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