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, channel_name):
  78     meta1 = response|G('metadata')|G('channelMetadataRenderer')
  79     meta2 = response|G('microformat')|G('microformatDataRenderer')
  80     title = meta1|G('title') or meta2|G('title') or channel_name
  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 = (
  85         response|G('continuationContents') or
  86         response|G('onResponseReceivedActions')
  87     )
  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             contents|G('richGridContinuation')|G('contents') or \
  94             contents|Select('appendContinuationItemsAction')|G('continuationItems') or \
  95             contents|G(-1)|G('reloadContinuationItemsCommand')|G('continuationItems') or []
  96     items, extra = parse_channel_items(unparsed, channel_id, title)
  97
  98     more = ( # videos, livestreams
  99         unparsed
 100         |Select('continuationItemRenderer')
 101         |G('continuationEndpoint')
 102         |G('continuationCommand')
 103         |G('token')
 104     ) or ( # playlists, search
 105         contents
 106         |G('gridContinuation', 'sectionListContinuation')
 107         |G('continuations')
 108         |Select('nextContinuationData')
 109         |G('continuation')
 110     )
 111
 112     return title, descr, thumb, items, more
 113
 114 def prepare_playlist(result):
 115     contents = result['continuationContents']
 116     unparsed = contents['playlistVideoListContinuation'].get('contents',[])
 117     more = (
 118         contents
 119         |G('playlistVideoListContinuation')
 120         |G('continuations')
 121         |Select('nextContinuationData')
 122         |G('continuation')
 123     )
 124
 125     meta = result|G('sidebar')|G('playlistSidebarRenderer')|G('items')
 126     meta1 = meta|Select('playlistSidebarPrimaryInfoRenderer')
 127     meta2 = meta|Select('playlistSidebarSecondaryInfoRenderer') \
 128         |G('videoOwner')|G('videoOwnerRenderer')
 129     title = meta1|G('title')|G.text
 130     author = meta2|G('title')|G.text
 131     channel_id = meta2|G('navigationEndpoint')|G('browseEndpoint')|G('browseId')
 132
 133     return title, author, channel_id, list(filter(None, map(parse_playlist, unparsed))), more
 134
 135 def mkthumbs(thumbs):
 136     output = {str(e['height']): e['url'] for e in thumbs}
 137     largest=next(iter(sorted(output.keys(),reverse=True,key=int)),None)
 138     return {**output, 'largest': largest}
 139
 140 def clean_url(url):
 141     # externals URLs are redirected through youtube.com/redirect, but we
 142     # may encounter internal URLs, too
 143     return parse_qs(urlparse(url).query).get('q',[url])[0]
 144
 145 def toInt(s, fallback=0):
 146     if s is None:
 147         return fallback
 148     try:
 149         return int(''.join(filter(str.isdigit, s)))
 150     except ValueError:
 151         return fallback
 152
 153 # Remove left-/rightmost word from string:
 154 delL = lambda s: s.partition(' ')[2]
 155
 156 def age(s):
 157     if s is None: # missing from autogen'd music, some livestreams
 158         return None
 159     # Some livestreams have "Streamed 7 hours ago"
 160     s = s.replace("Streamed ","")
 161     # Now, everything should be in the form "1 year ago"
 162     value, unit, _ = s.split(" ")
 163     suffix = dict(
 164         minute='min',
 165         minutes='min',
 166     ).get(unit, unit[0]) # first letter otherwise (e.g. year(s) => y)
 167
 168     return f"{value}{suffix}"
 169
 170 def log_unknown_card(data):
 171     import json
 172     try:
 173         from flask import request
 174         source = request.url
 175     except: source = "unknown"
 176     with open("/tmp/innertube.err", "a", encoding="utf-8", errors="backslashreplace") as f:
 177         f.write(f"\n/***** {source} *****/\n")
 178         json.dump(data, f, indent=2)
 179
 180 def parse_result_items(items):
 181     # TODO: use .get() for most non-essential attributes
 182     """
 183     parses youtube search response into an easier to use format.
 184     """
 185     results = []
 186     extras = []
 187     for item in items:
 188         key = next(iter(item.keys()), None)
 189         content = item[key]
 190         if key == 'videoRenderer':
 191             results.append({'type': 'VIDEO', 'content': {
 192                 'video_id': content['videoId'],
 193                 'title': content['title']|G.text,
 194                 'author': content|G('longBylineText','shortBylineText')|G.text,
 195                 'channel_id': content|G('ownerText')|G('runs')|G(0) \
 196                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 197                 'length': content|G('lengthText')|G.text, # "44:07", "1:41:50"
 198                 'views': content|G('viewCountText')|G.text|A.int or 0, # "1,234 {views|watching}", absent on 0 views
 199                 'published': content|G('publishedTimeText')|G('simpleText')|A(age),
 200                 'live': content|G('badges')|Select('metadataBadgeRenderer')|G('style')=='BADGE_STYLE_TYPE_LIVE_NOW',
 201             }})
 202         elif key in ['playlistRenderer', 'radioRenderer', 'showRenderer']: # radio == "Mix" playlist, show == normal playlist, specially displayed
 203             results.append({'type': 'PLAYLIST', 'content': {
 204                 'playlist_id': content['navigationEndpoint']|G('watchEndpoint')|G('playlistId'),
 205                 'video_id': content['navigationEndpoint']|G('watchEndpoint')|G('videoId'),
 206                 'title': content['title']|G.text,
 207                 'author': content|G('longBylineText','shortBylineText')|G.text,
 208                 'channel_id': content|G('longBylineText','shortBylineText')|G('runs')|G(0) \
 209                     |G('navigationEndpoint')|G('browseEndpoint')|G('browseId'),
 210                 'n_videos': content|G('videoCount')|A.int or \
 211                     content|G('videoCountShortText','videoCountText')|G.text, # "Mix" playlists
 212             }})
 213         elif key == 'channelRenderer':
 214             results.append({'type': 'CHANNEL', 'content': {
 215                 'channel_id': content['channelId'],
 216                 'title': content['title']|G.text,
 217                 'icons': content['thumbnail']['thumbnails']|A(mkthumbs),
 218                 'subscribers': content|G('subscriberCountText')|G('simpleText'), # "2.47K subscribers"
 219             }})
 220         elif key == 'shelfRenderer':
 221             subkey = next(iter(content['content'].keys()), None) #verticalListRenderer/horizontalMovieListRenderer
 222             r, e = parse_result_items(content['content'][subkey]['items'])
 223             results.extend(r)
 224             extras.extend(e)
 225         elif key == 'reelShelfRenderer': # XXX: seems to be Shorts only
 226             pass # TODO?
 227         elif key in ['movieRenderer', 'gridMovieRenderer']: # movies to buy/rent
 228             pass # gMR.{videoId,title.runs[].text,lengthText.simpleText}
 229         elif key in ['carouselAdRenderer','searchPyvRenderer','promotedSparklesTextSearchRenderer',
 230                 'promotedSparklesWebRenderer','compactPromotedItemRenderer']: # haha, no.
 231             pass
 232         elif key == 'horizontalCardListRenderer':
 233             # suggested searches: .cards[].searchRefinementCardRenderer.query.runs[].text
 234             pass
 235         elif key == 'emergencyOneboxRenderer': # suicide prevention hotline
 236             pass
 237         elif key in ['clarificationRenderer', 'infoPanelContainerRenderer']: # COVID-19/conspiracy theory infos
 238             pass
 239         elif key == 'webAnswerRenderer': # "Result from the web"
 240             pass
 241         elif key == 'infoPanelContentRenderer': # "These results may be new or changing quickly"
 242             pass
 243         elif key == 'hashtagTileRenderer': # link to '/hashtag/<search_query>'
 244             pass
 245         elif key in ['didYouMeanRenderer', 'showingResultsForRenderer', 'includingResultsForRenderer']:
 246             extras.append({
 247                 'type': 'spelling',
 248                 'query': content['correctedQueryEndpoint']['searchEndpoint']['query'], # non-misspelled query
 249                 'autocorrected': key in ['showingResultsForRenderer', 'includingResultsForRenderer'],
 250             })
 251         elif key == 'messageRenderer': # "No more results"
 252             extras.append({
 253                 'type': 'message',
 254                 'message': content|G('title','text')|G.text,
 255             })
 256         elif key == 'backgroundPromoRenderer': # e.g. "no results"
 257             extras.append({
 258                 'type': content['icon']['iconType'],
 259                 'message': content['title']|G.text,
 260             })
 261         else:
 262             log_unknown_card(item)
 263     return results, extras
 264
 265 def parse_infocard(card):
 266     """
 267     parses a single infocard into a format that's easier to handle.
 268     """
 269     card = card['cardRenderer']
 270     if not 'content' in card:
 271         return None  # probably the "View corrections" card, ignore.
 272     ctype = list(card['content'].keys())[0]
 273     content = card['content'][ctype]
 274     if ctype == "pollRenderer":
 275         return {'type': "POLL", 'content': {
 276             'question': content['question']['simpleText'],
 277             'answers': [(a['text']['simpleText'],a['numVotes']) \
 278                 for a in content['choices']],
 279         }}
 280     elif ctype == "videoInfoCardContentRenderer":
 281         is_live = content.get('badge',{}).get('liveBadgeRenderer') is not None
 282         return {'type': "VIDEO", 'content': {
 283             'video_id': content['action']['watchEndpoint']['videoId'],
 284             'title': content['videoTitle']['simpleText'],
 285             'author': delL(content['channelName']['simpleText']),
 286             'length': content.get('lengthString',{}).get('simpleText') \
 287                 if not is_live else "LIVE", # "23:03"
 288             'views': toInt(content.get('viewCountText',{}).get('simpleText')),
 289                 # XXX: views sometimes "Starts: July 31, 2020 at 1:30 PM"
 290         }}
 291     elif ctype == "playlistInfoCardContentRenderer":
 292         return {'type': "PLAYLIST", 'content': {
 293             'playlist_id': content['action']['watchEndpoint']['playlistId'],
 294             'video_id': content['action']['watchEndpoint']['videoId'],
 295             'title': content['playlistTitle']['simpleText'],
 296             'author': delL(content['channelName']['simpleText']),
 297             'n_videos': toInt(content['playlistVideoCount']['simpleText']),
 298         }}
 299     elif ctype == "simpleCardContentRenderer" and \
 300             'urlEndpoint' in content['command']:
 301         return {'type': "WEBSITE", 'content': {
 302             'url': clean_url(content['command']['urlEndpoint']['url']),
 303             'domain': content['displayDomain']['simpleText'],
 304             'title': content['title']['simpleText'],
 305             # XXX: no thumbnails for infocards
 306         }}
 307     elif ctype == "collaboratorInfoCardContentRenderer":
 308         return {'type': "CHANNEL", 'content': {
 309             'channel_id': content['endpoint']['browseEndpoint']['browseId'],
 310             'title': content['channelName']['simpleText'],
 311             'icons': mkthumbs(content['channelAvatar']['thumbnails']),
 312             'subscribers': content.get('subscriberCountText',{}).get('simpleText',''), # "545K subscribers"
 313         }}
 314     else:
 315         log_unknown_card(card)
 316         return None
 317
 318 def parse_endcard(card):
 319     """
 320     parses a single endcard into a format that's easier to handle.
 321     """
 322     card = card.get('endscreenElementRenderer', card) #only sometimes nested
 323     ctype = card['style']
 324     if ctype == "CHANNEL":
 325         return {'type': ctype, 'content': {
 326             'channel_id': card['endpoint']['browseEndpoint']['browseId'],
 327             'title': card['title']|G.text,
 328             'icons': mkthumbs(card['image']['thumbnails']),
 329         }}
 330     elif ctype == "VIDEO":
 331         if not 'endpoint' in card: return None # title == "This video is unavailable."
 332         return {'type': ctype, 'content': {
 333             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 334             'title': card['title']|G.text,
 335             'length': card|G('videoDuration')|G.text,  # '12:21'
 336             'views': toInt(card['metadata']|G.text),
 337             # XXX: no channel name
 338         }}
 339     elif ctype == "PLAYLIST":
 340         return {'type': ctype, 'content': {
 341             'playlist_id': card['endpoint']['watchEndpoint']['playlistId'],
 342             'video_id': card['endpoint']['watchEndpoint']['videoId'],
 343             'title': card['title']|G.text,
 344             'author': delL(card['metadata']|G.text),
 345             'n_videos': toInt(card['playlistLength']|G.text),
 346         }}
 347     elif ctype == "WEBSITE" or ctype == "CREATOR_MERCHANDISE":
 348         url = clean_url(card['endpoint']['urlEndpoint']['url'])
 349         return {'type': "WEBSITE", 'content': {
 350             'url': url,
 351             'domain': urlparse(url).netloc,
 352             'title': card['title']|G.text,
 353             'icons': mkthumbs(card['image']['thumbnails']),
 354         }}
 355     else:
 356         log_unknown_card(card)
 357         return None
 358
 359 def parse_channel_items(items, channel_id, author):
 360     result = []
 361     extra = []
 362     for item in items:
 363         key = next(iter(item.keys()), None)
 364         content = item[key]
 365         if key in ["gridVideoRenderer", "videoRenderer", "videoCardRenderer", 'reelItemRenderer']: # reel==youtube-shorts
 366             # only videoCardRenderer (topic channels) has author and channel, others fall back to supplied ones.
 367             result.append({'type': 'VIDEO', 'content': {
 368                 'video_id': content['videoId'],
 369                 'title': content|G('title')|G.text or content|G('headline')|G.text,
 370                 'author': content|G('bylineText')|G.text or author,
 371                 'channel_id': (content|G('bylineText')|G('runs')
 372                     |Select('navigationEndpoint')
 373                     |G('browseEndpoint')|G('browseId') or channel_id),
 374                 'length': (content|G('lengthText')|G.text or # topic channel
 375                     content|G('thumbnailOverlays')
 376                     |Select('thumbnailOverlayTimeStatusRenderer')
 377                     |G('text')|G.text),
 378                 # topic channel: .metadataText.simpleText = "22M views \u00b7 2 months ago"
 379                 'views': content|G('viewCountText')|G.text|A.int,
 380                 'published': content|G('publishedTimeText')|G.text|A(age),
 381             }})
 382         elif key in ["gridPlaylistRenderer", "playlistRenderer", "gridRadioRenderer"]:
 383             result.append({'type': 'PLAYLIST', 'content': {
 384                 'playlist_id': content|G('navigationEndpoint')|G('watchEndpoint')|G('playlistId'),
 385                 'video_id': content|G('navigationEndpoint')|G('watchEndpoint')|G('videoId'),
 386                 'title': content|G('title')|G.text,
 387                 'author': author, # Note: gridRadioRenderer is by 'Youtube' without channel_id, ignoring that.
 388                 'channel_id': channel_id,
 389                 'n_videos': (content|G('videoCount')|A.int or # playlistRenderer
 390                     content|G('videoCountShortText','videoCountText')|G.text|A.int) # grid
 391             }})
 392         elif key == "showRenderer":
 393             result.append({'type': 'PLAYLIST', 'content': {
 394                 'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 395                 'video_id': content['navigationEndpoint']['watchEndpoint']['videoId'],
 396                 'title': content['title']['simpleText'],
 397                 'author': author,
 398                 'channel_id': channel_id,
 399                 'n_videos': None,
 400             }})
 401         elif key in ["gridShowRenderer"]:
 402             result.append({'type': 'PLAYLIST', 'content': {
 403                 'playlist_id': (content|G('navigationEndpoint')
 404                     |G('browseEndpoint')|G('browseId'))[2:],
 405                 #^: playlistId prefixed with 'VL', which must be removed
 406                 'video_id': None,
 407                 'title': content|G('title')|G.text,
 408                 'author': author,
 409                 'channel_id': channel_id,
 410                 'n_videos': content|G('thumbnailOverlays')|G(0)
 411                     |G('thumbnailOverlayBottomPanelRenderer')|G('text')|G.text,
 412             }})
 413         elif key in ["itemSectionRenderer", "gridRenderer", "horizontalCardListRenderer", "horizontalListRenderer"]:
 414                 newkey = {
 415                     "itemSectionRenderer": 'contents',
 416                     "gridRenderer": 'items',
 417                     "horizontalCardListRenderer": 'cards',
 418                     "horizontalListRenderer": 'items',
 419                 }.get(key)
 420                 r, e = parse_channel_items(content[newkey], channel_id, author)
 421                 result.extend(r)
 422                 extra.extend(e)
 423         elif key in ["shelfRenderer", "richItemRenderer"]:
 424                 r, e = parse_channel_items([content['content']], channel_id, author)
 425                 result.extend(r)
 426                 extra.extend(e)
 427         elif key == "messageRenderer":
 428             # e.g. {'messageRenderer': {'text': {'runs': [{'text': 'This channel has no playlists.'}]}}}
 429             pass
 430         elif key == "gameCardRenderer":
 431             pass
 432         elif key == "gridChannelRenderer":
 433             pass # don't care; related channels, e.g. on UCMsgXPD3wzzt8RxHJmXH7hQ
 434         elif key == 'continuationItemRenderer': # handled in parent function
 435             pass
 436         else:
 437             log_unknown_card(item)
 438
 439     return result, extra
 440
 441 def parse_playlist(item):
 442     key = next(iter(item.keys()), None)
 443     content = item[key]
 444     if key == "playlistVideoRenderer":
 445         if not content.get('isPlayable', False):
 446             return None # private or deleted video
 447
 448         return {'type': 'VIDEO', 'content': {
 449             'video_id': content['videoId'],
 450             'title': (content['title'].get('simpleText') or # playable videos
 451                 content['title'].get('runs',[{}])[0].get('text')), # "[Private video]"
 452             'playlist_id': content['navigationEndpoint']['watchEndpoint']['playlistId'],
 453             'index': content['navigationEndpoint']['watchEndpoint'].get('index',0), #or int(content['index']['simpleText']) (absent on course intros; e.g. PL96C35uN7xGJu6skU4TBYrIWxggkZBrF5)
 454             # rest is missing from unplayable videos:
 455             'author': content.get('shortBylineText',{}).get('runs',[{}])[0].get('text'),
 456             'channel_id':content.get('shortBylineText',{}).get('runs',[{}])[0].get('navigationEndpoint',{}).get('browseEndpoint',{}).get('browseId'),
 457             'length': (content.get("lengthText",{}).get("simpleText") or # "8:51"
 458                 int(content.get("lengthSeconds", 0))), # "531"
 459             'starttime': content['navigationEndpoint']['watchEndpoint'].get('startTimeSeconds'),
 460         }}
 461     else:
 462         raise Exception(item) # XXX TODO