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     meta = result|G('sidebar')|G('playlistSidebarRenderer')|G('items')
 113     meta1 = meta|Select('playlistSidebarPrimaryInfoRenderer')
 114     meta2 = meta|Select('playlistSidebarSecondaryInfoRenderer') \
 115         |G('videoOwner')|G('videoOwnerRenderer')
 116     title = meta1|G('title')|G.text
 117     author = meta2|G('title')|G.text
 118     channel_id = meta2|G('navigationEndpoint')|G('browseEndpoint')|G('browseId')
 119
 120     return title, author, channel_id, list(filter(None, map(parse_playlist, unparsed))), more
 121
 122 def mkthumbs(thumbs):
 123     output = {str(e['height']): e['url'] for e in thumbs}
 124     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
 125     return {**output, 'largest': largest}
 126
 127 def clean_url(url):
 128     # externals URLs are redirected through youtube.com/redirect, but we
 129     # may encounter internal URLs, too
 130     return parse_qs(urlparse(url).query).get('q',[url])[0]
 131
 132 def toInt(s, fallback=0):
 133     if s is None:
 134         return fallback
 135     try:
 136         return int(''.join(filter(str.isdigit, s)))
 137     except ValueError:
 138         return fallback
 139
 140 # Remove left-/rightmost word from string:
 141 delL = lambda s: s.partition(' ')[2]
 142
 143 def age(s):
 144     if s is None: # missing from autogen'd music, some livestreams
 145         return None
 146     # Some livestreams have "Streamed 7 hours ago"
 147     s = s.replace("Streamed ","")
 148     # Now, everything should be in the form "1 year ago"
 149     value, unit, _ = s.split(" ")
 150     suffix = dict(
 151         minute='min',
 152         minutes='min',
 153     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 154
 155     return f"{value}{suffix}"
 156
 157 def log_unknown_card(data):
 158     import json
 159     try:
 160         from flask import request
 161         source = request.url
 162     except: source = "unknown"
 163     with open("/tmp/innertube.err", "a", encoding="utf-8", errors="backslashreplace") as f:
 164         f.write(f"\n/***** {source} *****/\n")
 165         json.dump(data, f, indent=2)
 166
 167 def parse_result_items(items):
 168     # TODO: use .get() for most non-essential attributes
 169     """
 170     parses youtube search response into an easier to use format.
 171     """
 172     results = []
 173     extras = []
 174     for item in items:
 175         key = next(iter(item.keys()), None)
 176         content = item[key]
 177         if key == 'videoRenderer':
 178             results.append({'type': 'VIDEO', 'content': {
 179                 'video_id': content['videoId'],
 180                 'title': content['title']|G.text,
 181                 'author': content|G('longBylineText','shortBylineText')|G.text,
 182                 'channel_id': content|G('ownerText')|G('runs')|G(0) \
 183                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 184                 'length': content|G('lengthText')|G.text, # "44:07", "1:41:50"
 185                 'views': content|G('viewCountText')|G.text|A.int or 0, # "1,234 {views|watching}", absent on 0 views
 186                 'published': content|G('publishedTimeText')|G('simpleText')|A(age),
 187                 'live': content|G('badges')|Select('metadataBadgeRenderer')|G('style')=='BADGE_STYLE_TYPE_LIVE_NOW',
 188             }})
 189         elif key in ['playlistRenderer', 'radioRenderer', 'showRenderer']: # radio == "Mix" playlist, show == normal playlist, specially displayed
 190             results.append({'type': 'PLAYLIST', 'content': {
 191                 'playlist_id': content['navigationEndpoint']|G('watchEndpoint')|G('playlistId'),
 192                 'video_id': content['navigationEndpoint']|G('watchEndpoint')|G('videoId'),
 193                 'title': content['title']|G.text,
 194                 'author': content|G('longBylineText','shortBylineText')|G.text,
 195                 'channel_id': content|G('longBylineText','shortBylineText')|G('runs')|G(0) \
 196                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 197                 'n_videos': content|G('videoCount')|A.int or \
 198                     content|G('videoCountShortText','videoCountText')|G.text, # "Mix" playlists
 199             }})
 200         elif key == 'channelRenderer':
 201             results.append({'type': 'CHANNEL', 'content': {
 202                 'channel_id': content['channelId'],
 203                 'title': content['title']|G.text,
 204                 'icons': content['thumbnail']['thumbnails']|A(mkthumbs),
 205                 'subscribers': content|G('subscriberCountText')|G('simpleText'), # "2.47K subscribers"
 206             }})
 207         elif key == 'shelfRenderer':
 208             subkey = next(iter(content['content'].keys()), None) #verticalListRenderer/horizontalMovieListRenderer
 209             r, e = parse_result_items(content['content'][subkey]['items'])
 210             results.extend(r)
 211             extras.extend(e)
 212         elif key in ['movieRenderer', 'gridMovieRenderer']: # movies to buy/rent
 213             pass # gMR.{videoId,title.runs[].text,lengthText.simpleText}
 214         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer',
 215                 'promotedSparklesWebRenderer','compactPromotedItemRenderer']: # haha, no.
 216             pass
 217         elif key == 'horizontalCardListRenderer':
 218             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 219             pass
 220         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 221             pass
 222         elif key in ['clarificationRenderer', 'infoPanelContainerRenderer']: # COVID-19/conspiracy theory infos
 223             pass
 224         elif key == 'webAnswerRenderer': # "Result from the web"
 225             pass
 226         elif key == 'infoPanelContentRenderer': # "These results may be new or changing quickly"
 227             pass
 228         elif key == 'hashtagTileRenderer': # link to '/hashtag/<search_query>'
 229             pass
 230         elif key in ['didYouMeanRenderer', 'showingResultsForRenderer', 'includingResultsForRenderer']:
 231             extras.append({
 232                 'type': 'spelling',
 233                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 234                 'autocorrected': key in ['showingResultsForRenderer', 'includingResultsForRenderer'],
 235             })
 236         elif key == 'messageRenderer': # "No more results"
 237             extras.append({
 238                 'type': 'message',
 239                 'message': content|G('title','text')|G.text,
 240             })
 241         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 242             extras.append({
 243                 'type': content['icon']['iconType'],
 244                 'message': content['title']|G.text,
 245             })
 246         else:
 247             log_unknown_card(item)
 248     return results, extras
 249
 250 def parse_infocard(card):
 251     """
 252     parses a single infocard into a format that's easier to handle.
 253     """
 254     card = card['cardRenderer']
 255     if not 'content' in card:
 256         return None  # probably the "View corrections" card, ignore.
 257     ctype = list(card['content'].keys())[0]
 258     content = card['content'][ctype]
 259     if ctype == "pollRenderer":
 260         return {'type': "POLL", 'content': {
 261             'question': content['question']['simpleText'],
 262             'answers': [(a['text']['simpleText'],a['numVotes']) \
 263                 for a in content['choices']],
 264         }}
 265     elif ctype == "videoInfoCardContentRenderer":
 266         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 267         return {'type': "VIDEO", 'content': {
 268             'video_id': content['action']['watchEndpoint']['videoId'],
 269             'title': content['videoTitle']['simpleText'],
 270             'author': delL(content['channelName']['simpleText']),
 271             'length': content.get('lengthString',{}).get('simpleText') \
 272                 if not is_live else "LIVE", # "23:03"
 273             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 274                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 275         }}
 276     elif ctype == "playlistInfoCardContentRenderer":
 277         return {'type': "PLAYLIST", 'content': {
 278             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 279             'video_id': content['action']['watchEndpoint']['videoId'],
 280             'title': content['playlistTitle']['simpleText'],
 281             'author': delL(content['channelName']['simpleText']),
 282             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 283         }}
 284     elif ctype == "simpleCardContentRenderer" and \
 285             'urlEndpoint' in content['command']:
 286         return {'type': "WEBSITE", 'content': {
 287             'url': clean_url(content['command']['urlEndpoint']['url']),
 288             'domain': content['displayDomain']['simpleText'],
 289             'title': content['title']['simpleText'],
 290             # XXX: no thumbnails for infocards
 291         }}
 292     elif ctype == "collaboratorInfoCardContentRenderer":
 293         return {'type': "CHANNEL", 'content': {
 294             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 295             'title': content['channelName']['simpleText'],
 296             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 297             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 298         }}
 299     else:
 300         log_unknown_card(card)
 301         return None
 302
 303 def parse_endcard(card):
 304     """
 305     parses a single endcard into a format that's easier to handle.
 306     """
 307     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 308     ctype = card['style']
 309     if ctype == "CHANNEL":
 310         return {'type': ctype, 'content': {
 311             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 312             'title': card['title']|G.text,
 313             'icons': mkthumbs(card['image']['thumbnails']),
 314         }}
 315     elif ctype == "VIDEO":
 316         if not 'endpoint' in card: return None # title == "This video is unavailable."
 317         return {'type': ctype, 'content': {
 318             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 319             'title': card['title']|G.text,
 320             'length': card|G('videoDuration')|G.text,  # '12:21'
 321             'views': toInt(card['metadata']|G.text),
 322             # XXX: no channel name
 323         }}
 324     elif ctype == "PLAYLIST":
 325         return {'type': ctype, 'content': {
 326             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 327             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 328             'title': card['title']|G.text,
 329             'author': delL(card['metadata']|G.text),
 330             'n_videos': toInt(card['playlistLength']|G.text),
 331         }}
 332     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 333         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 334         return {'type': "WEBSITE", 'content': {
 335             'url': url,
 336             'domain': urlparse(url).netloc,
 337             'title': card['title']|G.text,
 338             'icons': mkthumbs(card['image']['thumbnails']),
 339         }}
 340     else:
 341         log_unknown_card(card)
 342         return None
 343
 344 def parse_channel_items(items, channel_id, author):
 345     result = []
 346     extra = []
 347     for item in items:
 348         key = next(iter(item.keys()), None)
 349         content = item[key]
 350         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer"]:
 351             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 352             result.append({'type': 'VIDEO', 'content': {
 353                 'video_id': content['videoId'],
 354                 'title': content|G('title')|G.text,
 355                 'author': content|G('bylineText')|G.text or author,
 356                 'channel_id': (content|G('bylineText')|G('runs')
 357                     |Select('navigationEndpoint')
 358                     |G('browseEndpoint')|G('browseId') or channel_id),
 359                 'length': (content|G('lengthText')|G.text or # topic channel
 360                     content|G('thumbnailOverlays')
 361                     |Select('thumbnailOverlayTimeStatusRenderer')
 362                     |G('text')|G.text),
 363                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 364                 'views': content|G('viewCountText')|G.text|A.int,
 365                 'published': content|G('publishedTimeText')|G.text|A(age),
 366             }})
 367         elif key in ["gridPlaylistRenderer", "playlistRenderer", "gridRadioRenderer"]:
 368             result.append({'type': 'PLAYLIST', 'content': {
 369                 'playlist_id': content|G('navigationEndpoint')|G('watchEndpoint')|G('playlistId'),
 370                 'video_id': content|G('navigationEndpoint')|G('watchEndpoint')|G('videoId'),
 371                 'title': content|G('title')|G.text,
 372                 'author': author, # Note: gridRadioRenderer is by 'Youtube' without channel_id, ignoring that.
 373                 'channel_id': channel_id,
 374                 'n_videos': (content|G('videoCount')|A.int or # playlistRenderer
 375                     content|G('videoCountShortText','videoCountText')|G.text|A.int) # grid
 376             }})
 377         elif key == "showRenderer":
 378             result.append({'type': 'PLAYLIST', 'content': {
 379                 'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 380                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 381                 'title': content['title']['simpleText'],
 382                 'author': author,
 383                 'channel_id': channel_id,
 384                 'n_videos': None,
 385             }})
 386         elif key in ["gridShowRenderer"]:
 387             result.append({'type': 'PLAYLIST', 'content': {
 388                 'playlist_id': (content|G('navigationEndpoint')
 389                     |G('browseEndpoint')|G('browseId'))[2:],
 390                 #^: playlistId prefixed with 'VL', which must be removed
 391                 'video_id': None,
 392                 'title': content|G('title')|G.text,
 393                 'author': author,
 394                 'channel_id': channel_id,
 395                 'n_videos': content|G('thumbnailOverlays')|G(0)
 396                     |G('thumbnailOverlayBottomPanelRenderer')|G('text')|G.text,
 397             }})
 398         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer", "horizontalListRenderer"]:
 399                 newkey = {
 400                     "itemSectionRenderer": 'contents',
 401                     "gridRenderer": 'items',
 402                     "horizontalCardListRenderer": 'cards',
 403                     "horizontalListRenderer": 'items',
 404                 }.get(key)
 405                 r, e = parse_channel_items(content[newkey], channel_id, author)
 406                 result.extend(r)
 407                 extra.extend(e)
 408         elif key == "shelfRenderer":
 409                 r, e = parse_channel_items([content['content']], channel_id, author)
 410                 result.extend(r)
 411                 extra.extend(e)
 412         elif key == "messageRenderer":
 413             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 414             pass
 415         elif key == "gameCardRenderer":
 416             pass
 417         elif key == "gridChannelRenderer":
 418             pass # don't care; related channels, e.g. on UCMsgXPD3wzzt8RxHJmXH7hQ
 419         else:
 420             log_unknown_card(item)
 421
 422     return result, extra
 423
 424 def parse_playlist(item):
 425     key = next(iter(item.keys()), None)
 426     content = item[key]
 427     if key == "playlistVideoRenderer":
 428         if not content.get('isPlayable', False):
 429             return None # private or deleted video
 430
 431         return {'type': 'VIDEO', 'content': {
 432             'video_id': content['videoId'],
 433             'title': (content['title'].get('simpleText') or # playable videos
 434                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 435             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 436             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 437             # rest is missing from unplayable videos:
 438             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 439             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 440             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 441                 int(content.get("lengthSeconds", 0))), # "531"
 442             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 443         }}
 444     else:
 445         raise Exception(item) # XXX TODO