yt_dlp/extractor/gedidigital.py

   1 import re
   2
   3 from .common import InfoExtractor
   4 from ..utils import (
   5     base_url,
   6     determine_ext,
   7     int_or_none,
   8     url_basename,
   9     urljoin,
  10 )
  11
  12
  13 class GediDigitalIE(InfoExtractor):
  14     _VALID_URL = r'''(?x:(?P<base_url>(?:https?:)//video\.
  15         (?:
  16             (?:
  17                 (?:espresso\.)?repubblica
  18                 |lastampa
  19                 |ilsecoloxix
  20                 |huffingtonpost
  21             )|
  22             (?:
  23                 iltirreno
  24                 |messaggeroveneto
  25                 |ilpiccolo
  26                 |gazzettadimantova
  27                 |mattinopadova
  28                 |laprovinciapavese
  29                 |tribunatreviso
  30                 |nuovavenezia
  31                 |gazzettadimodena
  32                 |lanuovaferrara
  33                 |corrierealpi
  34                 |lasentinella
  35             )\.gelocal
  36         )\.it(?:/[^/]+){2,4}/(?P<id>\d+))(?:$|[?&].*))'''
  37     _EMBED_REGEX = [rf'''(?x)
  38             (?:
  39                 data-frame-src=|
  40                 <iframe[^\n]+src=
  41             )
  42             (["'])(?P<url>{_VALID_URL})\1''']
  43     _TESTS = [{
  44         'url': 'https://video.lastampa.it/politica/il-paradosso-delle-regionali-la-lega-vince-ma-sembra-aver-perso/121559/121683',
  45         'md5': '84658d7fb9e55a6e57ecc77b73137494',
  46         'info_dict': {
  47             'id': '121683',
  48             'ext': 'mp4',
  49             'title': 'Il paradosso delle Regionali: ecco perché la Lega vince ma sembra aver perso',
  50             'description': 'md5:de7f4d6eaaaf36c153b599b10f8ce7ca',
  51             'thumbnail': r're:^https://www\.repstatic\.it/video/photo/.+?-thumb-full-.+?\.jpg$',
  52             'duration': 125,
  53         },
  54     }, {
  55         'url': 'https://video.huffingtonpost.it/embed/politica/cotticelli-non-so-cosa-mi-sia-successo-sto-cercando-di-capire-se-ho-avuto-un-malore/29312/29276?responsive=true&el=video971040871621586700',
  56         'only_matching': True,
  57     }, {
  58         'url': 'https://video.espresso.repubblica.it/embed/tutti-i-video/01-ted-villa/14772/14870&width=640&height=360',
  59         'only_matching': True,
  60     }, {
  61         'url': 'https://video.repubblica.it/motori/record-della-pista-a-spa-francorchamps-la-pagani-huayra-roadster-bc-stupisce/367415/367963',
  62         'only_matching': True,
  63     }, {
  64         'url': 'https://video.ilsecoloxix.it/sport/cassani-e-i-brividi-azzurri-ai-mondiali-di-imola-qui-mi-sono-innamorato-del-ciclismo-da-ragazzino-incredibile-tornarci-da-ct/66184/66267',
  65         'only_matching': True,
  66     }, {
  67         'url': 'https://video.iltirreno.gelocal.it/sport/dentro-la-notizia-ferrari-cosa-succede-a-maranello/141059/142723',
  68         'only_matching': True,
  69     }, {
  70         'url': 'https://video.messaggeroveneto.gelocal.it/locale/maria-giovanna-elmi-covid-vaccino/138155/139268',
  71         'only_matching': True,
  72     }, {
  73         'url': 'https://video.ilpiccolo.gelocal.it/dossier/big-john/dinosauro-big-john-al-via-le-visite-guidate-a-trieste/135226/135751',
  74         'only_matching': True,
  75     }, {
  76         'url': 'https://video.gazzettadimantova.gelocal.it/locale/dal-ponte-visconteo-di-valeggio-l-and-8217sos-dei-ristoratori-aprire-anche-a-cena/137310/137818',
  77         'only_matching': True,
  78     }, {
  79         'url': 'https://video.mattinopadova.gelocal.it/dossier/coronavirus-in-veneto/covid-a-vo-un-anno-dopo-un-cuore-tricolore-per-non-dimenticare/138402/138964',
  80         'only_matching': True,
  81     }, {
  82         'url': 'https://video.laprovinciapavese.gelocal.it/locale/mede-zona-rossa-via-alle-vaccinazioni-per-gli-over-80/137545/138120',
  83         'only_matching': True,
  84     }, {
  85         'url': 'https://video.tribunatreviso.gelocal.it/dossier/coronavirus-in-veneto/ecco-le-prima-vaccinazioni-di-massa-nella-marca/134485/135024',
  86         'only_matching': True,
  87     }, {
  88         'url': 'https://video.nuovavenezia.gelocal.it/locale/camion-troppo-alto-per-il-ponte-ferroviario-perde-il-carico/135734/136266',
  89         'only_matching': True,
  90     }, {
  91         'url': 'https://video.gazzettadimodena.gelocal.it/locale/modena-scoperta-la-proteina-che-predice-il-livello-di-gravita-del-covid/139109/139796',
  92         'only_matching': True,
  93     }, {
  94         'url': 'https://video.lanuovaferrara.gelocal.it/locale/due-bombole-di-gpl-aperte-e-abbandonate-i-vigili-bruciano-il-gas/134391/134957',
  95         'only_matching': True,
  96     }, {
  97         'url': 'https://video.corrierealpi.gelocal.it/dossier/cortina-2021-i-mondiali-di-sci-alpino/mondiali-di-sci-il-timelapse-sulla-splendida-olympia/133760/134331',
  98         'only_matching': True,
  99     }, {
 100         'url': 'https://video.lasentinella.gelocal.it/locale/vestigne-centra-un-auto-e-si-ribalta/138931/139466',
 101         'only_matching': True,
 102     }, {
 103         'url': 'https://video.espresso.repubblica.it/tutti-i-video/01-ted-villa/14772',
 104         'only_matching': True,
 105     }]
 106
 107     @staticmethod
 108     def _sanitize_urls(urls):
 109         # add protocol if missing
 110         for i, e in enumerate(urls):
 111             if e.startswith('//'):
 112                 urls[i] = f'https:{e}'
 113         # clean iframes urls
 114         for i, e in enumerate(urls):
 115             urls[i] = urljoin(base_url(e), url_basename(e))
 116         return urls
 117
 118     @classmethod
 119     def _extract_embed_urls(cls, url, webpage):
 120         return cls._sanitize_urls(tuple(super()._extract_embed_urls(url, webpage)))
 121
 122     @staticmethod
 123     def _clean_formats(formats):
 124         format_urls = set()
 125         clean_formats = []
 126         for f in formats:
 127             if f['url'] not in format_urls:
 128                 if f.get('audio_ext') != 'none' and not f.get('acodec'):
 129                     continue
 130                 format_urls.add(f['url'])
 131                 clean_formats.append(f)
 132         formats[:] = clean_formats
 133
 134     def _real_extract(self, url):
 135         video_id, url = self._match_valid_url(url).group('id', 'base_url')
 136         webpage = self._download_webpage(url, video_id)
 137         title = self._html_search_meta(
 138             ['twitter:title', 'og:title'], webpage, fatal=True)
 139         player_data = re.findall(
 140             r"PlayerFactory\.setParam\('(?P<type>format|param)',\s*'(?P<name>[^']+)',\s*'(?P<val>[^']+)'\);",
 141             webpage)
 142
 143         formats = []
 144         duration = thumb = None
 145         for t, n, v in player_data:
 146             if t == 'format':
 147                 if n in ('video-hds-vod-ec', 'video-hls-vod-ec', 'video-viralize', 'video-youtube-pfp'):
 148                     continue
 149                 elif n.endswith('-vod-ak'):
 150                     formats.extend(self._extract_akamai_formats(
 151                         v, video_id, {'http': 'media.gedidigital.it'}))
 152                 else:
 153                     ext = determine_ext(v)
 154                     if ext == 'm3u8':
 155                         formats.extend(self._extract_m3u8_formats(
 156                             v, video_id, 'mp4', 'm3u8_native', m3u8_id=n, fatal=False))
 157                         continue
 158                     f = {
 159                         'format_id': n,
 160                         'url': v,
 161                     }
 162                     if ext == 'mp3':
 163                         abr = int_or_none(self._search_regex(
 164                             r'-mp3-audio-(\d+)', v, 'abr', default=None))
 165                         f.update({
 166                             'abr': abr,
 167                             'tbr': abr,
 168                             'acodec': ext,
 169                             'vcodec': 'none',
 170                         })
 171                     else:
 172                         mobj = re.match(r'^video-rrtv-(\d+)(?:-(\d+))?$', n)
 173                         if mobj:
 174                             f.update({
 175                                 'height': int(mobj.group(1)),
 176                                 'vbr': int_or_none(mobj.group(2)),
 177                             })
 178                         if not f.get('vbr'):
 179                             f['vbr'] = int_or_none(self._search_regex(
 180                                 r'-video-rrtv-(\d+)', v, 'abr', default=None))
 181                     formats.append(f)
 182             elif t == 'param':
 183                 if n in ['image_full', 'image']:
 184                     thumb = v
 185                 elif n == 'videoDuration':
 186                     duration = int_or_none(v)
 187
 188         self._clean_formats(formats)
 189
 190         return {
 191             'id': video_id,
 192             'title': title,
 193             'description': self._html_search_meta(
 194                 ['twitter:description', 'og:description', 'description'], webpage),
 195             'thumbnail': thumb or self._og_search_thumbnail(webpage),
 196             'formats': formats,
 197             'duration': duration,
 198         }