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(response, channel_id):
  81     meta1 = response|G('metadata')|G('channelMetadataRenderer')
  82     meta2 = response|G('microformat')|G('microformatDataRenderer')
  83     title = meta1|G('title') or meta2|G('title')
  84     descr = meta1|G('description') or meta2|G('description') # meta2.description is capped at 160chars
  85     thumb = mkthumbs((meta2|G('thumbnail') or meta1|G('avatar'))|G('thumbnails') or {}) # .avatar ~ 900px
  86
  87     contents = response|G('continuationContents')
  88     if not contents: # overran end of list
  89         return title, descr, thumb, [], False
  90
  91     unparsed = contents|G('gridContinuation')|G('items') or \
  92             contents|G('sectionListContinuation')|G('contents') or []
  93     items, extra = parse_channel_items(unparsed, channel_id, title)
  94     has_more = 'continuations' in \
  95             (contents|G('gridContinuation', 'sectionListContinuation') or {})
  96
  97     return title, descr, thumb, items, has_more
  98
  99 def prepare_playlist(result):
 100     contents = result['continuationContents']
 101     unparsed = contents['playlistVideoListContinuation'].get('contents',[])
 102     has_more = 'continuations' in contents.get('playlistVideoListContinuation')
 103
 104     return list(filter(None, map(parse_playlist, unparsed))), has_more
 105
 106 def mkthumbs(thumbs):
 107     output = {str(e['height']): e['url'] for e in thumbs}
 108     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
 109     return {**output, 'largest': largest}
 110
 111 def clean_url(url):
 112     # externals URLs are redirected through youtube.com/redirect, but we
 113     # may encounter internal URLs, too
 114     return parse_qs(urlparse(url).query).get('q',[url])[0]
 115
 116 def toInt(s, fallback=0):
 117     if s is None:
 118         return fallback
 119     try:
 120         return int(''.join(filter(str.isdigit, s)))
 121     except ValueError:
 122         return fallback
 123
 124 # Remove left-/rightmost word from string:
 125 delL = lambda s: s.partition(' ')[2]
 126
 127 def age(s):
 128     if s is None: # missing from autogen'd music, some livestreams
 129         return None
 130     # Some livestreams have "Streamed 7 hours ago"
 131     s = s.replace("Streamed ","")
 132     # Now, everything should be in the form "1 year ago"
 133     value, unit, _ = s.split(" ")
 134     suffix = dict(
 135         minute='min',
 136         minutes='min',
 137     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 138
 139     return f"{value}{suffix}"
 140
 141 def log_unknown_card(data):
 142     import json
 143     try:
 144         from flask import request
 145         source = request.url
 146     except: source = "unknown"
 147     with open("/tmp/innertube.err", "a") as f:
 148         f.write(f"\n/***** {source} *****/\n")
 149         json.dump(data, f, indent=2)
 150
 151 def parse_result_items(items):
 152     # TODO: use .get() for most non-essential attributes
 153     """
 154     parses youtube search response into an easier to use format.
 155     """
 156     results = []
 157     extras = []
 158     for item in items:
 159         key = next(iter(item.keys()), None)
 160         content = item[key]
 161         if key == 'videoRenderer':
 162             results.append({'type': 'VIDEO', 'content': {
 163                 'video_id': content['videoId'],
 164                 'title': content['title']|G.text,
 165                 'author': content|G('longBylineText','shortBylineText')|G.text,
 166                 'channel_id': content|G('ownerText')|G('runs')|G(0) \
 167                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 168                 'length': content|G('lengthText')|G.text, # "44:07", "1:41:50"
 169                 'views': content|G('viewCountText')|G.text|A.int or 0, # "1,234 {views|watching}", absent on 0 views
 170                 'published': content|G('publishedTimeText')|G('simpleText')|A(age),
 171                 'live': content|G('badges')|Select('metadataBadgeRenderer')|G('style')=='BADGE_STYLE_TYPE_LIVE_NOW',
 172             }})
 173         elif key in ['playlistRenderer', 'radioRenderer', 'showRenderer']: # radio == "Mix" playlist, show == normal playlist, specially displayed
 174             results.append({'type': 'PLAYLIST', 'content': {
 175                 'playlist_id': content['navigationEndpoint']|G('watchEndpoint')|G('playlistId') or \
 176                     content|G('playlistId'), # COURSE/"learning playlist"
 177                 'video_id': content['navigationEndpoint']|G('watchEndpoint')|G('videoId') or \
 178                     videoid_from_thumbnail(content), # learning playlist
 179                 'title': content['title']|G.text,
 180                 # Note: learning playlists have no author/channel_id
 181                 'author': content|G('longBylineText','shortBylineText')|G.text,
 182                 'channel_id': content|G('longBylineText','shortBylineText')|G('runs')|G(0) \
 183                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 184                 'n_videos': content|G('videoCount')|A.int or \
 185                     content|G('videoCountShortText','videoCountText')|G.text, # "Mix" playlists
 186             }})
 187         elif key == 'channelRenderer':
 188             results.append({'type': 'CHANNEL', 'content': {
 189                 'channel_id': content['channelId'],
 190                 'title': content['title']|G.text,
 191                 'icons': content['thumbnail']['thumbnails']|A(mkthumbs),
 192                 'subscribers': content|G('subscriberCountText')|G('simpleText'), # "2.47K subscribers"
 193             }})
 194         elif key == 'shelfRenderer':
 195             subkey = next(iter(content['content'].keys()), None) #verticalListRenderer/horizontalMovieListRenderer
 196             r, e = parse_result_items(content['content'][subkey]['items'])
 197             results.extend(r)
 198             extras.extend(e)
 199         elif key in ['movieRenderer', 'gridMovieRenderer']: # movies to buy/rent
 200             pass # gMR.{videoId,title.runs[].text,lengthText.simpleText}
 201         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer']: # haha, no.
 202             pass
 203         elif key == 'horizontalCardListRenderer':
 204             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 205             pass
 206         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 207             pass
 208         elif key in ['clarificationRenderer', 'infoPanelContainerRenderer']: # COVID-19/conspiracy theory infos
 209             pass
 210         elif key == 'webAnswerRenderer': # "Result from the web"
 211             pass
 212         elif key == 'didYouMeanRenderer' or key == 'showingResultsForRenderer':
 213             extras.append({
 214                 'type': 'spelling',
 215                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 216                 'autocorrected': key == 'showingResultsForRenderer',
 217             })
 218         elif key == 'messageRenderer': # "No more results"
 219             extras.append({
 220                 'type': 'message',
 221                 'message': content|G('title','text')|G.text,
 222             })
 223         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 224             extras.append({
 225                 'type': content['icon']['iconType'],
 226                 'message': content['title']|G.text,
 227             })
 228         else:
 229             log_unknown_card(item)
 230     return results, extras
 231
 232 def parse_infocard(card):
 233     """
 234     parses a single infocard into a format that's easier to handle.
 235     """
 236     card = card['cardRenderer']
 237     ctype = list(card['content'].keys())[0]
 238     content = card['content'][ctype]
 239     if ctype == "pollRenderer":
 240         return {'type': "POLL", 'content': {
 241             'question': content['question']['simpleText'],
 242             'answers': [(a['text']['simpleText'],a['numVotes']) \
 243                 for a in content['choices']],
 244         }}
 245     elif ctype == "videoInfoCardContentRenderer":
 246         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 247         return {'type': "VIDEO", 'content': {
 248             'video_id': content['action']['watchEndpoint']['videoId'],
 249             'title': content['videoTitle']['simpleText'],
 250             'author': delL(content['channelName']['simpleText']),
 251             'length': content.get('lengthString',{}).get('simpleText') \
 252                 if not is_live else "LIVE", # "23:03"
 253             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 254                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 255         }}
 256     elif ctype == "playlistInfoCardContentRenderer":
 257         return {'type': "PLAYLIST", 'content': {
 258             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 259             'video_id': content['action']['watchEndpoint']['videoId'],
 260             'title': content['playlistTitle']['simpleText'],
 261             'author': delL(content['channelName']['simpleText']),
 262             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 263         }}
 264     elif ctype == "simpleCardContentRenderer" and \
 265             'urlEndpoint' in content['command']:
 266         return {'type': "WEBSITE", 'content': {
 267             'url': clean_url(content['command']['urlEndpoint']['url']),
 268             'domain': content['displayDomain']['simpleText'],
 269             'title': content['title']['simpleText'],
 270             # XXX: no thumbnails for infocards
 271         }}
 272     elif ctype == "collaboratorInfoCardContentRenderer":
 273         return {'type': "CHANNEL", 'content': {
 274             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 275             'title': content['channelName']['simpleText'],
 276             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 277             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 278         }}
 279     else:
 280         log_unknown_card(card)
 281         return None
 282
 283 def parse_endcard(card):
 284     """
 285     parses a single endcard into a format that's easier to handle.
 286     """
 287     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 288     ctype = card['style']
 289     if ctype == "CHANNEL":
 290         return {'type': ctype, 'content': {
 291             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 292             'title': card['title']['simpleText'],
 293             'icons': mkthumbs(card['image']['thumbnails']),
 294         }}
 295     elif ctype == "VIDEO":
 296         if not 'endpoint' in card: return None # title == "This video is unavailable."
 297         return {'type': ctype, 'content': {
 298             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 299             'title': card['title']['simpleText'],
 300             'length': card['videoDuration']['simpleText'],  # '12:21'
 301             'views': toInt(card['metadata']['simpleText']),
 302             # XXX: no channel name
 303         }}
 304     elif ctype == "PLAYLIST":
 305         return {'type': ctype, 'content': {
 306             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 307             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 308             'title': card['title']['simpleText'],
 309             'author': delL(card['metadata']['simpleText']),
 310             'n_videos': toInt(card['playlistLength']['simpleText']),
 311         }}
 312     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 313         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 314         return {'type': "WEBSITE", 'content': {
 315             'url': url,
 316             'domain': urlparse(url).netloc,
 317             'title': card['title']['simpleText'],
 318             'icons': mkthumbs(card['image']['thumbnails']),
 319         }}
 320     else:
 321         log_unknown_card(card)
 322         return None
 323
 324 def videoid_from_thumbnail(content):
 325     # learning playlist; example: PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5 (/user/enyay/playlists)
 326     return re.match(r"https?://i.ytimg.com/vi/([-_0-9a-zA-Z]{11})|()",
 327         listget(listget(content.get('thumbnails',[]),0,{}).get('thumbnails',[]),0,{}).get('url','')
 328     ).group(1)
 329
 330 def parse_channel_items(items, channel_id, author):
 331     result = []
 332     extra = []
 333     for item in items:
 334         key = next(iter(item.keys()), None)
 335         content = item[key]
 336         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer"]:
 337             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 338             result.append({'type': 'VIDEO', 'content': {
 339                 'video_id': content['videoId'],
 340                 'title': content['title'].get('simpleText') or content['title'].get('runs',[{}])[0].get('text'),
 341                 'author': content.get('bylineText',{}).get('runs',[{}])[0].get('text') or author,
 342                 'channel_id': content.get('bylineText',{}).get('runs',[{}])[0] \
 343                     .get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId') or channel_id,
 344                 'length': (content.get('lengthText',{}).get('simpleText') or # topic channel
 345                     listfind(content.get('thumbnailOverlays',[]),'thumbnailOverlayTimeStatusRenderer')
 346                         .get('text',{}).get('simpleText')),
 347                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 348                 'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 349                 'published': age(content.get('publishedTimeText',{}).get('simpleText')),
 350             }})
 351         elif key == "gridPlaylistRenderer" or key == "playlistRenderer":
 352             result.append({'type': 'PLAYLIST', 'content': {
 353                 'playlist_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('playlistId') or content.get('playlistId'),
 354                 'video_id': content['navigationEndpoint'].get('watchEndpoint',{}).get('videoId',{}) or videoid_from_thumbnail(content),
 355                 'title': (content['title'].get('simpleText') or # playlistRenderer
 356                     content['title']['runs'][0]['text']), # gridPlaylistRenderer
 357                 'author': author,
 358                 'channel_id': channel_id,
 359                 'n_videos': toInt(content.get('videoCount') or # playlistRenderer
 360                     content.get('videoCountShortText',{}).get('simpleText') or # grid(1)
 361                     content.get('videoCountText',{}).get('runs',[{}])[0].get('text')), # grid(2)
 362             }})
 363         elif key == "showRenderer":
 364             result.append({'type': 'PLAYLIST', 'content': {
 365                 'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 366                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 367                 'title': content['title']['simpleText'],
 368                 'author': author,
 369                 'channel_id': channel_id,
 370                 'n_videos': None,
 371             }})
 372         elif key in ["gridShowRenderer"]:
 373             result.append({'type': 'PLAYLIST', 'content': {
 374                 'playlist_id': (content|G('navigationEndpoint')
 375                     |G('browseEndpoint')|G('browseId'))[2:],
 376                 #^: playlistId prefixed with 'VL', which must be removed
 377                 'video_id': None,
 378                 'title': content|G('title')|G.text,
 379                 'author': author,
 380                 'channel_id': channel_id,
 381                 'n_videos': content|G('thumbnailOverlays')|G(0)
 382                     |G('thumbnailOverlayBottomPanelRenderer')|G('text')|G.text,
 383             }})
 384         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer", "horizontalListRenderer"]:
 385                 newkey = {
 386                     "itemSectionRenderer": 'contents',
 387                     "gridRenderer": 'items',
 388                     "horizontalCardListRenderer": 'cards',
 389                     "horizontalListRenderer": 'items',
 390                 }.get(key)
 391                 r, e = parse_channel_items(content[newkey], channel_id, author)
 392                 result.extend(r)
 393                 extra.extend(e)
 394         elif key == "shelfRenderer":
 395                 r, e = parse_channel_items([content['content']], channel_id, author)
 396                 result.extend(r)
 397                 extra.extend(e)
 398         elif key == "messageRenderer":
 399             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 400             pass
 401         elif key == "gameCardRenderer":
 402             pass
 403         elif key == "gridChannelRenderer":
 404             pass # don't care; related channels, e.g. on UCMsgXPD3wzzt8RxHJmXH7hQ
 405         else:
 406             log_unknown_card(item)
 407
 408     return result, extra
 409
 410 def parse_playlist(item):
 411     key = next(iter(item.keys()), None)
 412     content = item[key]
 413     if key == "playlistVideoRenderer":
 414         if not content.get('isPlayable', False):
 415             return None # private or deleted video
 416
 417         return {'type': 'VIDEO', 'content': {
 418             'video_id': content['videoId'],
 419             'title': (content['title'].get('simpleText') or # playable videos
 420                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 421             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 422             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 423             # rest is missing from unplayable videos:
 424             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 425             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 426             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 427                 int(content.get("lengthSeconds", 0))), # "531"
 428             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 429         }}
 430     else:
 431         raise Exception(item) # XXX TODO