X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=blobdiff_plain;f=youtube_dl%2Fextractor%2Fyoutube.py;h=9b09793eb307b4ca899942e57cfa3dda88ec5caa;hb=3f8ced5144a76a3f9ab7ee8cd06cc79bb75dc564;hp=ec1cf8d30c8d2b65d4e17016ad02226058b2a68f;hpb=6f56389b8836301fc64f849e43ebd05043c0a66d;p=youtube-dl diff --git a/youtube_dl/extractor/youtube.py b/youtube_dl/extractor/youtube.py index ec1cf8d30..9b09793eb 100644 --- a/youtube_dl/extractor/youtube.py +++ b/youtube_dl/extractor/youtube.py @@ -1,28 +1,39 @@ # coding: utf-8 +import collections +import errno +import io +import itertools import json -import netrc +import os.path import re import socket -import itertools +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, compat_urllib_parse, compat_urllib_request, + compat_urlparse, compat_str, clean_html, + get_cachedir, get_element_by_id, ExtractorError, unescapeHTML, unified_strdate, orderedSet, + write_json_file, ) class YoutubeBaseInfoExtractor(InfoExtractor): @@ -63,14 +74,8 @@ class YoutubeBaseInfoExtractor(InfoExtractor): self._downloader.report_warning(u'unable to fetch login page: %s' % compat_str(err)) return False - galx = None - dsh = None - match = re.search(re.compile(r'[a-zA-Z0-9_-]+)\.(?P[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 = 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 + + if player_type == 'js': + code = self._download_webpage( + player_url, video_id, + note=u'Downloading %s player %s' % (player_type, player_id), + errnote=u'Download of %s failed' % player_url) + 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, player_id), + errnote=u'Download of %s failed' % player_url) + code = urlh.read() + res = self._parse_sig_swf(code) + else: + assert False, 'Invalid player type %r' % player_type + + if cache_enabled: + try: + 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) + 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)) 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 + 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) + + 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') + + functions = {} + + def argidx(varname): + return string.lowercase.index(varname) + + def interpret_statement(stmt, local_vars, allow_recursion=20): + if allow_recursion < 0: + raise ExtractorError(u'Recursion limit reached') + + if stmt.startswith(u'var '): + stmt = stmt[len(u'var '):] + ass_m = re.match(r'^(?P[a-z]+)(?:\[(?P[^\]]+)\])?' + + r'=(?P.*)$', stmt) + if ass_m: + if ass_m.groupdict().get('index'): + def assign(val): + lvar = local_vars[ass_m.group('out')] + idx = interpret_expression(ass_m.group('index'), + local_vars, allow_recursion) + assert isinstance(idx, int) + lvar[idx] = val + return val + expr = ass_m.group('expr') + else: + def assign(val): + local_vars[ass_m.group('out')] = val + return val + expr = ass_m.group('expr') + elif stmt.startswith(u'return '): + assign = lambda v: v + expr = stmt[len(u'return '):] + else: + raise ExtractorError( + u'Cannot determine left side of statement in %r' % stmt) + + v = interpret_expression(expr, local_vars, allow_recursion) + return assign(v) + + def interpret_expression(expr, local_vars, allow_recursion): + if expr.isdigit(): + return int(expr) + + if expr.isalpha(): + return local_vars[expr] + + m = re.match(r'^(?P[a-z]+)\.(?P.*)$', expr) + if m: + member = m.group('member') + val = local_vars[m.group('in')] + if member == 'split("")': + return list(val) + if member == 'join("")': + return u''.join(val) + if member == 'length': + return len(val) + if member == 'reverse()': + return val[::-1] + slice_m = re.match(r'slice\((?P.*)\)', member) + if slice_m: + idx = interpret_expression( + slice_m.group('idx'), local_vars, allow_recursion-1) + return val[idx:] + + m = re.match( + r'^(?P[a-z]+)\[(?P.+)\]$', expr) + if m: + val = local_vars[m.group('in')] + idx = interpret_expression(m.group('idx'), local_vars, + allow_recursion-1) + return val[idx] + + m = re.match(r'^(?P.+?)(?P[%])(?P.+?)$', expr) + if m: + a = interpret_expression(m.group('a'), + local_vars, allow_recursion) + b = interpret_expression(m.group('b'), + local_vars, allow_recursion) + return a % b + + m = re.match( + r'^(?P[a-zA-Z]+)\((?P[a-z0-9,]+)\)$', expr) + if m: + fname = m.group('func') + if fname not in functions: + functions[fname] = extract_function(fname) + argvals = [int(v) if v.isdigit() else local_vars[v] + for v in m.group('args').split(',')] + return functions[fname](argvals) + raise ExtractorError(u'Unsupported JS expression %r' % expr) + + def extract_function(funcname): + func_m = re.search( + r'function ' + re.escape(funcname) + + r'\((?P[a-z,]+)\){(?P[^}]+)}', + jscode) + argnames = func_m.group('args').split(',') + + def resf(args): + local_vars = dict(zip(argnames, args)) + for stmt in func_m.group('code').split(';'): + res = interpret_statement(stmt, local_vars) + return res + return resf + + initial_function = 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('> 6 + tag_len = header16 & 0x3f + if tag_len == 0x3f: + tag_len = struct.unpack('> 4 + methods = {} + if kind in [0x00, 0x06]: # Slot or Const + u30() # Slot id + u30() # type_name_idx + vindex = u30() + if vindex != 0: + read_byte() # vkind + elif kind in [0x01, 0x02, 0x03]: # Method / Getter / Setter + u30() # disp_id + method_idx = u30() + methods[multinames[trait_name_idx]] = method_idx + elif kind == 0x04: # Class + u30() # slot_id + u30() # classi + elif kind == 0x05: # Function + 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 = u30() + for _c3 in range(metadata_count): + u30() # metadata index + + return methods + + # Classes + TARGET_CLASSNAME = u'SignatureDecipher' + searched_idx = multinames.index(TARGET_CLASSNAME) + searched_class_id = None + class_count = u30() + for class_id in range(class_count): + name_idx = u30() + if name_idx == searched_idx: + # We found the class we're looking for! + searched_class_id = class_id + u30() # super_name idx + flags = read_byte() + if flags & 0x08 != 0: # Protected namespace is present + u30() # protected_ns_idx + intrf_count = u30() + for _c2 in range(intrf_count): + u30() + u30() # iinit + trait_count = u30() + for _c2 in range(trait_count): + 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): + u30() # cinit + trait_count = u30() + for _c2 in range(trait_count): + 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 + script_count = u30() + for _c in range(script_count): + u30() # init + trait_count = u30() + for _c2 in range(trait_count): + parse_traits_info() + + # Method bodies + method_body_count = u30() + Method = collections.namedtuple('Method', ['code', 'local_count']) + methods = {} + for _c in range(method_body_count): + 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, local_count) + methods[method_idxs[method_idx]] = m + exception_count = u30() + for _c2 in range(exception_count): + u30() # from + u30() # to + u30() # target + u30() # exc_type + u30() # var_name + trait_count = u30() + for _c2 in range(trait_count): + parse_traits_info() + + 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] + 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) + 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)) + 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 + 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) + 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]) + + def _decrypt_signature(self, s, video_id, player_url, age_gate=False): """Turn the encrypted s field into a working signature""" + 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: + 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) + except Exception: + tb = traceback.format_exc() + self._downloader.report_warning( + u'Automatic signature extraction failed: ' + tb) + + 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) == 93: return s[86:29:-1] + s[88] + s[28:5:-1] elif len(s) == 92: @@ -431,15 +1077,15 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): elif len(s) == 87: return s[6:27] + s[4] + s[28:39] + s[27] + s[40:59] + s[2] + s[60:] elif len(s) == 86: - return s[80:72:-1] + s[16] + s[71:39:-1] + s[72] + s[38:16:-1] + s[82] + s[:16][::-1] + 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[81:64:-1] + s[82] + s[63:52:-1] + s[45] + s[51:45:-1] + s[1] + s[44:1:-1] + s[0] + return s[80:63:-1] + s[0] + s[62:0:-1] + s[63] elif len(s) == 82: - return s[80:73:-1] + s[81] + s[72:54:-1] + s[2] + s[53:43:-1] + s[0] + s[42:2:-1] + s[43] + s[1] + s[54] + 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: @@ -450,16 +1096,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): else: raise ExtractorError(u'Unable to decrypt signature, key length %d not supported; retrying might work' % (len(s))) - def _decrypt_signature_age_gate(self, s): - # 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] - else: - # Fallback to the other algortihms - return self._decrypt_signature(s) - - def _get_available_subtitles(self, video_id): + def _get_available_subtitles(self, video_id, webpage): try: sub_list = self._download_webpage( 'http://video.google.com/timedtext?hl=en&type=list&v=%s' % video_id, @@ -475,7 +1112,8 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): params = compat_urllib_parse.urlencode({ 'lang': lang, 'v': video_id, - 'fmt': self._downloader.params.get('subtitlesformat'), + 'fmt': self._downloader.params.get('subtitlesformat', 'srt'), + 'name': l[0].encode('utf-8'), }) url = u'http://www.youtube.com/api/timedtext?' + params sub_lang_list[lang] = url @@ -487,7 +1125,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): 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') + 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 @@ -509,7 +1147,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): list_page = self._download_webpage(list_url, video_id) caption_list = xml.etree.ElementTree.fromstring(list_page.encode('utf-8')) original_lang_node = caption_list.find('track') - if original_lang_node.attrib.get('kind') != 'asr' : + 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'] @@ -609,10 +1247,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): url_map[itag] = format_url return url_map - def _real_extract(self, url): - if re.match(r'(?:https?://)?[^/]+/watch\?feature=[a-z_]+$', url): - self._downloader.report_warning(u'Did you forget to quote the URL? Remember that & is a meta-character in most shells, so you want to put the URL in quotes, like youtube-dl \'http://www.youtube.com/watch?feature=foo&v=BaW_jenozKc\' (or simply youtube-dl BaW_jenozKc ).') + 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): # 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: @@ -631,7 +1270,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): video_webpage = video_webpage_bytes.decode('utf-8', 'ignore') # Attempt to extract SWF player URL - mobj = re.search(r'swfConfig.*?"(http:\\/\\/.*?watch.*?-.*?\.swf)"', video_webpage) + 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: @@ -645,7 +1284,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): # 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, - 'el': 'embedded', + 'el': 'player_embedded', 'gl': 'US', 'hl': 'en', 'eurl': 'https://youtube.googleapis.com/v/' + video_id, @@ -674,6 +1313,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): else: raise ExtractorError(u'"token" parameter not in video info for unknown reason') + 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') @@ -695,9 +1339,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): self._downloader.report_warning(u'unable to extract uploader nickname') # title - if 'title' not in video_info: - raise ExtractorError(u'Unable to extract video title') - video_title = compat_urllib_parse.unquote_plus(video_info['title'][0]) + if 'title' in video_info: + video_title = compat_urllib_parse.unquote_plus(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: @@ -707,7 +1353,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): 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 = '' + 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]) @@ -742,6 +1388,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): else: video_duration = 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: @@ -752,53 +1403,65 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): args = info['args'] # Easy way to know if the 's' value is in url_encoded_fmt_stream_map # this signatures are encrypted - m_s = re.search(r'[&,]s=', args['url_encoded_fmt_stream_map']) + 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.search(r'[&,]s=', args.get('adaptive_fmts', u'')) + m_s = re_signature.search(args.get('adaptive_fmts', u'')) if m_s is not None: - if 'url_encoded_fmt_stream_map' in video_info: - video_info['url_encoded_fmt_stream_map'][0] += ',' + args['adaptive_fmts'] + if 'adaptive_fmts' in video_info: + video_info['adaptive_fmts'][0] += ',' + args['adaptive_fmts'] else: - video_info['url_encoded_fmt_stream_map'] = [args['adaptive_fmts']] - elif 'adaptive_fmts' in video_info: - if 'url_encoded_fmt_stream_map' in video_info: - video_info['url_encoded_fmt_stream_map'][0] += ',' + video_info['adaptive_fmts'][0] - else: - video_info['url_encoded_fmt_stream_map'] = video_info['adaptive_fmts'] + video_info['adaptive_fmts'] = [args['adaptive_fmts']] except ValueError: pass if 'conn' in video_info and video_info['conn'][0].startswith('rtmp'): self.report_rtmp_download() video_url_list = [(None, video_info['conn'][0])] - elif 'url_encoded_fmt_stream_map' in video_info and len(video_info['url_encoded_fmt_stream_map']) >= 1: - if 'rtmpe%3Dyes' in video_info['url_encoded_fmt_stream_map'][0]: + 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 video_info['url_encoded_fmt_stream_map'][0].split(','): + 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'): - s = url_data['s'][0] if age_gate: - player = 'flash player' + 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: - player = u'html5 player %s' % self._search_regex(r'html5player-(.+?)\.js', video_webpage, + player_version = self._search_regex( + r'html5player-(.+?)\.js', video_webpage, 'html5 player', fatal=False) - parts_sizes = u'.'.join(compat_str(len(part)) for part in s.split('.')) + 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(s), parts_sizes, url_data['itag'][0], player)) - encrypted_sig = url_data['s'][0] - if age_gate: - signature = self._decrypt_signature_age_gate(encrypted_sig) - else: - signature = self._decrypt_signature(encrypted_sig) + (len(encrypted_sig), parts_sizes, url_data['itag'][0], player_desc)) + + if not age_gate: + jsplayer_url_json = self._search_regex( + r'"assets":.+?"js":\s*("[^"]+")', + video_webpage, u'JS player URL') + player_url = json.loads(jsplayer_url_json) + + signature = self._decrypt_signature( + encrypted_sig, video_id, player_url, age_gate) url += '&signature=' + signature if 'ratebypass' not in url: url += '&ratebypass=yes' @@ -814,16 +1477,16 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): return else: - raise ExtractorError(u'no conn or url_encoded_fmt_stream_map information found in video info') + raise ExtractorError(u'no conn, hlsvp or url_encoded_fmt_stream_map information found in video info') results = [] - for format_param, video_real_url in video_url_list: + for itag, video_real_url in video_url_list: # Extension - video_extension = self._video_extensions.get(format_param, 'flv') + video_extension = self._video_extensions.get(itag, 'flv') - video_format = '{0} - {1}{2}'.format(format_param if format_param else video_extension, - self._video_dimensions.get(format_param, '???'), - ' ('+self._special_itags[format_param]+')' if format_param in self._special_itags else '') + video_format = '{0} - {1}{2}'.format(itag if itag else video_extension, + self._video_dimensions.get(itag, '???'), + ' ('+self._special_itags[itag]+')' if itag in self._special_itags else '') results.append({ 'id': video_id, @@ -834,15 +1497,20 @@ class YoutubeIE(YoutubeBaseInfoExtractor, SubtitlesInfoExtractor): 'title': video_title, 'ext': video_extension, 'format': video_format, + 'format_id': itag, 'thumbnail': video_thumbnail, 'description': video_description, 'player_url': player_url, 'subtitles': video_subtitles, - 'duration': video_duration + 'duration': video_duration, + 'age_limit': 18 if age_gate else 0, + 'annotations': video_annotations, + 'webpage_url': 'https://www.youtube.com/watch?v=%s' % video_id, + 'view_count': view_count, }) return results -class YoutubePlaylistIE(InfoExtractor): +class YoutubePlaylistIE(YoutubeBaseInfoExtractor): IE_DESC = u'YouTube.com playlists' _VALID_URL = r"""(?: (?:https?://)? @@ -858,8 +1526,9 @@ class YoutubePlaylistIE(InfoExtractor): | ((?:PL|EC|UU|FL)[0-9A-Za-z-_]{10,}) )""" - _TEMPLATE_URL = 'https://gdata.youtube.com/feeds/api/playlists/%s?max-results=%i&start-index=%i&v=2&alt=json&safeSearch=none' - _MAX_RESULTS = 50 + _TEMPLATE_URL = 'https://www.youtube.com/playlist?list=%s&page=%s' + _MORE_PAGES_INDICATOR = r'data-link-type="next"' + _VIDEO_RE = r'href="/watch\?v=([0-9A-Za-z_-]{11})&' IE_NAME = u'youtube:playlist' @classmethod @@ -867,54 +1536,48 @@ class YoutubePlaylistIE(InfoExtractor): """Receives a URL and returns True if suitable for this IE.""" return re.match(cls._VALID_URL, url, re.VERBOSE) is not None + def _real_initialize(self): + self._login() + def _real_extract(self, url): # Extract playlist id mobj = re.match(self._VALID_URL, url, re.VERBOSE) if mobj is None: raise ExtractorError(u'Invalid URL: %s' % url) - - # Download playlist videos from API playlist_id = mobj.group(1) or mobj.group(2) - videos = [] + + # 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('https://www.youtube.com/watch?v=' + video_id, 'Youtube') + else: + self.to_screen(u'Downloading playlist PL%s - add --no-playlist to just download video %s' % (playlist_id, video_id)) + + # Extract the video ids from the playlist pages + ids = [] for page_num in itertools.count(1): - start_index = self._MAX_RESULTS * (page_num - 1) + 1 - if start_index >= 1000: - self._downloader.report_warning(u'Max number of results reached') - break - url = self._TEMPLATE_URL % (playlist_id, self._MAX_RESULTS, start_index) + url = self._TEMPLATE_URL % (playlist_id, page_num) page = self._download_webpage(url, playlist_id, u'Downloading page #%s' % page_num) + # The ids are duplicated + new_ids = orderedSet(re.findall(self._VIDEO_RE, page)) + ids.extend(new_ids) - try: - response = json.loads(page) - except ValueError as err: - raise ExtractorError(u'Invalid JSON in API response: ' + compat_str(err)) - - if 'feed' not in response: - raise ExtractorError(u'Got a malformed response from YouTube API') - playlist_title = response['feed']['title']['$t'] - if 'entry' not in response['feed']: - # Number of videos is a multiple of self._MAX_RESULTS + if re.search(self._MORE_PAGES_INDICATOR, page) is None: break - for entry in response['feed']['entry']: - index = entry['yt$position']['$t'] - if 'media$group' in entry and 'yt$videoid' in entry['media$group']: - videos.append(( - index, - 'https://www.youtube.com/watch?v=' + entry['media$group']['yt$videoid']['$t'] - )) - - videos = [v[1] for v in sorted(videos)] + playlist_title = self._og_search_title(page) - url_results = [self.url_result(vurl, 'Youtube') for vurl in videos] - return [self.playlist_result(url_results, playlist_id, playlist_title)] + url_results = [self.url_result(vid, 'Youtube') for vid in ids] + return self.playlist_result(url_results, playlist_id, playlist_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_-]+)" - _TEMPLATE_URL = 'http://www.youtube.com/channel/%s/videos?sort=da&flow=list&view=0&page=%s&gl=US&hl=en' _MORE_PAGES_INDICATOR = 'yt-uix-load-more' _MORE_PAGES_URL = 'http://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' @@ -935,29 +1598,30 @@ class YoutubeChannelIE(InfoExtractor): # Download channel page channel_id = mobj.group(1) video_ids = [] - pagenum = 1 - - url = self._TEMPLATE_URL % (channel_id, pagenum) - page = self._download_webpage(url, channel_id, - u'Downloading page #%s' % pagenum) - - # Extract video identifiers - ids_in_page = self.extract_videos_from_page(page) - video_ids.extend(ids_in_page) + url = 'https://www.youtube.com/channel/%s/videos' % channel_id + channel_page = self._download_webpage(url, channel_id) + if re.search(r'channel-header-autogenerated-label', channel_page) is not None: + autogenerated = True + else: + autogenerated = False - # Download any subsequent channel pages using the json-based channel_ajax query - if self._MORE_PAGES_INDICATOR in page: + 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_webpage(url, channel_id, u'Downloading page #%s' % pagenum) - + page = json.loads(page) - + 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']: + + 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))) @@ -969,7 +1633,7 @@ class YoutubeChannelIE(InfoExtractor): class YoutubeUserIE(InfoExtractor): IE_DESC = u'YouTube.com user videos (URL or "ytuser" keyword)' - _VALID_URL = r'(?:(?:(?:https?://)?(?:\w+\.)?youtube\.com/(?:user/)?)|ytuser:)(?!feed/)([A-Za-z0-9_-]+)' + _VALID_URL = r'(?:(?:(?:https?://)?(?:\w+\.)?youtube\.com/(?:user/)?(?!(?:attribution_link|watch)(?:$|[^a-z_A-Z0-9-])))|ytuser:)(?!feed/)([A-Za-z0-9_-]+)' _TEMPLATE_URL = 'http://gdata.youtube.com/feeds/api/users/%s' _GDATA_PAGE_SIZE = 50 _GDATA_URL = 'http://gdata.youtube.com/feeds/api/users/%s/uploads?max-results=%d&start-index=%d&alt=json' @@ -1074,6 +1738,10 @@ class YoutubeSearchIE(SearchInfoExtractor): videos = [self.url_result('http://www.youtube.com/watch?v=%s' % id, 'Youtube') for id in video_ids] return self.playlist_result(videos, query) +class YoutubeSearchDateIE(YoutubeSearchIE): + _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 YoutubeShowIE(InfoExtractor): IE_DESC = u'YouTube.com (multi-season) shows' @@ -1162,3 +1830,18 @@ class YoutubeFavouritesIE(YoutubeBaseInfoExtractor): 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'(?:https?://)?[^/]+/watch\?feature=[a-z_]+$' + + 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)