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 class G:
  20     """
  21     null-coalescing version of dict.get() that also works on lists.
  22
  23     the | operator is overloaded to achieve similar looking code to jq(1) filters.
  24     the first found key is used: dict(foo=1)|G('bar','foo') returns 1.
  25     """
  26     def __init__(self, *keys):
  27         self.keys = keys
  28     def __ror__(self, other):
  29         for key in self.keys:
  30             try:    return other[key]
  31             except: continue
  32         return None
  33     class _Text:
  34         """ parses youtube's .runs[].text and .simpleText variants """
  35         def __ror__(self, other): # Note: only returning runs[0], not concat'ing all!
  36             return other|G('simpleText') or other|G('runs')|G(0)|G('text')
  37     text = _Text()
  38 class Select:
  39     """ |Select('foo') returns the first foo in list, |Select(all='foo') returns all foos. """
  40     def __init__(self, key=None, *, all=None):
  41         self.key = key or all
  42         self.all = all
  43     def __ror__(self, other):
  44         try: items = [ other[self.key] for other in other if self.key in other.keys() ]
  45         except: items = []
  46         return items if self.all else items|G(0)
  47 class A:
  48     """ apply """
  49     def __init__(self, f, *args):
  50         self.f = f
  51         self.args = args
  52     def __ror__(self, other):
  53         return self.f(other, *self.args)
  54     class _Int:
  55         def __ror__(self, other):
  56             try:    return int(''.join(filter(str.isdigit, other)))
  57             except: return None
  58     int = _Int()
  59
  60
  61 def prepare_searchresults(yt_results):
  62     contents = listfind(yt_results, 'response') \
  63         .get('contents',{})\
  64         .get('twoColumnSearchResultsRenderer',{})\
  65         .get('primaryContents',{})\
  66         .get('sectionListRenderer',{})\
  67         .get('contents',[])
  68     contents = flatten([c.get('contents',[]) for c in contents|Select(all='itemSectionRenderer')])
  69
  70     return parse_result_items(contents)
  71
  72 def prepare_infocards(metadata):
  73     cards = metadata.get('cards',{}).get('cardCollectionRenderer',{}).get('cards',[])
  74     return list(filter(None, map(parse_infocard, cards)))
  75
  76 def prepare_endcards(metadata):
  77     endsc = metadata.get('endscreen',{}).get('endscreenRenderer',{}).get('elements',[])
  78     return list(filter(None, map(parse_endcard, endsc)))
  79
  80 def prepare_channel(result, channel_id):
  81     response = listfind(result,'response')
  82
  83     if 'alerts' in response: # possibly got an error back
  84         from flask import current_app
  85         current_app.logger.error([(alert['alertRenderer']['type'],alert['alertRenderer']['text']['simpleText']) for alert in response['alerts']])
  86         return None,None,[],[],False
  87
  88     meta1 = response.get('metadata',{}).get('channelMetadataRenderer',{})
  89     meta2 = response.get('microformat',{}).get('microformatDataRenderer',{})
  90     title = meta1.get('title', meta2.get('title'))
  91     descr = meta1.get('description', meta2.get('description')) # meta2.description is capped at 160chars
  92     thumb = mkthumbs(meta2.get('thumbnail',meta1.get('avatar',{})).get('thumbnails',{})) # .avatar ~ 900px
  93
  94     contents = response.get('continuationContents')
  95     if not contents: # overran end of list
  96         return title, descr, thumb, [], False
  97
  98     unparsed = contents.get('gridContinuation',{}).get('items') or \
  99             contents.get('sectionListContinuation',{}).get('contents') or []
 100     items, extra = parse_channel_items(unparsed, channel_id, title)
 101     has_more = 'continuations' in (contents.get('gridContinuation') or
 102             contents.get('sectionListContinuation') or {})
 103
 104     return title, descr, thumb, items, has_more
 105
 106 def prepare_playlist(result):
 107     contents = listfind(result,'response')['continuationContents']
 108     unparsed = contents['playlistVideoListContinuation'].get('contents',[])
 109     has_more = 'continuations' in contents.get('playlistVideoListContinuation')
 110
 111     return list(filter(None, map(parse_playlist, unparsed))), has_more
 112
 113 def mkthumbs(thumbs):
 114     output = {str(e['height']): e['url'] for e in thumbs}
 115     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
 116     return {**output, 'largest': largest}
 117
 118 def clean_url(url):
 119     # externals URLs are redirected through youtube.com/redirect, but we
 120     # may encounter internal URLs, too
 121     return parse_qs(urlparse(url).query).get('q',[url])[0]
 122
 123 def toInt(s, fallback=0):
 124     if s is None:
 125         return fallback
 126     try:
 127         return int(''.join(filter(str.isdigit, s)))
 128     except ValueError:
 129         return fallback
 130
 131 # Remove left-/rightmost word from string:
 132 delL = lambda s: s.partition(' ')[2]
 133
 134 def age(s):
 135     if s is None: # missing from autogen'd music, some livestreams
 136         return None
 137     # Some livestreams have "Streamed 7 hours ago"
 138     s = s.replace("Streamed ","")
 139     # Now, everything should be in the form "1 year ago"
 140     value, unit, _ = s.split(" ")
 141     suffix = dict(
 142         month='mn',
 143         months='mn',
 144     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 145
 146     return f"{value}{suffix}"
 147
 148 def log_unknown_card(data):
 149     import json
 150     try:
 151         from flask import request
 152         source = request.url
 153     except: source = "unknown"
 154     with open("/tmp/innertube.err", "a") as f:
 155         f.write(f"\n/***** {source} *****/\n")
 156         json.dump(data, f, indent=2)
 157
 158 def parse_result_items(items):
 159     # TODO: use .get() for most non-essential attributes
 160     """
 161     parses youtube search response into an easier to use format.
 162     """
 163     results = []
 164     extras = []
 165     for item in items:
 166         key = next(iter(item.keys()), None)
 167         content = item[key]
 168         if key == 'videoRenderer':
 169             results.append({'type': 'VIDEO', 'content': {
 170                 'video_id': content['videoId'],
 171                 'title': content['title']|G.text,
 172                 'author': content|G('longBylineText','shortBylineText')|G.text,
 173                 'channel_id': content|G('ownerText')|G('runs')|G(0) \
 174                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 175                 'length': content|G('lengthText')|G.text, # "44:07", "1:41:50"
 176                 'views': content|G('viewCountText')|G.text|A.int or 0, # "1,234 {views|watching}", absent on 0 views
 177                 'published': content|G('publishedTimeText')|G('simpleText')|A(age),
 178                 'live': content|G('badges')|Select('metadataBadgeRenderer')|G('style')=='BADGE_STYLE_TYPE_LIVE_NOW',
 179             }})
 180         elif key in ['playlistRenderer', 'radioRenderer', 'showRenderer']: # radio == "Mix" playlist, show == normal playlist, specially displayed
 181             results.append({'type': 'PLAYLIST', 'content': {
 182                 'playlist_id': content['navigationEndpoint']|G('watchEndpoint')|G('playlistId') or \
 183                     content|G('playlistId'), # COURSE/"learning playlist"
 184                 'video_id': content['navigationEndpoint']|G('watchEndpoint')|G('videoId') or \
 185                     videoid_from_thumbnail(content), # learning playlist
 186                 'title': content['title']|G.text,
 187                 # Note: learning playlists have no author/channel_id
 188                 'author': content|G('longBylineText','shortBylineText')|G.text,
 189                 'channel_id': content|G('longBylineText','shortBylineText')|G('runs')|G(0) \
 190                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 191                 'n_videos': content|G('videoCount')|A.int or \
 192                     content|G('videoCountShortText','videoCountText')|G.text, # "Mix" playlists
 193             }})
 194         elif key == 'channelRenderer':
 195             results.append({'type': 'CHANNEL', 'content': {
 196                 'channel_id': content['channelId'],
 197                 'title': content['title']|G.text,
 198                 'icons': content['thumbnail']['thumbnails']|A(mkthumbs),
 199                 'subscribers': content|G('subscriberCountText')|G('simpleText'), # "2.47K subscribers"
 200             }})
 201         elif key == 'shelfRenderer':
 202             subkey = next(iter(content['content'].keys()), None) #verticalListRenderer/horizontalMovieListRenderer
 203             r, e = parse_result_items(content['content'][subkey]['items'])
 204             results.extend(r)
 205             extras.extend(e)
 206         elif key in ['movieRenderer', 'gridMovieRenderer']: # movies to buy/rent
 207             pass # gMR.{videoId,title.runs[].text,lengthText.simpleText}
 208         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer']: # haha, no.
 209             pass
 210         elif key == 'horizontalCardListRenderer':
 211             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 212             pass
 213         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 214             pass
 215         elif key in ['clarificationRenderer', 'infoPanelContainerRenderer']: # COVID-19/conspiracy theory infos
 216             pass
 217         elif key == 'webAnswerRenderer': # "Result from the web"
 218             pass
 219         elif key == 'didYouMeanRenderer' or key == 'showingResultsForRenderer':
 220             extras.append({
 221                 'type': 'spelling',
 222                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 223                 'autocorrected': key == 'showingResultsForRenderer',
 224             })
 225         elif key == 'messageRenderer': # "No more results"
 226             extras.append({
 227                 'type': 'message',
 228                 'message': content|G('title','text')|G.text,
 229             })
 230         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 231             extras.append({
 232                 'type': content['icon']['iconType'],
 233                 'message': content['title']|G.text,
 234             })
 235         else:
 236             log_unknown_card(item)
 237     return results, extras
 238
 239 def parse_infocard(card):
 240     """
 241     parses a single infocard into a format that's easier to handle.
 242     """
 243     card = card['cardRenderer']
 244     ctype = list(card['content'].keys())[0]
 245     content = card['content'][ctype]
 246     if ctype == "pollRenderer":
 247         return {'type': "POLL", 'content': {
 248             'question': content['question']['simpleText'],
 249             'answers': [(a['text']['simpleText'],a['numVotes']) \
 250                 for a in content['choices']],
 251         }}
 252     elif ctype == "videoInfoCardContentRenderer":
 253         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 254         return {'type': "VIDEO", 'content': {
 255             'video_id': content['action']['watchEndpoint']['videoId'],
 256             'title': content['videoTitle']['simpleText'],
 257             'author': delL(content['channelName']['simpleText']),
 258             'length': content.get('lengthString',{}).get('simpleText') \
 259                 if not is_live else "LIVE", # "23:03"
 260             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 261                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 262         }}
 263     elif ctype == "playlistInfoCardContentRenderer":
 264         return {'type': "PLAYLIST", 'content': {
 265             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 266             'video_id': content['action']['watchEndpoint']['videoId'],
 267             'title': content['playlistTitle']['simpleText'],
 268             'author': delL(content['channelName']['simpleText']),
 269             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 270         }}
 271     elif ctype == "simpleCardContentRenderer" and \
 272             'urlEndpoint' in content['command']:
 273         return {'type': "WEBSITE", 'content': {
 274             'url': clean_url(content['command']['urlEndpoint']['url']),
 275             'domain': content['displayDomain']['simpleText'],
 276             'title': content['title']['simpleText'],
 277             # XXX: no thumbnails for infocards
 278         }}
 279     elif ctype == "collaboratorInfoCardContentRenderer":
 280         return {'type': "CHANNEL", 'content': {
 281             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 282             'title': content['channelName']['simpleText'],
 283             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 284             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 285         }}
 286     else:
 287         log_unknown_card(card)
 288         return None
 289
 290 def parse_endcard(card):
 291     """
 292     parses a single endcard into a format that's easier to handle.
 293     """
 294     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 295     ctype = card['style']
 296     if ctype == "CHANNEL":
 297         return {'type': ctype, 'content': {
 298             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 299             'title': card['title']['simpleText'],
 300             'icons': mkthumbs(card['image']['thumbnails']),
 301         }}
 302     elif ctype == "VIDEO":
 303         return {'type': ctype, 'content': {
 304             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 305             'title': card['title']['simpleText'],
 306             'length': card['videoDuration']['simpleText'],  # '12:21'
 307             'views': toInt(card['metadata']['simpleText']),
 308             # XXX: no channel name
 309         }}
 310     elif ctype == "PLAYLIST":
 311         return {'type': ctype, 'content': {
 312             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 313             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 314             'title': card['title']['simpleText'],
 315             'author': delL(card['metadata']['simpleText']),
 316             'n_videos': toInt(card['playlistLength']['simpleText']),
 317         }}
 318     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 319         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 320         return {'type': "WEBSITE", 'content': {
 321             'url': url,
 322             'domain': urlparse(url).netloc,
 323             'title': card['title']['simpleText'],
 324             'icons': mkthumbs(card['image']['thumbnails']),
 325         }}
 326     else:
 327         log_unknown_card(card)
 328         return None
 329
 330 def videoid_from_thumbnail(content):
 331     # learning playlist; example: PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5 (/user/enyay/playlists)
 332     return re.match(r"https?://i.ytimg.com/vi/([-_0-9a-zA-Z]{11})|()",
 333         listget(listget(content.get('thumbnails',[]),0,{}).get('thumbnails',[]),0,{}).get('url','')
 334     ).group(1)
 335
 336 def parse_channel_items(items, channel_id, author):
 337     result = []
 338     extra = []
 339     for item in items:
 340         key = next(iter(item.keys()), None)
 341         content = item[key]
 342         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer"]:
 343             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 344             result.append({'type': 'VIDEO', 'content': {
 345                 'video_id': content['videoId'],
 346                 'title': content['title'].get('simpleText') or content['title'].get('runs',[{}])[0].get('text'),
 347                 'author': content.get('bylineText',{}).get('runs',[{}])[0].get('text') or author,
 348                 'channel_id': content.get('bylineText',{}).get('runs',[{}])[0] \
 349                     .get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId') or channel_id,
 350                 'length': (content.get('lengthText',{}).get('simpleText') or # topic channel
 351                     listfind(content.get('thumbnailOverlays',[]),'thumbnailOverlayTimeStatusRenderer')
 352                         .get('text',{}).get('simpleText')),
 353                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 354                 'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 355                 'published': age(content.get('publishedTimeText',{}).get('simpleText')),
 356             }})
 357         elif key == "gridPlaylistRenderer" or key == "playlistRenderer":
 358             result.append({'type': 'PLAYLIST', 'content': {
 359                 'playlist_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('playlistId') or content.get('playlistId'),
 360                 'video_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('videoId',{}) or videoid_from_thumbnail(content),
 361                 'title': (content['title'].get('simpleText') or # playlistRenderer
 362                     content['title']['runs'][0]['text']), # gridPlaylistRenderer
 363                 'author': author,
 364                 'channel_id': channel_id,
 365                 'n_videos': toInt(content.get('videoCount') or # playlistRenderer
 366                     content.get('videoCountShortText',{}).get('simpleText') or # grid(1)
 367                     content.get('videoCountText',{}).get('runs',[{}])[0].get('text')), # grid(2)
 368             }})
 369         elif key == "showRenderer":
 370             result.append({'type': 'PLAYLIST', 'content': {
 371                 'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 372                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 373                 'title': content['title']['simpleText'],
 374                 'author': author,
 375                 'channel_id': channel_id,
 376                 'n_videos': None,
 377             }})
 378         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer"]:
 379                 newkey = {
 380                     "itemSectionRenderer": 'contents',
 381                     "gridRenderer": 'items',
 382                     "horizontalCardListRenderer": 'cards',
 383                 }.get(key)
 384                 r, e = parse_channel_items(content[newkey], channel_id, author)
 385                 result.extend(r)
 386                 extra.extend(e)
 387         elif key == "shelfRenderer":
 388                 r, e = parse_channel_items([content['content']], channel_id, author)
 389                 result.extend(r)
 390                 extra.extend(e)
 391         elif key == "messageRenderer":
 392             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 393             pass
 394         elif key == "gameCardRenderer":
 395             pass
 396         else:
 397             log_unknown_card(item)
 398
 399     return result, extra
 400
 401 def parse_playlist(item):
 402     key = next(iter(item.keys()), None)
 403     content = item[key]
 404     if key == "playlistVideoRenderer":
 405         if not content.get('isPlayable', False):
 406             return None # private or deleted video
 407
 408         return {'type': 'VIDEO', 'content': {
 409             'video_id': content['videoId'],
 410             'title': (content['title'].get('simpleText') or # playable videos
 411                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 412             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 413             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 414             # rest is missing from unplayable videos:
 415             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 416             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 417             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 418                 int(content.get("lengthSeconds", 0))), # "531"
 419             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 420         }}
 421     else:
 422         raise Exception(item) # XXX TODO