From 33a1ec950c97b03e742926f3d37e9aa4ce642633 Mon Sep 17 00:00:00 2001 From: remitamine Date: Sat, 30 Apr 2016 20:38:45 +0100 Subject: [PATCH] [discovery] extract http formats --- youtube_dl/extractor/discovery.py | 50 +++++++++++++++++++++++++++++-- 1 file changed, 47 insertions(+), 3 deletions(-) diff --git a/youtube_dl/extractor/discovery.py b/youtube_dl/extractor/discovery.py index 5f1275b39..6d1f8e670 100644 --- a/youtube_dl/extractor/discovery.py +++ b/youtube_dl/extractor/discovery.py @@ -1,11 +1,16 @@ from __future__ import unicode_literals +import re + from .common import InfoExtractor from ..utils import ( parse_duration, parse_iso8601, ) -from ..compat import compat_str +from ..compat import ( + compat_str, + compat_urlparse, +) class DiscoveryIE(InfoExtractor): @@ -66,9 +71,48 @@ class DiscoveryIE(InfoExtractor): entries = [] for idx, video_info in enumerate(info['playlist']): - formats = self._extract_m3u8_formats( - video_info['src'], display_id, 'mp4', 'm3u8_native', m3u8_id='hls', + m3u8_url = video_info['src'] + formats = m3u8_formats = self._extract_m3u8_formats( + m3u8_url, display_id, 'mp4', 'm3u8_native', m3u8_id='hls', note='Download m3u8 information for video %d' % (idx + 1)) + qualities_basename = self._search_regex( + '/([^/]+)\.csmil/', m3u8_url, 'qualities basename', default=None) + if qualities_basename: + m3u8_path = compat_urlparse.urlparse(m3u8_url).path + QUALITIES_RE = r'((,\d+k)+,?)' + qualities = self._search_regex( + QUALITIES_RE, qualities_basename, + 'qualities', default=None) + if qualities: + qualities = list(map(lambda q: int(q[:-1]), qualities.strip(',').split(','))) + qualities.sort() + http_path = m3u8_path[1:].split('/', 1)[1] + http_template = re.sub(QUALITIES_RE, r'%dk', http_path) + http_template = http_template.replace('.csmil/master.m3u8', '') + http_template = compat_urlparse.urljoin( + 'http://discsmil.edgesuite.net/', http_template) + if m3u8_formats: + self._sort_formats(m3u8_formats) + m3u8_formats = list(filter( + lambda f: f.get('vcodec') != 'none' and f.get('resolution') != 'multiple', + m3u8_formats)) + if len(qualities) == len(m3u8_formats): + for q, m3u8_format in zip(qualities, m3u8_formats): + f = m3u8_format.copy() + f.update({ + 'url': http_template % q, + 'format_id': f['format_id'].replace('hls', 'http'), + 'protocol': 'http', + }) + formats.append(f) + else: + for q in qualities: + formats.append({ + 'url': http_template % q, + 'ext': 'mp4', + 'format_id': 'http-%d' % q, + 'tbr': q, + }) self._sort_formats(formats) entries.append({ 'id': compat_str(video_info['id']), -- 2.22.2