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 class G:
   7     """
   8     null-coalescing version of dict.get() that also works on lists.
   9
  10     the | operator is overloaded to achieve similar looking code to jq(1) filters.
  11     the first found key is used: dict(foo=1)|G('bar','foo') returns 1.
  12     """
  13     def __init__(self, *keys):
  14         self.keys = keys
  15     def __ror__(self, other):
  16         for key in self.keys:
  17             try:    return other[key]
  18             except: continue
  19         return None
  20     class _Text:
  21         """ parses youtube's .runs[].text and .simpleText variants """
  22         def __ror__(self, other): # Note: only returning runs[0], not concat'ing all!
  23             return other|G('simpleText') or other|G('runs')|G(0)|G('text')
  24     text = _Text()
  25 class Select:
  26     """ |Select('foo') returns the first foo in list, |Select(all='foo') returns all foos. """
  27     def __init__(self, key=None, *, all=None):
  28         self.key = key or all
  29         self.all = all
  30     def __ror__(self, other):
  31         try: items = [ other[self.key] for other in other if self.key in other.keys() ]
  32         except: items = []
  33         return items if self.all else items|G(0)
  34 class A:
  35     """ apply """
  36     def __init__(self, f, *args):
  37         self.f = f
  38         self.args = args
  39     def __ror__(self, other):
  40         return self.f(other, *self.args)
  41     class _Int:
  42         def __ror__(self, other):
  43             try:    return int(''.join(filter(str.isdigit, other)))
  44             except: return None
  45     int = _Int()
  46
  47
  48 def prepare_searchresults(yt_results):
  49     contents = ( # from continuation token
  50         yt_results
  51         |G('onResponseReceivedCommands')
  52         |Select('appendContinuationItemsAction')
  53         |G('continuationItems')
  54     ) or ( # from page 1
  55         yt_results
  56         |G('contents')
  57         |G('twoColumnSearchResultsRenderer')
  58         |G('primaryContents')
  59         |G('sectionListRenderer')
  60         |G('contents')
  61     )
  62     items = contents|Select('itemSectionRenderer')|G('contents')
  63     items, extra = parse_result_items(items)
  64     more = contents|Select("continuationItemRenderer")|G("continuationEndpoint")|G("continuationCommand")|G("token")
  65     estimatedResults = yt_results|G("estimatedResults")
  66
  67     return items, extra, more
  68
  69 def prepare_infocards(metadata):
  70     cards = metadata.get('cards',{}).get('cardCollectionRenderer',{}).get('cards',[])
  71     return list(filter(None, map(parse_infocard, cards)))
  72
  73 def prepare_endcards(metadata):
  74     endsc = metadata.get('endscreen',{}).get('endscreenRenderer',{}).get('elements',[])
  75     return list(filter(None, map(parse_endcard, endsc)))
  76
  77 def prepare_channel(response, channel_id):
  78     meta1 = response|G('metadata')|G('channelMetadataRenderer')
  79     meta2 = response|G('microformat')|G('microformatDataRenderer')
  80     title = meta1|G('title') or meta2|G('title')
  81     descr = meta1|G('description') or meta2|G('description') # meta2.description is capped at 160chars
  82     thumb = mkthumbs((meta2|G('thumbnail') or meta1|G('avatar'))|G('thumbnails') or {}) # .avatar ~ 900px
  83
  84     contents = response|G('continuationContents')
  85     if not contents: # overran end of list
  86         return title, descr, thumb, [], False
  87
  88     unparsed = contents|G('gridContinuation')|G('items') or \
  89             contents|G('sectionListContinuation')|G('contents') or []
  90     items, extra = parse_channel_items(unparsed, channel_id, title)
  91     more = (
  92         contents
  93         |G('gridContinuation', 'sectionListContinuation')
  94         |G('continuations')
  95         |Select('nextContinuationData')
  96         |G('continuation')
  97     )
  98
  99     return title, descr, thumb, items, more
 100
 101 def prepare_playlist(result):
 102     contents = result['continuationContents']
 103     unparsed = contents['playlistVideoListContinuation'].get('contents',[])
 104     more = (
 105         contents
 106         |G('playlistVideoListContinuation')
 107         |G('continuations')
 108         |Select('nextContinuationData')
 109         |G('continuation')
 110     )
 111
 112     return list(filter(None, map(parse_playlist, unparsed))), more
 113
 114 def mkthumbs(thumbs):
 115     output = {str(e['height']): e['url'] for e in thumbs}
 116     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
 117     return {**output, 'largest': largest}
 118
 119 def clean_url(url):
 120     # externals URLs are redirected through youtube.com/redirect, but we
 121     # may encounter internal URLs, too
 122     return parse_qs(urlparse(url).query).get('q',[url])[0]
 123
 124 def toInt(s, fallback=0):
 125     if s is None:
 126         return fallback
 127     try:
 128         return int(''.join(filter(str.isdigit, s)))
 129     except ValueError:
 130         return fallback
 131
 132 # Remove left-/rightmost word from string:
 133 delL = lambda s: s.partition(' ')[2]
 134
 135 def age(s):
 136     if s is None: # missing from autogen'd music, some livestreams
 137         return None
 138     # Some livestreams have "Streamed 7 hours ago"
 139     s = s.replace("Streamed ","")
 140     # Now, everything should be in the form "1 year ago"
 141     value, unit, _ = s.split(" ")
 142     suffix = dict(
 143         minute='min',
 144         minutes='min',
 145     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 146
 147     return f"{value}{suffix}"
 148
 149 def log_unknown_card(data):
 150     import json
 151     try:
 152         from flask import request
 153         source = request.url
 154     except: source = "unknown"
 155     with open("/tmp/innertube.err", "a", encoding="utf-8", errors="backslashreplace") as f:
 156         f.write(f"\n/***** {source} *****/\n")
 157         json.dump(data, f, indent=2)
 158
 159 def parse_result_items(items):
 160     # TODO: use .get() for most non-essential attributes
 161     """
 162     parses youtube search response into an easier to use format.
 163     """
 164     results = []
 165     extras = []
 166     for item in items:
 167         key = next(iter(item.keys()), None)
 168         content = item[key]
 169         if key == 'videoRenderer':
 170             results.append({'type': 'VIDEO', 'content': {
 171                 'video_id': content['videoId'],
 172                 'title': content['title']|G.text,
 173                 'author': content|G('longBylineText','shortBylineText')|G.text,
 174                 'channel_id': content|G('ownerText')|G('runs')|G(0) \
 175                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 176                 'length': content|G('lengthText')|G.text, # "44:07", "1:41:50"
 177                 'views': content|G('viewCountText')|G.text|A.int or 0, # "1,234 {views|watching}", absent on 0 views
 178                 'published': content|G('publishedTimeText')|G('simpleText')|A(age),
 179                 'live': content|G('badges')|Select('metadataBadgeRenderer')|G('style')=='BADGE_STYLE_TYPE_LIVE_NOW',
 180             }})
 181         elif key in ['playlistRenderer', 'radioRenderer', 'showRenderer']: # radio == "Mix" playlist, show == normal playlist, specially displayed
 182             results.append({'type': 'PLAYLIST', 'content': {
 183                 'playlist_id': content['navigationEndpoint']|G('watchEndpoint')|G('playlistId') or \
 184                     content|G('playlistId'), # COURSE/"learning playlist"
 185                 'video_id': content['navigationEndpoint']|G('watchEndpoint')|G('videoId') or \
 186                     videoid_from_thumbnail(content), # learning playlist
 187                 'title': content['title']|G.text,
 188                 # Note: learning playlists have no author/channel_id
 189                 'author': content|G('longBylineText','shortBylineText')|G.text,
 190                 'channel_id': content|G('longBylineText','shortBylineText')|G('runs')|G(0) \
 191                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 192                 'n_videos': content|G('videoCount')|A.int or \
 193                     content|G('videoCountShortText','videoCountText')|G.text, # "Mix" playlists
 194             }})
 195         elif key == 'channelRenderer':
 196             results.append({'type': 'CHANNEL', 'content': {
 197                 'channel_id': content['channelId'],
 198                 'title': content['title']|G.text,
 199                 'icons': content['thumbnail']['thumbnails']|A(mkthumbs),
 200                 'subscribers': content|G('subscriberCountText')|G('simpleText'), # "2.47K subscribers"
 201             }})
 202         elif key == 'shelfRenderer':
 203             subkey = next(iter(content['content'].keys()), None) #verticalListRenderer/horizontalMovieListRenderer
 204             r, e = parse_result_items(content['content'][subkey]['items'])
 205             results.extend(r)
 206             extras.extend(e)
 207         elif key in ['movieRenderer', 'gridMovieRenderer']: # movies to buy/rent
 208             pass # gMR.{videoId,title.runs[].text,lengthText.simpleText}
 209         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer']: # haha, no.
 210             pass
 211         elif key == 'horizontalCardListRenderer':
 212             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 213             pass
 214         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 215             pass
 216         elif key in ['clarificationRenderer', 'infoPanelContainerRenderer']: # COVID-19/conspiracy theory infos
 217             pass
 218         elif key == 'webAnswerRenderer': # "Result from the web"
 219             pass
 220         elif key == 'didYouMeanRenderer' or key == 'showingResultsForRenderer':
 221             extras.append({
 222                 'type': 'spelling',
 223                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 224                 'autocorrected': key == 'showingResultsForRenderer',
 225             })
 226         elif key == 'messageRenderer': # "No more results"
 227             extras.append({
 228                 'type': 'message',
 229                 'message': content|G('title','text')|G.text,
 230             })
 231         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 232             extras.append({
 233                 'type': content['icon']['iconType'],
 234                 'message': content['title']|G.text,
 235             })
 236         else:
 237             log_unknown_card(item)
 238     return results, extras
 239
 240 def parse_infocard(card):
 241     """
 242     parses a single infocard into a format that's easier to handle.
 243     """
 244     card = card['cardRenderer']
 245     ctype = list(card['content'].keys())[0]
 246     content = card['content'][ctype]
 247     if ctype == "pollRenderer":
 248         return {'type': "POLL", 'content': {
 249             'question': content['question']['simpleText'],
 250             'answers': [(a['text']['simpleText'],a['numVotes']) \
 251                 for a in content['choices']],
 252         }}
 253     elif ctype == "videoInfoCardContentRenderer":
 254         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 255         return {'type': "VIDEO", 'content': {
 256             'video_id': content['action']['watchEndpoint']['videoId'],
 257             'title': content['videoTitle']['simpleText'],
 258             'author': delL(content['channelName']['simpleText']),
 259             'length': content.get('lengthString',{}).get('simpleText') \
 260                 if not is_live else "LIVE", # "23:03"
 261             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 262                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 263         }}
 264     elif ctype == "playlistInfoCardContentRenderer":
 265         return {'type': "PLAYLIST", 'content': {
 266             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 267             'video_id': content['action']['watchEndpoint']['videoId'],
 268             'title': content['playlistTitle']['simpleText'],
 269             'author': delL(content['channelName']['simpleText']),
 270             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 271         }}
 272     elif ctype == "simpleCardContentRenderer" and \
 273             'urlEndpoint' in content['command']:
 274         return {'type': "WEBSITE", 'content': {
 275             'url': clean_url(content['command']['urlEndpoint']['url']),
 276             'domain': content['displayDomain']['simpleText'],
 277             'title': content['title']['simpleText'],
 278             # XXX: no thumbnails for infocards
 279         }}
 280     elif ctype == "collaboratorInfoCardContentRenderer":
 281         return {'type': "CHANNEL", 'content': {
 282             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 283             'title': content['channelName']['simpleText'],
 284             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 285             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 286         }}
 287     else:
 288         log_unknown_card(card)
 289         return None
 290
 291 def parse_endcard(card):
 292     """
 293     parses a single endcard into a format that's easier to handle.
 294     """
 295     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 296     ctype = card['style']
 297     if ctype == "CHANNEL":
 298         return {'type': ctype, 'content': {
 299             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 300             'title': card['title']|G.text,
 301             'icons': mkthumbs(card['image']['thumbnails']),
 302         }}
 303     elif ctype == "VIDEO":
 304         if not 'endpoint' in card: return None # title == "This video is unavailable."
 305         return {'type': ctype, 'content': {
 306             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 307             'title': card['title']|G.text,
 308             'length': card['videoDuration']|G.text,  # '12:21'
 309             'views': toInt(card['metadata']|G.text),
 310             # XXX: no channel name
 311         }}
 312     elif ctype == "PLAYLIST":
 313         return {'type': ctype, 'content': {
 314             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 315             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 316             'title': card['title']|G.text,
 317             'author': delL(card['metadata']|G.text),
 318             'n_videos': toInt(card['playlistLength']|G.text),
 319         }}
 320     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 321         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 322         return {'type': "WEBSITE", 'content': {
 323             'url': url,
 324             'domain': urlparse(url).netloc,
 325             'title': card['title']|G.text,
 326             'icons': mkthumbs(card['image']['thumbnails']),
 327         }}
 328     else:
 329         log_unknown_card(card)
 330         return None
 331
 332 def videoid_from_thumbnail(content):
 333     # learning playlist; example: PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5 (/user/enyay/playlists)
 334     return re.match(r"https?://i.ytimg.com/vi/([-_0-9a-zA-Z]{11})|()",
 335         content|G('thumbnails')|G(0)|G('thumbnails')|G(0)|G('url') or ''
 336     ).group(1)
 337
 338 def parse_channel_items(items, channel_id, author):
 339     result = []
 340     extra = []
 341     for item in items:
 342         key = next(iter(item.keys()), None)
 343         content = item[key]
 344         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer"]:
 345             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 346             result.append({'type': 'VIDEO', 'content': {
 347                 'video_id': content['videoId'],
 348                 'title': content|G('title')|G.text,
 349                 'author': content|G('bylineText')|G.text or author,
 350                 'channel_id': (content|G('bylineText')|G('runs')
 351                     |Select('navigationEndpoint')
 352                     |G('browseEndpoint')|G('browseId') or channel_id),
 353                 'length': (content|G('lengthText')|G.text or # topic channel
 354                     content|G('thumbnailOverlays')
 355                     |Select('thumbnailOverlayTimeStatusRenderer')
 356                     |G('text')|G.text),
 357                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 358                 'views': content|G('viewCountText')|G.text|A.int,
 359                 'published': content|G('publishedTimeText')|G.text|A(age),
 360             }})
 361         elif key == "gridPlaylistRenderer" or key == "playlistRenderer":
 362             result.append({'type': 'PLAYLIST', 'content': {
 363                 'playlist_id': (content|G('navigationEndpoint')|G('watchEndpoint')|G('playlistId')
 364                     or content|G('playlistId')),
 365                 'video_id': (content|G('navigationEndpoint')|G('watchEndpoint')|G('videoId')
 366                     or videoid_from_thumbnail(content)),
 367                 'title': content|G('title')|G.text,
 368                 'author': author,
 369                 'channel_id': channel_id,
 370                 'n_videos': (content|G('videoCount')|A.int or # playlistRenderer
 371                     content|G('videoCountShortText','videoCountText')|G.text|A.int) # grid
 372             }})
 373         elif key == "showRenderer":
 374             result.append({'type': 'PLAYLIST', 'content': {
 375                 'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 376                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 377                 'title': content['title']['simpleText'],
 378                 'author': author,
 379                 'channel_id': channel_id,
 380                 'n_videos': None,
 381             }})
 382         elif key in ["gridShowRenderer"]:
 383             result.append({'type': 'PLAYLIST', 'content': {
 384                 'playlist_id': (content|G('navigationEndpoint')
 385                     |G('browseEndpoint')|G('browseId'))[2:],
 386                 #^: playlistId prefixed with 'VL', which must be removed
 387                 'video_id': None,
 388                 'title': content|G('title')|G.text,
 389                 'author': author,
 390                 'channel_id': channel_id,
 391                 'n_videos': content|G('thumbnailOverlays')|G(0)
 392                     |G('thumbnailOverlayBottomPanelRenderer')|G('text')|G.text,
 393             }})
 394         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer", "horizontalListRenderer"]:
 395                 newkey = {
 396                     "itemSectionRenderer": 'contents',
 397                     "gridRenderer": 'items',
 398                     "horizontalCardListRenderer": 'cards',
 399                     "horizontalListRenderer": 'items',
 400                 }.get(key)
 401                 r, e = parse_channel_items(content[newkey], channel_id, author)
 402                 result.extend(r)
 403                 extra.extend(e)
 404         elif key == "shelfRenderer":
 405                 r, e = parse_channel_items([content['content']], channel_id, author)
 406                 result.extend(r)
 407                 extra.extend(e)
 408         elif key == "messageRenderer":
 409             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 410             pass
 411         elif key == "gameCardRenderer":
 412             pass
 413         elif key == "gridChannelRenderer":
 414             pass # don't care; related channels, e.g. on UCMsgXPD3wzzt8RxHJmXH7hQ
 415         else:
 416             log_unknown_card(item)
 417
 418     return result, extra
 419
 420 def parse_playlist(item):
 421     key = next(iter(item.keys()), None)
 422     content = item[key]
 423     if key == "playlistVideoRenderer":
 424         if not content.get('isPlayable', False):
 425             return None # private or deleted video
 426
 427         return {'type': 'VIDEO', 'content': {
 428             'video_id': content['videoId'],
 429             'title': (content['title'].get('simpleText') or # playable videos
 430                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 431             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 432             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 433             # rest is missing from unplayable videos:
 434             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 435             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 436             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 437                 int(content.get("lengthSeconds", 0))), # "531"
 438             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 439         }}
 440     else:
 441         raise Exception(item) # XXX TODO