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