X-Git-Url: http://git.cielonegro.org/gitweb.cgi?a=blobdiff_plain;f=youtube_dl%2FInfoExtractors.py;h=3f7950cb834c984d97c9c4c1c2899930f9a0a682;hb=1447f728b57601b109ac1c3ff913e9c0d5238c9c;hp=490f30491a5c1f878a58ea9e337040c927ab28ae;hpb=89de9eb125abd273d50c9c5cc2d2ce80fa8922f3;p=youtube-dl.git diff --git a/youtube_dl/InfoExtractors.py b/youtube_dl/InfoExtractors.py index 490f30491..3f7950cb8 100755 --- a/youtube_dl/InfoExtractors.py +++ b/youtube_dl/InfoExtractors.py @@ -48,7 +48,7 @@ class InfoExtractor(object): uploader_id: Nickname or id of the video uploader. location: Physical location of the video. player_url: SWF Player URL (used for rtmpdump). - subtitles: The .srt file contents. + subtitles: The subtitle file contents. urlhandle: [internal] The urlHandle to be used to download the file, like returned by urllib.request.urlopen @@ -126,8 +126,14 @@ class InfoExtractor(object): def _download_webpage(self, url_or_request, video_id, note=None, errnote=None): """ Returns the data of the page as a string """ urlh = self._request_webpage(url_or_request, video_id, note, errnote) + content_type = urlh.headers.get('Content-Type', '') + m = re.match(r'[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+\s*;\s*charset=(.+)', content_type) + if m: + encoding = m.group(1) + else: + encoding = 'utf-8' webpage_bytes = urlh.read() - return webpage_bytes.decode('utf-8', 'replace') + return webpage_bytes.decode(encoding, 'replace') class YoutubeIE(InfoExtractor): @@ -218,7 +224,16 @@ class YoutubeIE(InfoExtractor): def report_video_subtitles_download(self, video_id): """Report attempt to download video info webpage.""" - self._downloader.to_screen(u'[youtube] %s: Downloading video subtitles' % video_id) + self._downloader.to_screen(u'[youtube] %s: Checking available subtitles' % video_id) + + def report_video_subtitles_request(self, video_id, sub_lang, format): + """Report attempt to download video info webpage.""" + self._downloader.to_screen(u'[youtube] %s: Downloading video subtitles for %s.%s' % (video_id, sub_lang, format)) + + def report_video_subtitles_available(self, video_id, sub_lang_list): + """Report available subtitles.""" + sub_lang = ",".join(list(sub_lang_list.keys())) + self._downloader.to_screen(u'[youtube] %s: Available subtitles for video: %s' % (video_id, sub_lang)) def report_information_extraction(self, video_id): """Report attempt to extract video information.""" @@ -232,55 +247,63 @@ class YoutubeIE(InfoExtractor): """Indicate the download will use the RTMP protocol.""" self._downloader.to_screen(u'[youtube] RTMP download detected') - def _closed_captions_xml_to_srt(self, xml_string): - srt = '' - texts = re.findall(r'([^<]+)', xml_string, re.MULTILINE) - # TODO parse xml instead of regex - for n, (start, dur_tag, dur, caption) in enumerate(texts): - if not dur: dur = '4' - start = float(start) - end = start + float(dur) - start = "%02i:%02i:%02i,%03i" %(start/(60*60), start/60%60, start%60, start%1*1000) - end = "%02i:%02i:%02i,%03i" %(end/(60*60), end/60%60, end%60, end%1*1000) - caption = unescapeHTML(caption) - caption = unescapeHTML(caption) # double cycle, intentional - srt += str(n+1) + '\n' - srt += start + ' --> ' + end + '\n' - srt += caption + '\n\n' - return srt - - def _extract_subtitles(self, video_id): + def _get_available_subtitles(self, video_id): self.report_video_subtitles_download(video_id) request = compat_urllib_request.Request('http://video.google.com/timedtext?hl=en&type=list&v=%s' % video_id) try: - srt_list = compat_urllib_request.urlopen(request).read().decode('utf-8') + sub_list = compat_urllib_request.urlopen(request).read().decode('utf-8') except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: return (u'WARNING: unable to download video subtitles: %s' % compat_str(err), None) - srt_lang_list = re.findall(r'name="([^"]*)"[^>]+lang_code="([\w\-]+)"', srt_list) - srt_lang_list = dict((l[1], l[0]) for l in srt_lang_list) - if not srt_lang_list: - return (u'WARNING: video has no closed captions', None) - if self._downloader.params.get('subtitleslang', False): - srt_lang = self._downloader.params.get('subtitleslang') - elif 'en' in srt_lang_list: - srt_lang = 'en' - else: - srt_lang = list(srt_lang_list.keys())[0] - if not srt_lang in srt_lang_list: - return (u'WARNING: no closed captions found in the specified language', None) + sub_lang_list = re.findall(r'name="([^"]*)"[^>]+lang_code="([\w\-]+)"', sub_list) + sub_lang_list = dict((l[1], l[0]) for l in sub_lang_list) + if not sub_lang_list: + return (u'WARNING: video doesn\'t have subtitles', None) + return sub_lang_list + + def _list_available_subtitles(self, video_id): + sub_lang_list = self._get_available_subtitles(video_id) + self.report_video_subtitles_available(video_id, sub_lang_list) + + def _request_subtitle(self, sub_lang, sub_name, video_id, format): + self.report_video_subtitles_request(video_id, sub_lang, format) params = compat_urllib_parse.urlencode({ - 'lang': srt_lang, - 'name': srt_lang_list[srt_lang].encode('utf-8'), + 'lang': sub_lang, + 'name': sub_name, 'v': video_id, + 'fmt': format, }) url = 'http://www.youtube.com/api/timedtext?' + params try: - srt_xml = compat_urllib_request.urlopen(url).read().decode('utf-8') + sub = compat_urllib_request.urlopen(url).read().decode('utf-8') except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: return (u'WARNING: unable to download video subtitles: %s' % compat_str(err), None) - if not srt_xml: + if not sub: return (u'WARNING: Did not fetch video subtitles', None) - return (None, self._closed_captions_xml_to_srt(srt_xml)) + return (None, sub_lang, sub) + + def _extract_subtitle(self, video_id): + sub_lang_list = self._get_available_subtitles(video_id) + sub_format = self._downloader.params.get('subtitlesformat') + if self._downloader.params.get('subtitleslang', False): + sub_lang = self._downloader.params.get('subtitleslang') + elif 'en' in sub_lang_list: + sub_lang = 'en' + else: + sub_lang = list(sub_lang_list.keys())[0] + if not sub_lang in sub_lang_list: + return (u'WARNING: no closed captions found in the specified language "%s"' % sub_lang, None) + + subtitle = self._request_subtitle(sub_lang, sub_lang_list[sub_lang].encode('utf-8'), video_id, sub_format) + return [subtitle] + + def _extract_all_subtitles(self, video_id): + sub_lang_list = self._get_available_subtitles(video_id) + sub_format = self._downloader.params.get('subtitlesformat') + subtitles = [] + for sub_lang in sub_lang_list: + subtitle = self._request_subtitle(sub_lang, sub_lang_list[sub_lang].encode('utf-8'), video_id, sub_format) + subtitles.append(subtitle) + return subtitles def _print_formats(self, formats): print('Available formats:') @@ -308,7 +331,7 @@ class YoutubeIE(InfoExtractor): else: raise netrc.NetrcParseError('No authenticators for %s' % self._NETRC_MACHINE) except (IOError, netrc.NetrcParseError) as err: - self._downloader.to_stderr(u'WARNING: parsing .netrc: %s' % compat_str(err)) + self._downloader.report_warning(u'parsing .netrc: %s' % compat_str(err)) return # Set language @@ -317,7 +340,7 @@ class YoutubeIE(InfoExtractor): self.report_lang() compat_urllib_request.urlopen(request).read() except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: - self._downloader.to_stderr(u'WARNING: unable to set language: %s' % compat_str(err)) + self._downloader.report_warning(u'unable to set language: %s' % compat_str(err)) return # No authentication to be performed @@ -328,7 +351,7 @@ class YoutubeIE(InfoExtractor): try: login_page = compat_urllib_request.urlopen(request).read().decode('utf-8') except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: - self._downloader.to_stderr(u'WARNING: unable to fetch login page: %s' % compat_str(err)) + self._downloader.report_warning(u'unable to fetch login page: %s' % compat_str(err)) return galx = None @@ -372,10 +395,10 @@ class YoutubeIE(InfoExtractor): self.report_login() login_results = compat_urllib_request.urlopen(request).read().decode('utf-8') if re.search(r'(?i)]* id="gaia_loginform"', login_results) is not None: - self._downloader.to_stderr(u'WARNING: unable to log in: bad username or password') + self._downloader.report_warning(u'unable to log in: bad username or password') return except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: - self._downloader.to_stderr(u'WARNING: unable to log in: %s' % compat_str(err)) + self._downloader.report_warning(u'unable to log in: %s' % compat_str(err)) return # Confirm age @@ -501,12 +524,26 @@ class YoutubeIE(InfoExtractor): else: video_description = '' - # closed captions + # subtitles video_subtitles = None + if self._downloader.params.get('writesubtitles', False): - (srt_error, video_subtitles) = self._extract_subtitles(video_id) - if srt_error: - self._downloader.trouble(srt_error) + video_subtitles = self._extract_subtitle(video_id) + if video_subtitles: + (sub_error, sub_lang, sub) = video_subtitles[0] + if sub_error: + self._downloader.trouble(sub_error) + + if self._downloader.params.get('allsubtitles', False): + video_subtitles = self._extract_all_subtitles(video_id) + for video_subtitle in video_subtitles: + (sub_error, sub_lang, sub) = video_subtitle + if sub_error: + self._downloader.trouble(sub_error) + + if self._downloader.params.get('listsubtitles', False): + sub_lang_list = self._list_available_subtitles(video_id) + return if 'length_seconds' not in video_info: self._downloader.trouble(u'WARNING: unable to extract video duration') @@ -1281,7 +1318,8 @@ class GenericIE(InfoExtractor): def report_download_webpage(self, video_id): """Report webpage download.""" - self._downloader.to_screen(u'WARNING: Falling back on generic information extractor.') + if not self._downloader.params.get('test', False): + self._downloader.to_screen(u'WARNING: Falling back on generic information extractor.') self._downloader.to_screen(u'[generic] %s: Downloading webpage' % video_id) def report_extraction(self, video_id): @@ -1351,13 +1389,8 @@ class GenericIE(InfoExtractor): if self._test_redirect(url): return video_id = url.split('/')[-1] - request = compat_urllib_request.Request(url) try: - self.report_download_webpage(video_id) - webpage = compat_urllib_request.urlopen(request).read() - except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: - self._downloader.trouble(u'ERROR: Unable to retrieve video webpage: %s' % compat_str(err)) - return + webpage = self._download_webpage(url, video_id) except ValueError as err: # since this is the last-resort InfoExtractor, if # this error is thrown, it'll be thrown here @@ -1456,7 +1489,7 @@ class YoutubeSearchIE(InfoExtractor): self._downloader.trouble(u'ERROR: invalid download number %s for query "%s"' % (n, query)) return elif n > self._max_youtube_results: - self._downloader.to_stderr(u'WARNING: ytsearch returns max %i results (you requested %i)' % (self._max_youtube_results, n)) + self._downloader.report_warning(u'ytsearch returns max %i results (you requested %i)' % (self._max_youtube_results, n)) n = self._max_youtube_results self._download_n_results(query, n) return @@ -1482,6 +1515,10 @@ class YoutubeSearchIE(InfoExtractor): return api_response = json.loads(data)['data'] + if not 'items' in api_response: + self._downloader.trouble(u'[youtube] No video results') + return + new_ids = list(video['id'] for video in api_response['items']) video_ids += new_ids @@ -1534,7 +1571,7 @@ class GoogleSearchIE(InfoExtractor): self._downloader.trouble(u'ERROR: invalid download number %s for query "%s"' % (n, query)) return elif n > self._max_google_results: - self._downloader.to_stderr(u'WARNING: gvsearch returns max %i results (you requested %i)' % (self._max_google_results, n)) + self._downloader.report_warning(u'gvsearch returns max %i results (you requested %i)' % (self._max_google_results, n)) n = self._max_google_results self._download_n_results(query, n) return @@ -1618,7 +1655,7 @@ class YahooSearchIE(InfoExtractor): self._downloader.trouble(u'ERROR: invalid download number %s for query "%s"' % (n, query)) return elif n > self._max_yahoo_results: - self._downloader.to_stderr(u'WARNING: yvsearch returns max %i results (you requested %i)' % (self._max_yahoo_results, n)) + self._downloader.report_warning(u'yvsearch returns max %i results (you requested %i)' % (self._max_yahoo_results, n)) n = self._max_yahoo_results self._download_n_results(query, n) return @@ -1737,8 +1774,7 @@ class YoutubePlaylistIE(InfoExtractor): break page_num += 1 - videos = map(operator.itemgetter(1), sorted(videos)) - + videos = [v[1] for v in sorted(videos)] total = len(videos) playliststart = self._downloader.params.get('playliststart', 1) - 1 @@ -2077,7 +2113,7 @@ class FacebookIE(InfoExtractor): else: raise netrc.NetrcParseError('No authenticators for %s' % self._NETRC_MACHINE) except (IOError, netrc.NetrcParseError) as err: - self._downloader.to_stderr(u'WARNING: parsing .netrc: %s' % compat_str(err)) + self._downloader.report_warning(u'parsing .netrc: %s' % compat_str(err)) return if useremail is None: @@ -2094,10 +2130,10 @@ class FacebookIE(InfoExtractor): self.report_login() login_results = compat_urllib_request.urlopen(request).read() if re.search(r'', login_results) is not None: - self._downloader.to_stderr(u'WARNING: unable to log in: bad username/password, or exceded login rate limit (~3/min). Check credentials or wait.') + self._downloader.report_warning(u'unable to log in: bad username/password, or exceded login rate limit (~3/min). Check credentials or wait.') return except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err: - self._downloader.to_stderr(u'WARNING: unable to log in: %s' % compat_str(err)) + self._downloader.report_warning(u'unable to log in: %s' % compat_str(err)) return def _real_extract(self, url): @@ -2162,6 +2198,17 @@ class BlipTVIE(InfoExtractor): self._downloader.trouble(u'ERROR: invalid URL: %s' % url) return + urlp = compat_urllib_parse_urlparse(url) + if urlp.path.startswith('/play/'): + request = compat_urllib_request.Request(url) + response = compat_urllib_request.urlopen(request) + redirecturl = response.geturl() + rurlp = compat_urllib_parse_urlparse(redirecturl) + file_id = compat_parse_qs(rurlp.fragment)['file'][0].rpartition('/')[2] + url = 'http://blip.tv/a/a-' + file_id + return self._real_extract(url) + + if '?' in url: cchar = '&' else: @@ -2543,7 +2590,7 @@ class EscapistIE(InfoExtractor): 'uploader': showName, 'upload_date': None, 'title': showName, - 'ext': 'flv', + 'ext': 'mp4', 'thumbnail': imgUrl, 'description': description, 'player_url': playerUrl, @@ -3577,55 +3624,6 @@ class FunnyOrDieIE(InfoExtractor): } return [info] -class TweetReelIE(InfoExtractor): - _VALID_URL = r'^(?:https?://)?(?:www\.)?tweetreel\.com/[?](?P[0-9a-z]+)$' - - def _real_extract(self, url): - mobj = re.match(self._VALID_URL, url) - if mobj is None: - self._downloader.trouble(u'ERROR: invalid URL: %s' % url) - return - - video_id = mobj.group('id') - webpage = self._download_webpage(url, video_id) - - m = re.search(r'
', webpage) - if not m: - self._downloader.trouble(u'ERROR: Cannot find status ID') - status_id = m.group(1) - - m = re.search(r'
(.*?)
', webpage, flags=re.DOTALL) - if not m: - self._downloader.trouble(u'WARNING: Cannot find description') - desc = unescapeHTML(re.sub('', '', m.group(1))).strip() - - m = re.search(r'
.*?from (?P.+?)', webpage, flags=re.DOTALL) - if not m: - self._downloader.trouble(u'ERROR: Cannot find uploader') - uploader = unescapeHTML(m.group('uploader')) - uploader_id = unescapeHTML(m.group('uploader_id')) - - m = re.search(r'video|app)/ #If the page is only for videos or for a game @@ -3764,7 +3762,7 @@ class YouPornIE(InfoExtractor): # Get the video date result = re.search(r'Date:(?P.*) ', webpage) if result is None: - self._downloader.to_stderr(u'WARNING: unable to extract video date') + self._downloader.report_warning(u'unable to extract video date') upload_date = None else: upload_date = result.group('date').strip() @@ -3772,7 +3770,7 @@ class YouPornIE(InfoExtractor): # Get the video uploader result = re.search(r'Submitted:(?P.*)', webpage) if result is None: - self._downloader.to_stderr(u'WARNING: unable to extract uploader') + self._downloader.report_warning(u'unable to extract uploader') video_uploader = None else: video_uploader = result.group('uploader').strip() @@ -3988,11 +3986,11 @@ class KeekIE(InfoExtractor): webpage = self._download_webpage(url, video_id) m = re.search(r'[\s\n]+

(?P\w+)

', webpage) - uploader = unescapeHTML(m.group('uploader')) + m = re.search(r'
[\S\s]+?

(?P.+?)

', webpage) + uploader = clean_html(m.group('uploader')) info = { - 'id':video_id, - 'url':video_url, + 'id': video_id, + 'url': video_url, 'ext': 'mp4', 'title': title, 'thumbnail': thumbnail, @@ -4129,6 +4127,40 @@ class MySpassIE(InfoExtractor): } return [info] +class SpiegelIE(InfoExtractor): + _VALID_URL = r'https?://(?:www\.)?spiegel\.de/video/[^/]*-(?P[0-9]+)(?:\.html)?(?:#.*)$' + + def _real_extract(self, url): + m = re.match(self._VALID_URL, url) + video_id = m.group('videoID') + + webpage = self._download_webpage(url, video_id) + m = re.search(r'
(.*?)
', webpage) + if not m: + raise ExtractorError(u'Cannot find title') + video_title = unescapeHTML(m.group(1)) + + xml_url = u'http://video2.spiegel.de/flash/' + video_id + u'.xml' + xml_code = self._download_webpage(xml_url, video_id, + note=u'Downloading XML', errnote=u'Failed to download XML') + + idoc = xml.etree.ElementTree.fromstring(xml_code) + last_type = idoc[-1] + filename = last_type.findall('./filename')[0].text + duration = float(last_type.findall('./duration')[0].text) + + video_url = 'http://video2.spiegel.de/flash/' + filename + video_ext = filename.rpartition('.')[2] + info = { + 'id': video_id, + 'url': video_url, + 'ext': video_ext, + 'title': video_title, + 'duration': duration, + } + return [info] + + def gen_extractors(): """ Return a list of an instance of every supported extractor. The order does matter; the first extractor matched is the one handling the URL. @@ -4170,7 +4202,6 @@ def gen_extractors(): NBAIE(), JustinTVIE(), FunnyOrDieIE(), - TweetReelIE(), SteamIE(), UstreamIE(), RBMARadioIE(), @@ -4178,6 +4209,7 @@ def gen_extractors(): KeekIE(), TEDIE(), MySpassIE(), + SpiegelIE(), GenericIE() ]