False)
else u'https')
# Extract original video URL from URL with redirection, like age verification, using next_url parameter
mobj = re.search(self._NEXT_URL_RE, url)
if mobj:
url = proto + '://www.youtube.com/' + compat_urllib_parse.unquote(mobj.group(1)).lstrip('/')
video_id = self.extract_id(url)
# Get video webpage
url = proto + '://www.youtube.com/watch?v=%s&gl=US&hl=en&has_verified=1' % video_id
video_webpage = self._download_webpage(url, video_id)
# Attempt to extract SWF player URL
mobj = re.search(r'swfConfig.*?"(https?:\\/\\/.*?watch.*?-.*?\.swf)"', video_webpage)
if mobj is not None:
player_url = re.sub(r'\\(.)', r'\1', mobj.group(1))
else:
player_url = None
# Get video info
self.report_video_info_webpage_download(video_id)
if re.search(r'player-age-gate-content">', video_webpage) is not None:
self.report_age_confirmation()
age_gate = True
# We simulate the access to the video from www.youtube.com/v/{video_id}
# this can be viewed without login into Youtube
data = compat_urllib_parse.urlencode({'video_id': video_id,
2013-11-18 12:05:18 +00:00
'el': 'player_embedded',
'gl': 'US',
'hl': 'en',
'eurl': 'https://youtube.googleapis.com/v/' + video_id,
'asv': 3,
'sts':'1588',
})
video_info_url = proto + '://www.youtube.com/get_video_info?' + data
video_info_webpage = self._download_webpage(video_info_url, video_id,
note=False,
errnote='unable to download video info webpage')
video_info = compat_parse_qs(video_info_webpage)
else:
age_gate = False
for el_type in ['&el=embedded', '&el=detailpage', '&el=vevo', '']:
video_info_url = (proto + '://www.youtube.com/get_video_info?&video_id=%s%s&ps=default&eurl=&gl=US&hl=en'
% (video_id, el_type))
video_info_webpage = self._download_webpage(video_info_url, video_id,
note=False,
errnote='unable to download video info webpage')
video_info = compat_parse_qs(video_info_webpage)
if 'token' in video_info:
break
if 'token' not in video_info:
if 'reason' in video_info:
raise ExtractorError(
u'YouTube said: %s' % video_info['reason'][0],
expected=True, video_id=video_id)
else:
raise ExtractorError(
u'"token" parameter not in video info for unknown reason',
video_id=video_id)
2013-11-17 10:06:16 +00:00
if 'view_count' in video_info:
view_count = int(video_info['view_count'][0])
else:
view_count = None
# Check for "rental" videos
if 'ypc_video_rental_bar_text' in video_info and 'author' not in video_info:
raise ExtractorError(u'"rental" videos not supported')
# Start extracting information
self.report_information_extraction(video_id)
# uploader
if 'author' not in video_info:
raise ExtractorError(u'Unable to extract uploader name')
video_uploader = compat_urllib_parse.unquote_plus(video_info['author'][0])
# uploader_id
video_uploader_id = None
mobj = re.search(r'<link itemprop="url" href="http://www.youtube.com/(?:user|channel)/([^"]+)">', video_webpage)
if mobj is not None:
video_uploader_id = mobj.group(1)
else:
self._downloader.report_warning(u'unable to extract uploader nickname')
# title
if 'title' in video_info:
video_title = video_info['title'][0]
else:
self._downloader.report_warning(u'Unable to extract video title')
video_title = u'_'
# thumbnail image
# We try first to get a high quality image:
m_thumb = re.search(r'<span itemprop="thumbnail".*?href="(.*?)">',
video_webpage, re.DOTALL)
if m_thumb is not None:
video_thumbnail = m_thumb.group(1)
elif 'thumbnail_url' not in video_info:
self._downloader.report_warning(u'unable to extract video thumbnail')
video_thumbnail = None
else: # don't panic if we can't find it
video_thumbnail = compat_urllib_parse.unquote_plus(video_info['thumbnail_url'][0])
# upload date
upload_date = None
mobj = re.search(r'id="eow-date.*?>(.*?)</span>', video_webpage, re.DOTALL)
if mobj is not None:
upload_date = ' '.join(re.sub(r'[/,-]', r' ', mobj.group(1)).split())
upload_date = unified_strdate(upload_date)
# description
video_description = get_element_by_id("eow-description", video_webpage)
if video_description:
2013-12-04 13:16:52 +00:00
video_description = re.sub(r'''(?x)
<a\s+
(?:[a-zA-Z-]+="[^"]+"\s+)*?
title="([^"]+)"\s+
(?:[a-zA-Z-]+="[^"]+"\s+)*?
class="yt-uix-redirect-link"\s*>
[^<]+
</a>
''', r'\1', video_description)
video_description = clean_html(video_description)
else:
fd_mobj = re.search(r'<meta name="description" content="([^"]+)"', video_webpage)
if fd_mobj:
video_description = unescapeHTML(fd_mobj.group(1))
else:
video_description = u''
def _extract_count(klass):
count = self._search_regex(
r'class="%s">([\d,]+)</span>' % re.escape(klass),
video_webpage, klass, default=None)
if count is not None:
return int(count.replace(',', ''))
return None
like_count = _extract_count(u'likes-count')
dislike_count = _extract_count(u'dislikes-count')
# subtitles
video_subtitles = self.extract_subtitles(video_id, video_webpage)
if self._downloader.params.get('listsubtitles', False):
self._list_available_subtitles(video_id, video_webpage)
return
if 'length_seconds' not in video_info:
self._downloader.report_warning(u'unable to extract video duration')
video_duration = None
else:
video_duration = int(compat_urllib_parse.unquote_plus(video_info['length_seconds'][0]))
# annotations
video_annotations = None
if self._downloader.params.get('writeannotations', False):
video_annotations = self._extract_annotations(video_id)
# Decide which formats to download
try:
2014-03-10 12:35:40 +00:00
mobj = re.search(r';ytplayer\.config\s*=\s*({.*?});', video_webpage)
if not mobj:
raise ValueError('Could not find vevo ID')
2014-03-10 12:35:40 +00:00
json_code = uppercase_escape(mobj.group(1))
ytplayer_config = json.loads(json_code)
args = ytplayer_config['args']
# Easy way to know if the 's' value is in url_encoded_fmt_stream_map
# this signatures are encrypted
if 'url_encoded_fmt_stream_map' not in args:
raise ValueError(u'No stream_map present') # caught below
re_signature = re.compile(r'[&,]s=')
m_s = re_signature.search(args['url_encoded_fmt_stream_map'])
if m_s is not None:
self.to_screen(u'%s: Encrypted signatures detected.' % video_id)
video_info['url_encoded_fmt_stream_map'] = [args['url_encoded_fmt_stream_map']]
m_s = re_signature.search(args.get('adaptive_fmts', u''))
2013-08-21 04:57:32 +00:00
if m_s is not None:
if 'adaptive_fmts' in video_info:
video_info['adaptive_fmts'][0] += ',' + args['adaptive_fmts']
2013-08-21 06:51:05 +00:00
else:
video_info['adaptive_fmts'] = [args['adaptive_fmts']]
except ValueError:
pass
def _map_to_format_list(urlmap):
formats = []
for itag, video_real_url in urlmap.items():
dct = {
'format_id': itag,
'url': video_real_url,
'player_url': player_url,
}
if itag in self._formats:
dct.update(self._formats[itag])
formats.append(dct)
return formats
if 'conn' in video_info and video_info['conn'][0].startswith('rtmp'):
self.report_rtmp_download()
formats = [{
'format_id': '_rtmp',
'protocol': 'rtmp',
'url': video_info['conn'][0],
'player_url': player_url,
}]
elif len(video_info.get('url_encoded_fmt_stream_map', [])) >= 1 or len(video_info.get('adaptive_fmts', [])) >= 1:
encoded_url_map = video_info.get('url_encoded_fmt_stream_map', [''])[0] + ',' + video_info.get('adaptive_fmts',[''])[0]
if 'rtmpe%3Dyes' in encoded_url_map:
raise ExtractorError('rtmpe downloads are not supported, see https://github.com/rg3/youtube-dl/issues/343 for more information.', expected=True)
url_map = {}
for url_data_str in encoded_url_map.split(','):
url_data = compat_parse_qs(url_data_str)
if 'itag' in url_data and 'url' in url_data:
url = url_data['url'][0]
if 'sig' in url_data:
url += '&signature=' + url_data['sig'][0]
elif 's' in url_data:
encrypted_sig = url_data['s'][0]
if self._downloader.params.get('verbose'):
if age_gate:
if player_url is None:
player_version = 'unknown'
else:
player_version = self._search_regex(
r'-(.+)\.swf$', player_url,
u'flash player', fatal=False)
player_desc = 'flash player %s' % player_version
else:
2013-09-21 13:19:48 +00:00
player_version = self._search_regex(
r'html5player-(.+?)\.js', video_webpage,
'html5 player', fatal=False)
player_desc = u'html5 player %s' % player_version
parts_sizes = u'.'.join(compat_str(len(part)) for part in encrypted_sig.split('.'))
self.to_screen(u'encrypted signature length %d (%s), itag %s, %s' %
(len(encrypted_sig), parts_sizes, url_data['itag'][0], player_desc))
2013-09-21 13:19:48 +00:00
if not age_gate:
jsplayer_url_json = self._search_regex(
r'"assets":.+?"js":\s*("[^"]+")',
video_webpage, u'JS player URL')
2013-09-21 13:19:48 +00:00
player_url = json.loads(jsplayer_url_json)
2013-09-21 13:19:48 +00:00
signature = self._decrypt_signature(
encrypted_sig, video_id, player_url, age_gate)
url += '&signature=' + signature
if 'ratebypass' not in url:
url += '&ratebypass=yes'
url_map[url_data['itag'][0]] = url
formats = _map_to_format_list(url_map)
elif video_info.get('hlsvp'):
manifest_url = video_info['hlsvp'][0]
url_map = self._extract_from_m3u8(manifest_url, video_id)
formats = _map_to_format_list(url_map)
else:
raise ExtractorError(u'no conn, hlsvp or url_encoded_fmt_stream_map information found in video info')
# Look for the DASH manifest
if (self._downloader.params.get('youtube_include_dash_manifest', False)):
try:
# The DASH manifest used needs to be the one from the original video_webpage.
# The one found in get_video_info seems to be using different signatures.
# However, in the case of an age restriction there won't be any embedded dashmpd in the video_webpage.
# Luckily, it seems, this case uses some kind of default signature (len == 86), so the
# combination of get_video_info and the _static_decrypt_signature() decryption fallback will work here.
if age_gate:
dash_manifest_url = video_info.get('dashmpd')[0]
else:
dash_manifest_url = ytplayer_config['args']['dashmpd']
def decrypt_sig(mobj):
s = mobj.group(1)
dec_s = self._decrypt_signature(s, video_id, player_url, age_gate)
return '/signature/%s' % dec_s
dash_manifest_url = re.sub(r'/s/([\w\.]+)', decrypt_sig, dash_manifest_url)
dash_doc = self._download_xml(
dash_manifest_url, video_id,
note=u'Downloading DASH manifest',
errnote=u'Could not download DASH manifest')
for r in dash_doc.findall(u'.//{urn:mpeg:DASH:schema:MPD:2011}Representation'):
url_el = r.find('{urn:mpeg:DASH:schema:MPD:2011}BaseURL')
if url_el is None:
continue
format_id = r.attrib['id']
video_url = url_el.text
filesize = int_or_none(url_el.attrib.get('{http://youtube.com/yt/2012/10/10}contentLength'))
f = {
'format_id': format_id,
'url': video_url,
'width': int_or_none(r.attrib.get('width')),
'tbr': int_or_none(r.attrib.get('bandwidth'), 1000),
'asr': int_or_none(r.attrib.get('audioSamplingRate')),
'filesize': filesize,
}
try:
existing_format = next(
fo for fo in formats
if fo['format_id'] == format_id)
except StopIteration:
f.update(self._formats.get(format_id, {}))
formats.append(f)
else:
existing_format.update(f)
except (ExtractorError, KeyError) as e:
self.report_warning(u'Skipping DASH manifest: %s' % e, video_id)
2013-12-23 03:51:42 +00:00
self._sort_formats(formats)
return {
'id': video_id,
'uploader': video_uploader,
'uploader_id': video_uploader_id,
'upload_date': upload_date,
'title': video_title,
'thumbnail': video_thumbnail,
'description': video_description,
'subtitles': video_subtitles,
'duration': video_duration,
'age_limit': 18 if age_gate else 0,
'annotations': video_annotations,
'webpage_url': proto + '://www.youtube.com/watch?v=%s' % video_id,
'view_count': view_count,
'like_count': like_count,
'dislike_count': dislike_count,
'formats': formats,
}
class YoutubePlaylistIE(YoutubeBaseInfoExtractor):
IE_DESC = u'YouTube.com playlists'
_VALID_URL = r"""(?x)(?:
(?:https?://)?
(?:\w+\.)?
youtube\.com/
(?:
(?:course|view_play_list|my_playlists|artist|playlist|watch)
\? (?:.*?&)*? (?:p|a|list)=
| p/
)
(
(?:PL|EC|UU|FL|RD)?[0-9A-Za-z-_]{10,}
# Top tracks, they can also include dots
|(?:MC)[\w\.]*
)
.*
|
((?:PL|EC|UU|FL|RD)[0-9A-Za-z-_]{10,})
)"""
_TEMPLATE_URL = 'https://www.youtube.com/playlist?list=%s'
_MORE_PAGES_INDICATOR = r'data-link-type="next"'
_VIDEO_RE = r'href="\s*/watch\?v=(?P<id>[0-9A-Za-z_-]{11})&amp;[^"]*?index=(?P<index>\d+)'
IE_NAME = u'youtube:playlist'
def _real_initialize(self):
self._login()
def _ids_to_results(self, ids):
return [self.url_result(vid_id, 'Youtube', video_id=vid_id)
for vid_id in ids]
def _extract_mix(self, playlist_id):
# The mixes are generated from a a single video
# the id of the playlist is just 'RD' + video_id
url = 'https://youtube.com/watch?v=%s&list=%s' % (playlist_id[-11:], playlist_id)
webpage = self._download_webpage(url, playlist_id, u'Downloading Youtube mix')
search_title = lambda class_name: get_element_by_attribute('class', class_name, webpage)
title_span = (search_title('playlist-title') or
search_title('title long-title') or search_title('title'))
title = clean_html(title_span)
video_re = r'''(?x)data-video-username="(.*?)".*?
href="/watch\?v=([0-9A-Za-z_-]{11})&amp;[^"]*?list=%s''' % re.escape(playlist_id)
matches = orderedSet(re.findall(video_re, webpage, flags=re.DOTALL))
# Some of the videos may have been deleted, their username field is empty
ids = [video_id for (username, video_id) in matches if username]
url_results = self._ids_to_results(ids)
return self.playlist_result(url_results, playlist_id, title)
def _real_extract(self, url):
# Extract playlist id
mobj = re.match(self._VALID_URL, url)
if mobj is None:
raise ExtractorError(u'Invalid URL: %s' % url)
playlist_id = mobj.group(1) or mobj.group(2)
# Check if it's a video-specific URL
query_dict = compat_urlparse.parse_qs(compat_urlparse.urlparse(url).query)
if 'v' in query_dict:
video_id = query_dict['v'][0]
if self._downloader.params.get('noplaylist'):
self.to_screen(u'Downloading just video %s because of --no-playlist' % video_id)
return self.url_result(video_id, 'Youtube', video_id=video_id)
else:
self.to_screen(u'Downloading playlist %s - add --no-playlist to just download video %s' % (playlist_id, video_id))
if playlist_id.startswith('RD'):
# Mixes require a custom extraction process
return self._extract_mix(playlist_id)
if playlist_id.startswith('TL'):
raise ExtractorError(u'For downloading YouTube.com top lists, use '
u'the "yttoplist" keyword, for example "youtube-dl \'yttoplist:music:Top Tracks\'"', expected=True)
url = self._TEMPLATE_URL % playlist_id
page = self._download_webpage(url, playlist_id)
more_widget_html = content_html = page
# Check if the playlist exists or is private
if re.search(r'<div class="yt-alert-message">[^<]*?(The|This) playlist (does not exist|is private)[^<]*?</div>', page) is not None:
raise ExtractorError(
u'The playlist doesn\'t exist or is private, use --username or '
'--netrc to access it.',
expected=True)
# Extract the video ids from the playlist pages
ids = []
for page_num in itertools.count(1):
matches = re.finditer(self._VIDEO_RE, content_html)
# We remove the duplicates and the link with index 0
# (it's not the first video of the playlist)
new_ids = orderedSet(m.group('id') for m in matches if m.group('index') != '0')
ids.extend(new_ids)
mobj = re.search(r'data-uix-load-more-href="/?(?P<more>[^"]+)"', more_widget_html)
if not mobj:
break
more = self._download_json(
'https://youtube.com/%s' % mobj.group('more'), playlist_id,
'Downloading page #%s' % page_num,
transform_source=uppercase_escape)
content_html = more['content_html']
more_widget_html = more['load_more_widget_html']
playlist_title = self._html_search_regex(
r'(?s)<h1 class="pl-header-title[^"]*">\s*(.*?)\s*</h1>',
page, u'title')
url_results = self._ids_to_results(ids)
return self.playlist_result(url_results, playlist_id, playlist_title)
class YoutubeTopListIE(YoutubePlaylistIE):
IE_NAME = u'youtube:toplist'
IE_DESC = (u'YouTube.com top lists, "yttoplist:{channel}:{list title}"'
u' (Example: "yttoplist:music:Top Tracks")')
_VALID_URL = r'yttoplist:(?P<chann>.*?):(?P<title>.*?)$'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
channel = mobj.group('chann')
title = mobj.group('title')
query = compat_urllib_parse.urlencode({'title': title})
playlist_re = 'href="([^"]+?%s.*?)"' % re.escape(query)
channel_page = self._download_webpage('https://www.youtube.com/%s' % channel, title)
link = self._html_search_regex(playlist_re, channel_page, u'list')
url = compat_urlparse.urljoin('https://www.youtube.com/', link)
video_re = r'data-index="\d+".*?data-video-id="([0-9A-Za-z_-]{11})"'
ids = []
# sometimes the webpage doesn't contain the videos
# retry until we get them
for i in itertools.count(0):
msg = u'Downloading Youtube mix'
if i > 0:
msg += ', retry #%d' % i
webpage = self._download_webpage(url, title, msg)
ids = orderedSet(re.findall(video_re, webpage))
if ids:
break
url_results = self._ids_to_results(ids)
return self.playlist_result(url_results, playlist_title=title)
class YoutubeChannelIE(InfoExtractor):
IE_DESC = u'YouTube.com channels'
_VALID_URL = r"^(?:https?://)?(?:youtu\.be|(?:\w+\.)?youtube(?:-nocookie)?\.com)/channel/([0-9A-Za-z_-]+)"
_MORE_PAGES_INDICATOR = 'yt-uix-load-more'
2014-01-25 16:11:55 +00:00
_MORE_PAGES_URL = 'https://www.youtube.com/c4_browse_ajax?action_load_more_videos=1&flow=list&paging=%s&view=0&sort=da&channel_id=%s'
IE_NAME = u'youtube:channel'
def extract_videos_from_page(self, page):
ids_in_page = []
for mobj in re.finditer(r'href="/watch\?v=([0-9A-Za-z_-]+)&?', page):
if mobj.group(1) not in ids_in_page:
ids_in_page.append(mobj.group(1))
return ids_in_page
def _real_extract(self, url):
# Extract channel id
mobj = re.match(self._VALID_URL, url)
if mobj is None:
raise ExtractorError(u'Invalid URL: %s' % url)
# Download channel page
channel_id = mobj.group(1)
video_ids = []
url = 'https://www.youtube.com/channel/%s/videos' % channel_id
channel_page = self._download_webpage(url, channel_id)
autogenerated = re.search(r'''(?x)
class="[^"]*?(?:
channel-header-autogenerated-label|
yt-channel-title-autogenerated
)[^"]*"''', channel_page) is not None
if autogenerated:
# The videos are contained in a single page
# the ajax pages can't be used, they are empty
video_ids = self.extract_videos_from_page(channel_page)
else:
# Download all channel pages using the json-based channel_ajax query
for pagenum in itertools.count(1):
url = self._MORE_PAGES_URL % (pagenum, channel_id)
page = self._download_json(
url, channel_id, note=u'Downloading page #%s' % pagenum,
transform_source=uppercase_escape)
ids_in_page = self.extract_videos_from_page(page['content_html'])
video_ids.extend(ids_in_page)
if self._MORE_PAGES_INDICATOR not in page['load_more_widget_html']:
break
self._downloader.to_screen(u'[youtube] Channel %s: Found %i videos' % (channel_id, len(video_ids)))
url_entries = [self.url_result(video_id, 'Youtube', video_id=video_id)
for video_id in video_ids]
return self.playlist_result(url_entries, channel_id)
class YoutubeUserIE(InfoExtractor):
IE_DESC = u'YouTube.com user videos (URL or "ytuser" keyword)'
_VALID_URL = r'(?:(?:(?:https?://)?(?:\w+\.)?youtube\.com/(?:user/)?(?!(?:attribution_link|watch|results)(?:$|[^a-z_A-Z0-9-])))|ytuser:)(?!feed/)([A-Za-z0-9_-]+)'
2014-01-25 16:11:55 +00:00
_TEMPLATE_URL = 'https://gdata.youtube.com/feeds/api/users/%s'
_GDATA_PAGE_SIZE = 50
2014-01-25 16:11:55 +00:00
_GDATA_URL = 'https://gdata.youtube.com/feeds/api/users/%s/uploads?max-results=%d&start-index=%d&alt=json'
IE_NAME = u'youtube:user'
@classmethod
def suitable(cls, url):
# Don't return True if the url can be extracted with other youtube
# extractor, the regex would is too permissive and it would match.
other_ies = iter(klass for (name, klass) in globals().items() if name.endswith('IE') and klass is not cls)
if any(ie.suitable(url) for ie in other_ies): return False
else: return super(YoutubeUserIE, cls).suitable(url)
def _real_extract(self, url):
# Extract username
mobj = re.match(self._VALID_URL, url)
if mobj is None:
raise ExtractorError(u'Invalid URL: %s' % url)
username = mobj.group(1)
# Download video ids using YouTube Data API. Result size per
# query is limited (currently to 50 videos) so we need to query
# page by page until there are no video ids - it means we got
# all of them.
def download_page(pagenum):
start_index = pagenum * self._GDATA_PAGE_SIZE + 1
gdata_url = self._GDATA_URL % (username, self._GDATA_PAGE_SIZE, start_index)
page = self._download_webpage(
gdata_url, username,
u'Downloading video ids from %d to %d' % (
start_index, start_index + self._GDATA_PAGE_SIZE))
try:
response = json.loads(page)
except ValueError as err:
raise ExtractorError(u'Invalid JSON in API response: ' + compat_str(err))
if 'entry' not in response['feed']:
return
# Extract video identifiers
2013-12-22 02:57:42 +00:00
entries = response['feed']['entry']
for entry in entries:
title = entry['title']['$t']
video_id = entry['id']['$t'].split('/')[-1]
yield {
2013-12-22 02:57:42 +00:00
'_type': 'url',
'url': video_id,
'ie_key': 'Youtube',
'id': video_id,
2013-12-22 02:57:42 +00:00
'title': title,
}
url_results = PagedList(download_page, self._GDATA_PAGE_SIZE)
return self.playlist_result(url_results, playlist_title=username)
class YoutubeSearchIE(SearchInfoExtractor):
IE_DESC = u'YouTube.com searches'
2014-03-30 13:53:06 +00:00
_API_URL = u'https://gdata.youtube.com/feeds/api/videos?q=%s&start-index=%i&max-results=50&v=2&alt=jsonc'
_MAX_RESULTS = 1000
IE_NAME = u'youtube:search'
_SEARCH_KEY = 'ytsearch'
def _get_n_results(self, query, n):
"""Get a specified number of results for a query"""
video_ids = []
pagenum = 0
limit = n
2014-03-30 13:53:06 +00:00
PAGE_SIZE = 50
2014-03-30 13:53:06 +00:00
while (PAGE_SIZE * pagenum) < limit:
result_url = self._API_URL % (
compat_urllib_parse.quote_plus(query.encode('utf-8')),
(PAGE_SIZE * pagenum) + 1)
data_json = self._download_webpage(
result_url, video_id=u'query "%s"' % query,
note=u'Downloading page %s' % (pagenum + 1),
errnote=u'Unable to download API page')
data = json.loads(data_json)
api_response = data['data']
if 'items' not in api_response:
raise ExtractorError(
u'[youtube] No video results', expected=True)
new_ids = list(video['id'] for video in api_response['items'])
video_ids += new_ids
limit = min(n, api_response['totalItems'])
pagenum += 1
if len(video_ids) > n:
video_ids = video_ids[:n]
videos = [self.url_result(video_id, 'Youtube', video_id=video_id)
for video_id in video_ids]
return self.playlist_result(videos, query)
class YoutubeSearchDateIE(YoutubeSearchIE):
IE_NAME = YoutubeSearchIE.IE_NAME + ':date'
_API_URL = 'https://gdata.youtube.com/feeds/api/videos?q=%s&start-index=%i&max-results=50&v=2&alt=jsonc&orderby=published'
_SEARCH_KEY = 'ytsearchdate'
IE_DESC = u'YouTube.com searches, newest videos first'
class YoutubeSearchURLIE(InfoExtractor):
IE_DESC = u'YouTube.com search URLs'
IE_NAME = u'youtube:search_url'
_VALID_URL = r'https?://(?:www\.)?youtube\.com/results\?(.*?&)?search_query=(?P<query>[^&]+)(?:[&]|$)'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
query = compat_urllib_parse.unquote_plus(mobj.group('query'))
webpage = self._download_webpage(url, query)
result_code = self._search_regex(
r'(?s)<ol id="search-results"(.*?)</ol>', webpage, u'result HTML')
part_codes = re.findall(
r'(?s)<h3 class="yt-lockup-title">(.*?)</h3>', result_code)
entries = []
for part_code in part_codes:
part_title = self._html_search_regex(
r'(?s)title="([^"]+)"', part_code, 'item title', fatal=False)
part_url_snippet = self._html_search_regex(
r'(?s)href="([^"]+)"', part_code, 'item URL')
part_url = compat_urlparse.urljoin(
'https://www.youtube.com/', part_url_snippet)
entries.append({
'_type': 'url',
'url': part_url,
'title': part_title,
})
return {
'_type': 'playlist',
'entries': entries,
'title': query,
}
class YoutubeShowIE(InfoExtractor):
IE_DESC = u'YouTube.com (multi-season) shows'
_VALID_URL = r'https?://www\.youtube\.com/show/(.*)'
IE_NAME = u'youtube:show'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
show_name = mobj.group(1)
webpage = self._download_webpage(url, show_name, u'Downloading show webpage')
# There's one playlist for each season of the show
m_seasons = list(re.finditer(r'href="(/playlist\?list=.*?)"', webpage))
self.to_screen(u'%s: Found %s seasons' % (show_name, len(m_seasons)))
return [self.url_result('https://www.youtube.com' + season.group(1), 'YoutubePlaylist') for season in m_seasons]
class YoutubeFeedsInfoExtractor(YoutubeBaseInfoExtractor):
"""
Base class for extractors that fetch info from
http://www.youtube.com/feed_ajax
Subclasses must define the _FEED_NAME and _PLAYLIST_TITLE properties.
"""
_LOGIN_REQUIRED = True
# use action_load_personal_feed instead of action_load_system_feed
_PERSONAL_FEED = False
@property
def _FEED_TEMPLATE(self):
action = 'action_load_system_feed'
if self._PERSONAL_FEED:
action = 'action_load_personal_feed'
2014-01-25 16:11:55 +00:00
return 'https://www.youtube.com/feed_ajax?%s=1&feed_name=%s&paging=%%s' % (action, self._FEED_NAME)
@property
def IE_NAME(self):
return u'youtube:%s' % self._FEED_NAME
def _real_initialize(self):
self._login()
def _real_extract(self, url):
feed_entries = []
paging = 0
for i in itertools.count(1):
info = self._download_json(self._FEED_TEMPLATE % paging,
u'%s feed' % self._FEED_NAME,
u'Downloading page %s' % i)
feed_html = info.get('feed_html') or info.get('content_html')
m_ids = re.finditer(r'"/watch\?v=(.*?)["&]', feed_html)
ids = orderedSet(m.group(1) for m in m_ids)
feed_entries.extend(
self.url_result(video_id, 'Youtube', video_id=video_id)
for video_id in ids)
if info['paging'] is None:
break
paging = info['paging']
return self.playlist_result(feed_entries, playlist_title=self._PLAYLIST_TITLE)
class YoutubeSubscriptionsIE(YoutubeFeedsInfoExtractor):
IE_DESC = u'YouTube.com subscriptions feed, "ytsubs" keyword (requires authentication)'
_VALID_URL = r'https?://www\.youtube\.com/feed/subscriptions|:ytsubs(?:criptions)?'
_FEED_NAME = 'subscriptions'
_PLAYLIST_TITLE = u'Youtube Subscriptions'
class YoutubeRecommendedIE(YoutubeFeedsInfoExtractor):
IE_DESC = u'YouTube.com recommended videos, "ytrec" keyword (requires authentication)'
_VALID_URL = r'https?://www\.youtube\.com/feed/recommended|:ytrec(?:ommended)?'
_FEED_NAME = 'recommended'
_PLAYLIST_TITLE = u'Youtube Recommended videos'
class YoutubeWatchLaterIE(YoutubeFeedsInfoExtractor):
IE_DESC = u'Youtube watch later list, "ytwatchlater" keyword (requires authentication)'
_VALID_URL = r'https?://www\.youtube\.com/feed/watch_later|:ytwatchlater'
_FEED_NAME = 'watch_later'
_PLAYLIST_TITLE = u'Youtube Watch Later'
_PERSONAL_FEED = True
class YoutubeHistoryIE(YoutubeFeedsInfoExtractor):
IE_DESC = u'Youtube watch history, "ythistory" keyword (requires authentication)'
_VALID_URL = u'https?://www\.youtube\.com/feed/history|:ythistory'
_FEED_NAME = 'history'
_PERSONAL_FEED = True
_PLAYLIST_TITLE = u'Youtube Watch History'
class YoutubeFavouritesIE(YoutubeBaseInfoExtractor):
IE_NAME = u'youtube:favorites'
IE_DESC = u'YouTube.com favourite videos, "ytfav" keyword (requires authentication)'
_VALID_URL = r'https?://www\.youtube\.com/my_favorites|:ytfav(?:ou?rites)?'
_LOGIN_REQUIRED = True
def _real_extract(self, url):
webpage = self._download_webpage('https://www.youtube.com/my_favorites', 'Youtube Favourites videos')
playlist_id = self._search_regex(r'list=(.+?)["&]', webpage, u'favourites playlist id')
return self.url_result(playlist_id, 'YoutubePlaylist')
class YoutubeTruncatedURLIE(InfoExtractor):
IE_NAME = 'youtube:truncated_url'
IE_DESC = False # Do not list
_VALID_URL = r'''(?x)
(?:https?://)?[^/]+/watch\?(?:feature=[a-z_]+)?$|
(?:https?://)?(?:www\.)?youtube\.com/attribution_link\?a=[^&]+$
'''
def _real_extract(self, url):
raise ExtractorError(
u'Did you forget to quote the URL? Remember that & is a meta '
u'character in most shells, so you want to put the URL in quotes, '
u'like youtube-dl '
u'"http://www.youtube.com/watch?feature=foo&v=BaW_jenozKc" '
u' or simply youtube-dl BaW_jenozKc .',
expected=True)