# coding: utf-8
import collections
-import itertools
+import errno
import io
+import itertools
import json
-import netrc
+import os.path
import re
import socket
import string
import struct
import traceback
+import xml.etree.ElementTree
import zlib
from .common import InfoExtractor, SearchInfoExtractor
from .subtitles import SubtitlesInfoExtractor
from ..utils import (
+ compat_chr,
compat_http_client,
compat_parse_qs,
compat_urllib_error,
unescapeHTML,
unified_strdate,
orderedSet,
+ write_json_file,
)
class YoutubeBaseInfoExtractor(InfoExtractor):
def __init__(self, *args, **kwargs):
super(YoutubeIE, self).__init__(*args, **kwargs)
- self._jsplayer_cache = {}
+ self._player_cache = {}
def report_video_webpage_download(self, video_id):
"""Report attempt to download video webpage."""
"""Indicate the download will use the RTMP protocol."""
self.to_screen(u'RTMP download detected')
- def _extract_signature_function(self, video_id, player_url):
- id_m = re.match(r'.*-(?P<id>[^.]+)\.(?P<ext>[^.]+)$', player_url)
+ def _extract_signature_function(self, video_id, player_url, slen):
+ id_m = re.match(r'.*-(?P<id>[a-zA-Z0-9_-]+)\.(?P<ext>[a-z]+)$',
+ 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 = self._downloader.params.get('cachedir',
+ u'~/.youtube-dl/cache')
+
+ if cache_dir != u'NONE':
+ 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
+
if player_type == 'js':
code = self._download_webpage(
player_url, video_id,
- note=u'Downloading %s player %s' % (player_type, jsplayer_id),
+ note=u'Downloading %s player %s' % (player_type, player_id),
errnote=u'Download of %s failed' % player_url)
- return self._parse_sig_js(code)
- elif player_tpye == 'swf':
+ res = self._parse_sig_js(code)
+ elif player_type == 'swf':
urlh = self._request_webpage(
player_url, video_id,
- note=u'Downloading %s player %s' % (player_type, jsplayer_id),
+ note=u'Downloading %s player %s' % (player_type, player_id),
errnote=u'Download of %s failed' % player_url)
code = urlh.read()
- return self._parse_sig_swf(code)
+ res = self._parse_sig_swf(code)
else:
assert False, 'Invalid player type %r' % player_type
+ if cache_dir is not False:
+ try:
+ cache_res = res(map(compat_chr, range(slen)))
+ 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)
+ except Exception:
+ tb = traceback.format_exc()
+ self._downloader.report_warning(
+ u'Writing cache to %r failed: %s' % (cache_fn, tb))
+
+ 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)
+ steps = u'' if step == 1 else (':%d' % step)
+ return u's[%s%s%s]' % (starts, ends, steps)
+
+ step = None
+ 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)
+
+ cache_res = func(map(compat_chr, range(slen)))
+ 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:\n' + code)
+
def _parse_sig_js(self, jscode):
funcname = self._search_regex(
r'signature=([a-zA-Z]+)', jscode,
def interpret_statement(stmt, local_vars, allow_recursion=20):
if allow_recursion < 0:
- raise ExctractorError(u'Recursion limit reached')
+ raise ExtractorError(u'Recursion limit reached')
if stmt.startswith(u'var '):
stmt = stmt[len(u'var '):]
for tag_code, tag in extract_tags(content)
if tag_code == 82)
p = code_tag.index(b'\0', 4) + 1
+ code_reader = io.BytesIO(code_tag[p:])
# Parse ABC (AVM2 ByteCode)
- def read_int(data=None, pos=None):
- if hasattr(data, 'read'):
- assert pos is None
-
- res = 0
- shift = 0
- for _ in range(5):
- buf = data.read(1)
- assert len(buf) == 1
- b = struct.unpack('<B', buf)[0]
- res = res | ((b & 0x7f) << shift)
- if b & 0x80 == 0:
- break
- shift += 7
- return res
-
- if data is None:
- data = code_tag
- if pos is None:
- pos = p
+ def read_int(reader=None):
+ if reader is None:
+ reader = code_reader
res = 0
shift = 0
for _ in range(5):
- b = struct.unpack('<B', data[pos:pos+1])[0]
- pos += 1
+ 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
- return (res, pos)
- assert read_int(b'\x00', 0) == (0, 1)
- assert read_int(b'\x10', 0) == (16, 1)
- assert read_int(b'\x34', 0) == (0x34, 1)
- assert read_int(b'\xb4\x12', 0) == (0x12 * 0x80 + 0x34, 2)
- assert read_int(b'\xff\xff\xff\x00', 0) == (0x1fffff, 4)
-
- def u30(*args, **kwargs):
- res = read_int(*args, **kwargs)
- if isinstance(res, tuple):
- assert res[0] & 0xf0000000 == 0
- else:
- assert res & 0xf0000000 == 0
+ return res
+
+ def u30(reader=None):
+ res = read_int(reader)
+ assert res & 0xf0000000 == 0
return res
u32 = read_int
- def s32(data=None, pos=None):
- v, pos = read_int(data, pos)
+ def s32(reader=None):
+ v = read_int(reader)
if v & 0x80000000 != 0:
v = - ((v ^ 0xffffffff) + 1)
- return (v, pos)
- assert s32(b'\xff\xff\xff\xff\x0f', 0) == (-1, 5)
-
- def string():
- slen, p = u30()
- return (code_tag[p:p+slen].decode('utf-8'), p + slen)
-
- def read_byte(data=None, pos=None):
- if data is None:
- data = code_tag
- if pos is None:
- pos = p
- res = struct.unpack('<B', data[pos:pos+1])[0]
- return (res, pos + 1)
+ return v
+
+ def read_string(reader=None):
+ 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
- p += 2 + 2
+ read_bytes(2 + 2)
# Constant pool
- int_count, p = u30()
+ int_count = u30()
for _c in range(1, int_count):
- _, p = s32()
- uint_count, p = u30()
+ s32()
+ uint_count = u30()
for _c in range(1, uint_count):
- _, p = u32()
- double_count, p = u30()
- p += (double_count-1) * 8
- string_count, p = u30()
+ u32()
+ double_count = u30()
+ read_bytes((double_count-1) * 8)
+ string_count = u30()
constant_strings = [u'']
for _c in range(1, string_count):
- s, p = string()
+ s = read_string()
constant_strings.append(s)
- namespace_count, p = u30()
+ namespace_count = u30()
for _c in range(1, namespace_count):
- p += 1 # kind
- _, p = u30() # name
- ns_set_count, p = u30()
+ read_bytes(1) # kind
+ u30() # name
+ ns_set_count = u30()
for _c in range(1, ns_set_count):
- count, p = u30()
+ count = u30()
for _c2 in range(count):
- _, p = u30()
- multiname_count, p = u30()
+ u30()
+ multiname_count = u30()
MULTINAME_SIZES = {
0x07: 2, # QName
0x0d: 2, # QNameA
}
multinames = [u'']
for _c in range(1, multiname_count):
- kind, p = u30()
+ kind = u30()
assert kind in MULTINAME_SIZES, u'Invalid multiname kind %r' % kind
if kind == 0x07:
- namespace_idx, p = u30()
- name_idx, p = u30()
+ u30() # namespace_idx
+ name_idx = u30()
multinames.append(constant_strings[name_idx])
else:
multinames.append('[MULTINAME kind: %d]' % kind)
for _c2 in range(MULTINAME_SIZES[kind]):
- _, p = u30()
+ u30()
# Methods
- method_count, p = u30()
+ method_count = u30()
MethodInfo = collections.namedtuple(
'MethodInfo',
['NEED_ARGUMENTS', 'NEED_REST'])
method_infos = []
for method_id in range(method_count):
- param_count, p = u30()
- _, p = u30() # return type
+ param_count = u30()
+ u30() # return type
for _ in range(param_count):
- _, p = u30() # param type
- _, p = u30() # name index (always 0 for youtube)
- flags, p = read_byte()
+ u30() # param type
+ u30() # name index (always 0 for youtube)
+ flags = read_byte()
if flags & 0x08 != 0:
# Options present
- option_count, p = u30()
+ option_count = u30()
for c in range(option_count):
- _, p = u30() # val
- p += 1 # kind
+ u30() # val
+ read_bytes(1) # kind
if flags & 0x80 != 0:
# Param names present
for _ in range(param_count):
- _, p = u30() # param name
+ u30() # param name
mi = MethodInfo(flags & 0x01 != 0, flags & 0x04 != 0)
method_infos.append(mi)
# Metadata
- metadata_count, p = u30()
+ metadata_count = u30()
for _c in range(metadata_count):
- _, p = u30() # name
- item_count, p = u30()
+ u30() # name
+ item_count = u30()
for _c2 in range(item_count):
- _, p = u30() # key
- _, p = u30() # value
-
- def parse_traits_info(pos=None):
- if pos is None:
- pos = p
- trait_name_idx, pos = u30(pos=pos)
- kind_full, pos = read_byte(pos=pos)
+ u30() # key
+ u30() # value
+
+ 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
- _, pos = u30(pos=pos) # Slot id
- type_name_idx, pos = u30(pos=pos)
- vindex, pos = u30(pos=pos)
+ u30() # Slot id
+ u30() # type_name_idx
+ vindex = u30()
if vindex != 0:
- _, pos = read_byte(pos=pos) # vkind
+ read_byte() # vkind
elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter
- _, pos = u30(pos=pos) # disp_id
- method_idx, pos = u30(pos=pos)
+ u30() # disp_id
+ method_idx = u30()
methods[multinames[trait_name_idx]] = method_idx
elif kind == 0x04: # Class
- _, pos = u30(pos=pos) # slot_id
- _, pos = u30(pos=pos) # classi
+ u30() # slot_id
+ u30() # classi
elif kind == 0x05: # Function
- _, pos = u30(pos=pos) # slot_id
- function_idx, pos = u30(pos=pos)
+ u30() # slot_id
+ 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
- metadata_count, pos = u30(pos=pos)
+ metadata_count = u30()
for _c3 in range(metadata_count):
- _, pos = u30(pos=pos)
+ u30() # metadata index
- return (methods, pos)
+ return methods
# Classes
TARGET_CLASSNAME = u'SignatureDecipher'
searched_idx = multinames.index(TARGET_CLASSNAME)
searched_class_id = None
- class_count, p = u30()
+ class_count = u30()
for class_id in range(class_count):
- name_idx, p = u30()
+ name_idx = u30()
if name_idx == searched_idx:
# We found the class we're looking for!
searched_class_id = class_id
- _, p = u30() # super_name idx
- flags, p = read_byte()
+ u30() # super_name idx
+ flags = read_byte()
if flags & 0x08 != 0: # Protected namespace is present
- protected_ns_idx, p = u30()
- intrf_count, p = u30()
+ u30() # protected_ns_idx
+ intrf_count = u30()
for _c2 in range(intrf_count):
- _, p = u30()
- _, p = u30() # iinit
- trait_count, p = u30()
+ u30()
+ u30() # iinit
+ trait_count = u30()
for _c2 in range(trait_count):
- _, p = parse_traits_info()
+ parse_traits_info()
if searched_class_id is None:
raise ExtractorError(u'Target class %r not found' %
method_names = {}
method_idxs = {}
for class_id in range(class_count):
- _, p = u30() # cinit
- trait_count, p = u30()
+ u30() # cinit
+ trait_count = u30()
for _c2 in range(trait_count):
- trait_methods, p = parse_traits_info()
+ trait_methods = parse_traits_info()
if class_id == searched_class_id:
method_names.update(trait_methods.items())
method_idxs.update(dict(
for name, idx in trait_methods.items()))
# Scripts
- script_count, p = u30()
+ script_count = u30()
for _c in range(script_count):
- _, p = u30() # init
- trait_count, p = u30()
+ u30() # init
+ trait_count = u30()
for _c2 in range(trait_count):
- _, p = parse_traits_info()
+ parse_traits_info()
# Method bodies
- method_body_count, p = u30()
+ method_body_count = u30()
Method = collections.namedtuple('Method', ['code', 'local_count'])
methods = {}
for _c in range(method_body_count):
- method_idx, p = u30()
- max_stack, p = u30()
- local_count, p = u30()
- init_scope_depth, p = u30()
- max_scope_depth, p = u30()
- code_length, p = u30()
+ method_idx = u30()
+ u30() # max_stack
+ local_count = u30()
+ u30() # init_scope_depth
+ u30() # max_scope_depth
+ code_length = u30()
+ code = read_bytes(code_length)
if method_idx in method_idxs:
- m = Method(code_tag[p:p+code_length], local_count)
+ m = Method(code, local_count)
methods[method_idxs[method_idx]] = m
- p += code_length
- exception_count, p = u30()
+ exception_count = u30()
for _c2 in range(exception_count):
- _, p = u30() # from
- _, p = u30() # to
- _, p = u30() # target
- _, p = u30() # exc_type
- _, p = u30() # var_name
- trait_count, p = u30()
+ u30() # from
+ u30() # to
+ u30() # target
+ u30() # exc_type
+ u30() # var_name
+ trait_count = u30()
for _c2 in range(trait_count):
- _, p = parse_traits_info()
+ parse_traits_info()
- assert p == len(code_tag)
+ assert p + code_reader.tell() == len(code_tag)
assert len(methods) == len(method_idxs)
method_pyfunctions = {}
m = methods[func_name]
def resfunc(args):
- print('Entering function %s(%r)' % (func_name, args))
registers = ['(this)'] + list(args) + [None] * m.local_count
stack = []
coder = io.BytesIO(m.code)
assert isinstance(obj, list)
stack.append(obj[idx])
elif opcode == 128: # coerce
- _ = u30(coder)
+ u30(coder)
elif opcode == 133: # coerce_s
assert isinstance(stack[-1], (type(None), compat_str))
elif opcode == 164: # modulo
initial_function = extract_function(u'decipher')
return lambda s: initial_function([s])
- def _decrypt_signature(self, s, video_id, jsplayer_url, age_gate=False):
+ def _decrypt_signature(self, s, video_id, player_url, age_gate=False):
"""Turn the encrypted s field into a working signature"""
- if jsplayer_url is not None:
+ if player_url is not None:
try:
- if jsplayer_url not in self._jsplayer_cache:
- self._jsplayer_cache[jsplayer_url] = self._extract_signature_function(
- video_id, jsplayer_url
+ if player_url not in self._player_cache:
+ func = self._extract_signature_function(
+ video_id, player_url, len(s)
)
- return self._jsplayer_cache[jsplayer_url]([s])
- except Exception as e:
+ self._player_cache[player_url] = func
+ func = self._player_cache[player_url]
+ if self._downloader.params.get('youtube_print_sig_code'):
+ self._print_sig_code(func, len(s))
+ return func(s)
+ except Exception:
tb = traceback.format_exc()
- self._downloader.report_warning(u'Automatic signature extraction failed: ' + tb)
+ self._downloader.report_warning(
+ u'Automatic signature extraction failed: ' + tb)
- self._downloader.report_warning(u'Warning: Falling back to static signature algorithm')
+ self._downloader.report_warning(
+ u'Warning: Falling back to static signature algorithm')
+ return self._static_decrypt_signature(
+ s, video_id, player_url, age_gate)
+ 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) == 92:
+ 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]
+ 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:
encrypted_sig = url_data['s'][0]
if self._downloader.params.get('verbose'):
if age_gate:
- player_version = self._search_regex(r'-(.+)\.swf$',
- player_url if player_url else 'NOT FOUND',
+ player_version = self._search_regex(
+ r'-(.+)\.swf$',
+ player_url if player_url else None,
'flash player', fatal=False)
player_desc = 'flash player %s' % player_version
else:
- player_version = self._search_regex(r'html5player-(.+?)\.js', video_webpage,
+ player_version = self._search_regex(
+ r'html5player-(.+?)\.js', video_webpage,
'html5 player', fatal=False)
player_desc = u'html5 player %s' % player_version
self.to_screen(u'encrypted signature length %d (%s), itag %s, %s' %
(len(encrypted_sig), parts_sizes, url_data['itag'][0], player_desc))
- if age_gate:
- jsplayer_url = None
- else:
+ if not age_gate:
jsplayer_url_json = self._search_regex(
r'"assets":.+?"js":\s*("[^"]+")',
video_webpage, u'JS player URL')
- jsplayer_url = json.loads(jsplayer_url_json)
+ player_url = json.loads(jsplayer_url_json)
- signature = self._decrypt_signature(encrypted_sig, video_id, jsplayer_url, age_gate)
+ signature = self._decrypt_signature(
+ encrypted_sig, video_id, player_url, age_gate)
url += '&signature=' + signature
if 'ratebypass' not in url:
url += '&ratebypass=yes'