X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/9f2b33881274af98a9145c533a1d295fad71521a..82a01b147e1b7241e1040c446b10b58b658c70cf:/youtube_dl/extractor/iprima.py diff --git a/youtube_dl/extractor/iprima.py b/youtube_dl/extractor/iprima.py index 61a0de4..788bbe0 100644 --- a/youtube_dl/extractor/iprima.py +++ b/youtube_dl/extractor/iprima.py @@ -1,4 +1,4 @@ -# -*- coding: utf-8 -*- +# coding: utf-8 from __future__ import unicode_literals import re @@ -6,6 +6,8 @@ import time from .common import InfoExtractor from ..utils import ( + determine_ext, + js_to_json, sanitized_Request, ) @@ -30,8 +32,7 @@ class IPrimaIE(InfoExtractor): }] def _real_extract(self, url): - mobj = re.match(self._VALID_URL, url) - video_id = mobj.group('id') + video_id = self._match_id(url) webpage = self._download_webpage(url, video_id) @@ -43,9 +44,42 @@ class IPrimaIE(InfoExtractor): req.add_header('Referer', url) playerpage = self._download_webpage(req, video_id, note='Downloading player') - m3u8_url = self._search_regex(r"'src': '([^']+\.m3u8)'", playerpage, 'm3u8 url') + formats = [] - formats = self._extract_m3u8_formats(m3u8_url, video_id, ext='mp4') + def extract_formats(format_url, format_key=None, lang=None): + ext = determine_ext(format_url) + new_formats = [] + if format_key == 'hls' or ext == 'm3u8': + new_formats = self._extract_m3u8_formats( + format_url, video_id, 'mp4', entry_protocol='m3u8_native', + m3u8_id='hls', fatal=False) + elif format_key == 'dash' or ext == 'mpd': + return + new_formats = self._extract_mpd_formats( + format_url, video_id, mpd_id='dash', fatal=False) + if lang: + for f in new_formats: + if not f.get('language'): + f['language'] = lang + formats.extend(new_formats) + + options = self._parse_json( + self._search_regex( + r'(?s)var\s+playerOptions\s*=\s*({.+?});', + playerpage, 'player options', default='{}'), + video_id, transform_source=js_to_json, fatal=False) + if options: + for key, tracks in options.get('tracks', {}).items(): + if not isinstance(tracks, list): + continue + for track in tracks: + src = track.get('src') + if src: + extract_formats(src, key.lower(), track.get('lang')) + + if not formats: + for _, src in re.findall(r'src["\']\s*:\s*(["\'])(.+?)\1', playerpage): + extract_formats(src) self._sort_formats(formats)