app/common/innertube.py

   1 # functions that deal with parsing data from youtube's internal API ("innertube")
   2
   3 from urllib.parse import parse_qs, urlparse
   4 import re
   5
   6 def findall(obj, key):
   7     """
   8     given a list of dicts, where one dict contains a given key, return said key.
   9     """
  10     if obj is None: return []
  11     return [ obj[key] for obj in obj if key in obj.keys() ]
  12 def listget(obj, index, fallback=None):
  13     if obj is None: return fallback
  14     return next(iter(obj[index:]), fallback)
  15 flatten = lambda l: [item for sublist in l for item in sublist] # https://stackoverflow.com/a/952952
  16 first = lambda l: next(iter(l),{})
  17 listfind = lambda obj,key: first(findall(obj,key))
  18
  19 def prepare_searchresults(yt_results):
  20     contents = listfind(yt_results, 'response') \
  21         .get('contents',{})\
  22         .get('twoColumnSearchResultsRenderer',{})\
  23         .get('primaryContents',{})\
  24         .get('sectionListRenderer',{})\
  25         .get('contents',[])
  26     contents = flatten([c.get('contents',[]) for c in findall(contents, 'itemSectionRenderer')])
  27
  28     return parse_result_items(contents)
  29
  30 def prepare_infocards(metadata):
  31     cards = metadata.get('cards',{}).get('cardCollectionRenderer',{}).get('cards',[])
  32     return list(filter(None, map(parse_infocard, cards)))
  33
  34 def prepare_endcards(metadata):
  35     endsc = metadata.get('endscreen',{}).get('endscreenRenderer',{}).get('elements',[])
  36     return list(filter(None, map(parse_endcard, endsc)))
  37
  38 def prepare_channel(result, channel_id):
  39     response = listfind(result,'response')
  40
  41     if 'alerts' in response: # possibly got an error back
  42         from flask import current_app
  43         current_app.logger.error([(alert['alertRenderer']['type'],alert['alertRenderer']['text']['simpleText']) for alert in response['alerts']])
  44         return None,None,[],[],False
  45
  46     meta1 = response.get('metadata',{}).get('channelMetadataRenderer',{})
  47     meta2 = response.get('microformat',{}).get('microformatDataRenderer',{})
  48     title = meta1.get('title', meta2.get('title'))
  49     descr = meta1.get('description', meta2.get('description')) # meta2.description is capped at 160chars
  50     thumb = mkthumbs(meta2.get('thumbnail',meta1.get('avatar',{})).get('thumbnails',{})) # .avatar ~ 900px
  51
  52     contents = response.get('continuationContents')
  53     if not contents: # overran end of list
  54         return title, descr, thumb, [], False
  55
  56     unparsed = contents.get('gridContinuation',{}).get('items') or \
  57             contents.get('sectionListContinuation',{}).get('contents') or []
  58     items, extra = parse_channel_items(unparsed, channel_id, title)
  59     has_more = 'continuations' in (contents.get('gridContinuation') or
  60             contents.get('sectionListContinuation') or {})
  61
  62     return title, descr, thumb, items, has_more
  63
  64 def prepare_playlist(result):
  65     contents = listfind(result,'response')['continuationContents']
  66     unparsed = contents['playlistVideoListContinuation'].get('contents',[])
  67     has_more = 'continuations' in contents.get('playlistVideoListContinuation')
  68
  69     return list(filter(None, map(parse_playlist, unparsed))), has_more
  70
  71 def mkthumbs(thumbs):
  72     output = {str(e['height']): e['url'] for e in thumbs}
  73     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
  74     return {**output, 'largest': largest}
  75
  76 def clean_url(url):
  77     # externals URLs are redirected through youtube.com/redirect, but we
  78     # may encounter internal URLs, too
  79     return parse_qs(urlparse(url).query).get('q',[url])[0]
  80
  81 def toInt(s, fallback=0):
  82     if s is None:
  83         return fallback
  84     try:
  85         return int(''.join(filter(str.isdigit, s)))
  86     except ValueError:
  87         return fallback
  88
  89 # Remove left-/rightmost word from string:
  90 delL = lambda s: s.partition(' ')[2]
  91
  92 def age(s):
  93     if s is None: # missing from autogen'd music, some livestreams
  94         return None
  95     # Some livestreams have "Streamed 7 hours ago"
  96     s = s.replace("Streamed ","")
  97     # Now, everything should be in the form "1 year ago"
  98     value, unit, _ = s.split(" ")
  99     suffix = dict(
 100         month='mn',
 101         months='mn',
 102     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 103
 104     return f"{value}{suffix}"
 105
 106 def log_unknown_card(data):
 107     import json
 108     try:
 109         from flask import request
 110         source = request.url
 111     except: source = "unknown"
 112     with open("/tmp/innertube.err", "a") as f:
 113         f.write(f"\n/***** {source} *****/\n")
 114         json.dump(data, f, indent=2)
 115
 116 def parse_result_items(items):
 117     # TODO: use .get() for most non-essential attributes
 118     """
 119     parses youtube search response into an easier to use format.
 120     """
 121     results = []
 122     extras = []
 123     for item in items:
 124         key = next(iter(item.keys()), None)
 125         content = item[key]
 126         if key == 'videoRenderer':
 127             is_live = listfind(content.get('badges',[]), 'metadataBadgeRenderer').get('style') == 'BADGE_STYLE_TYPE_LIVE_NOW'
 128             results.append({'type': 'VIDEO', 'content': {
 129                 'video_id': content['videoId'],
 130                 'title': content['title']['runs'][0]['text'],
 131                 'author': content['longBylineText']['runs'][0]['text'] or \
 132                           content['shortBylineText']['runs'][0]['text'],
 133                 'channel_id': content['ownerText']['runs'][0] \
 134                     ['navigationEndpoint']['browseEndpoint']['browseId'],
 135                 'length': content.get('lengthText',{}).get('simpleText') \
 136                     if not is_live else 'LIVE', # "44:07", "1:41:50"
 137                 'views': toInt(content.get('viewCountText',{}).get('simpleText') or # "123,456 views", ...
 138                     listget(content.get('viewCountText',{}).get('runs',[]),0,{}).get('text')) or 0, # ... "1,234 watching", absent on 0 views
 139                 'published': age(content.get('publishedTimeText',{}).get('simpleText')),
 140             }})
 141         elif key == 'playlistRenderer':
 142             results.append({'type': 'PLAYLIST', 'content': {
 143                 'playlist_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('playlistId') or \
 144                     content.get('playlistId'), # COURSE/"learning playlist"
 145                 'video_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('videoId') or \
 146                     videoid_from_thumbnail(content), # learning playlist
 147                 'title': content['title']['simpleText'],
 148                 # Note: learning playlists have no author/channel_id
 149                 'author': listget(content.get('longBylineText',{}).get('runs',[]),0,{}).get('text') or
 150                      listget(content.get('shortBylineText',{}).get('runs',[]),0,{}).get('text'),
 151                 'channel_id': listget(content.get('longBylineText',{}).get('runs',[]),0,{}) \
 152                     .get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'), # OR .shortBylineText
 153                 'n_videos': toInt(content['videoCount']),
 154             }})
 155         elif key == 'radioRenderer': # "Mix" playlists
 156             results.append({'type': 'PLAYLIST', 'content': {
 157                 'playlist_id': content['playlistId'],
 158                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 159                 'title': content['title']['simpleText'],
 160                 'author': content['longBylineText']['simpleText'] or \
 161                     content['shortBylineText']['simpleText'] , # always "YouTube"
 162                 'channel_id': None,
 163                 'n_videos': content['videoCountShortText']['runs'][0]['text'] or \
 164                     content['videoCountText']['runs'][0]['text'],
 165                     # videoCountShortText: "50+"; videoCountText: "50+ videos"
 166             }})
 167         elif key == 'channelRenderer':
 168             results.append({'type': 'CHANNEL', 'content': {
 169                 'channel_id': content['channelId'],
 170                 'title': content['title']['simpleText'],
 171                 'icons': mkthumbs(content['thumbnail']['thumbnails']),
 172                 'subscribers': content.get('subscriberCountText',{}).get('simpleText'), # "2.47K subscribers"
 173             }})
 174         elif key == 'shelfRenderer':
 175             subkey = next(iter(content['content'].keys()), {}) #verticalListRenderer/horizontalMovieListRenderer
 176             r, e = parse_result_items(content['content'][subkey]['items'])
 177             results.extend(r)
 178             extras.extend(e)
 179         elif key == 'movieRenderer': # movies to buy/rent
 180             pass
 181         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer']: # haha, no.
 182             pass
 183         elif key == 'horizontalCardListRenderer':
 184             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 185             pass
 186         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 187             pass
 188         elif key == 'clarificationRenderer': # COVID-19 infos
 189             pass
 190         elif key == 'didYouMeanRenderer' or key == 'showingResultsForRenderer':
 191             extras.append({
 192                 'type': 'spelling',
 193                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 194                 'autocorrected': key == 'showingResultsForRenderer',
 195             })
 196         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 197             extras.append({
 198                 'type': content['icon']['iconType'],
 199                 'message': content['title']['runs'][0]['text'],
 200             })
 201         else:
 202             log_unknown_card(item)
 203     return results, extras
 204
 205 def parse_infocard(card):
 206     """
 207     parses a single infocard into a format that's easier to handle.
 208     """
 209     card = card['cardRenderer']
 210     ctype = list(card['content'].keys())[0]
 211     content = card['content'][ctype]
 212     if ctype == "pollRenderer":
 213         return {'type': "POLL", 'content': {
 214             'question': content['question']['simpleText'],
 215             'answers': [(a['text']['simpleText'],a['numVotes']) \
 216                 for a in content['choices']],
 217         }}
 218     elif ctype == "videoInfoCardContentRenderer":
 219         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 220         return {'type': "VIDEO", 'content': {
 221             'video_id': content['action']['watchEndpoint']['videoId'],
 222             'title': content['videoTitle']['simpleText'],
 223             'author': delL(content['channelName']['simpleText']),
 224             'length': content.get('lengthString',{}).get('simpleText') \
 225                 if not is_live else "LIVE", # "23:03"
 226             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 227                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 228         }}
 229     elif ctype == "playlistInfoCardContentRenderer":
 230         return {'type': "PLAYLIST", 'content': {
 231             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 232             'video_id': content['action']['watchEndpoint']['videoId'],
 233             'title': content['playlistTitle']['simpleText'],
 234             'author': delL(content['channelName']['simpleText']),
 235             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 236         }}
 237     elif ctype == "simpleCardContentRenderer" and \
 238             'urlEndpoint' in content['command']:
 239         return {'type': "WEBSITE", 'content': {
 240             'url': clean_url(content['command']['urlEndpoint']['url']),
 241             'domain': content['displayDomain']['simpleText'],
 242             'title': content['title']['simpleText'],
 243             # XXX: no thumbnails for infocards
 244         }}
 245     elif ctype == "collaboratorInfoCardContentRenderer":
 246         return {'type': "CHANNEL", 'content': {
 247             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 248             'title': content['channelName']['simpleText'],
 249             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 250             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 251         }}
 252     else:
 253         log_unknown_card(card)
 254         return None
 255
 256 def parse_endcard(card):
 257     """
 258     parses a single endcard into a format that's easier to handle.
 259     """
 260     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 261     ctype = card['style']
 262     if ctype == "CHANNEL":
 263         return {'type': ctype, 'content': {
 264             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 265             'title': card['title']['simpleText'],
 266             'icons': mkthumbs(card['image']['thumbnails']),
 267         }}
 268     elif ctype == "VIDEO":
 269         return {'type': ctype, 'content': {
 270             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 271             'title': card['title']['simpleText'],
 272             'length': card['videoDuration']['simpleText'],  # '12:21'
 273             'views': toInt(card['metadata']['simpleText']),
 274             # XXX: no channel name
 275         }}
 276     elif ctype == "PLAYLIST":
 277         return {'type': ctype, 'content': {
 278             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 279             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 280             'title': card['title']['simpleText'],
 281             'author': delL(card['metadata']['simpleText']),
 282             'n_videos': toInt(card['playlistLength']['simpleText']),
 283         }}
 284     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 285         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 286         return {'type': "WEBSITE", 'content': {
 287             'url': url,
 288             'domain': urlparse(url).netloc,
 289             'title': card['title']['simpleText'],
 290             'icons': mkthumbs(card['image']['thumbnails']),
 291         }}
 292     else:
 293         log_unknown_card(card)
 294         return None
 295
 296 def videoid_from_thumbnail(content):
 297     # learning playlist; example: PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5 (/user/enyay/playlists)
 298     return re.match(r"https?://i.ytimg.com/vi/([-_0-9a-zA-Z]{11})|()",
 299         listget(listget(content.get('thumbnails',[]),0,{}).get('thumbnails',[]),0,{}).get('url','')
 300     ).group(1)
 301
 302 def parse_channel_items(items, channel_id, author):
 303     result = []
 304     extra = []
 305     for item in items:
 306         key = next(iter(item.keys()), None)
 307         content = item[key]
 308         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer"]:
 309             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 310             result.append({'type': 'VIDEO', 'content': {
 311                 'video_id': content['videoId'],
 312                 'title': content['title'].get('simpleText') or content['title'].get('runs',[{}])[0].get('text'),
 313                 'author': content.get('bylineText',{}).get('runs',[{}])[0].get('text') or author,
 314                 'channel_id': content.get('bylineText',{}).get('runs',[{}])[0] \
 315                     .get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId') or channel_id,
 316                 'length': (content.get('lengthText',{}).get('simpleText') or # topic channel
 317                     listfind(content.get('thumbnailOverlays',[]),'thumbnailOverlayTimeStatusRenderer')
 318                         .get('text',{}).get('simpleText')),
 319                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 320                 'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 321                 'published': age(content.get('publishedTimeText',{}).get('simpleText')),
 322             }})
 323         elif key == "gridPlaylistRenderer" or key == "playlistRenderer":
 324             result.append({'type': 'PLAYLIST', 'content': {
 325                 'playlist_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('playlistId') or content.get('playlistId'),
 326                 'video_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('videoId',{}) or videoid_from_thumbnail(content),
 327                 'title': (content['title'].get('simpleText') or # playlistRenderer
 328                     content['title']['runs'][0]['text']), # gridPlaylistRenderer
 329                 'author': author,
 330                 'channel_id': channel_id,
 331                 'n_videos': toInt(content.get('videoCount') or # playlistRenderer
 332                     content.get('videoCountShortText',{}).get('simpleText') or # grid(1)
 333                     content.get('videoCountText',{}).get('runs',[{}])[0].get('text')), # grid(2)
 334             }})
 335         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer"]:
 336                 newkey = {
 337                     "itemSectionRenderer": 'contents',
 338                     "gridRenderer": 'items',
 339                     "horizontalCardListRenderer": 'cards',
 340                 }.get(key)
 341                 r, e = parse_channel_items(content[newkey], channel_id, author)
 342                 result.extend(r)
 343                 extra.extend(e)
 344         elif key == "shelfRenderer":
 345                 r, e = parse_channel_items([content['content']], channel_id, author)
 346                 result.extend(r)
 347                 extra.extend(e)
 348         elif key == "messageRenderer":
 349             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 350             pass
 351         elif key == "gameCardRenderer":
 352             pass
 353         else:
 354             log_unknown_card(item)
 355
 356     return result, extra
 357
 358 def parse_playlist(item):
 359     key = next(iter(item.keys()), None)
 360     content = item[key]
 361     if key == "playlistVideoRenderer":
 362         if not content.get('isPlayable', False):
 363             return None # private or deleted video
 364
 365         return {'type': 'VIDEO', 'content': {
 366             'video_id': content['videoId'],
 367             'title': (content['title'].get('simpleText') or # playable videos
 368                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 369             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 370             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 371             # rest is missing from unplayable videos:
 372             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 373             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 374             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 375                 int(content.get("lengthSeconds", 0))), # "531"
 376             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 377         }}
 378     else:
 379         raise Exception(item) # XXX TODO