From 31f224008e39b629d793a8fc3f8fafc7a4baf417 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sergey=20M=E2=80=A4?= Date: Sun, 19 Apr 2015 04:07:45 +0600 Subject: [PATCH] [megavideozeu] Simplify (Closes #5454) --- youtube_dl/extractor/megavideozeu.py | 59 ++++++++++++++-------------- 1 file changed, 30 insertions(+), 29 deletions(-) diff --git a/youtube_dl/extractor/megavideozeu.py b/youtube_dl/extractor/megavideozeu.py index ee26b0f2e..f98080caa 100644 --- a/youtube_dl/extractor/megavideozeu.py +++ b/youtube_dl/extractor/megavideozeu.py @@ -1,51 +1,52 @@ # encoding: utf-8 from __future__ import unicode_literals +import re + from .common import InfoExtractor from ..utils import ( - int_or_none, - parse_filesize, - unified_strdate, + float_or_none, + xpath_text, ) class MegavideozeuIE(InfoExtractor): - _VALID_URL = r'https?://(?:www\.)?megavideoz\.eu/video/(?P.*)(?:.*)' - _TESTS = [ - { - 'url': 'http://megavideoz.eu/video/WM6UB919XMXH/SMPTE-Universal-Film-Leader', - 'info_dict': { - 'id': '48723', - 'ext': 'mp4', - 'duration': '10', - 'title': 'SMPTE Universal Film Leader', - } + _VALID_URL = r'https?://(?:www\.)?megavideoz\.eu/video/(?P[^/]+)(?:/(?P[^/]+))?' + _TEST = { + 'url': 'http://megavideoz.eu/video/WM6UB919XMXH/SMPTE-Universal-Film-Leader', + 'info_dict': { + 'id': '48723', + 'display_id': 'SMPTE-Universal-Film-Leader', + 'ext': 'mp4', + 'title': 'SMPTE Universal Film Leader', + 'thumbnail': 're:https?://.*?\.jpg', + 'duration': 10.93, } - ] - + } def _real_extract(self, url): - tmp_video_id = self._match_id(url) + mobj = re.match(self._VALID_URL, url) + video_id = mobj.group('id') + display_id = mobj.group('display_id') or video_id - webpage = self._download_webpage(url, tmp_video_id) + webpage = self._download_webpage(url, display_id) - config_php = self._html_search_regex( - r'var cnf = \'([^\']+)\'', webpage, 'config.php url') + config = self._download_xml( + self._search_regex( + r"var\s+cnf\s*=\s*'([^']+)'", webpage, 'cnf url'), + display_id) - configpage = self._download_webpage(config_php, tmp_video_id) - - video_id = self._html_search_regex( - r'([^<]+)', configpage, 'video id') - video_url = self._html_search_regex( - r'([^<]+)', configpage, 'video URL') - title = self._html_search_regex( - r'<!\[CDATA\[([^\]]+)', configpage, 'title') - duration = int_or_none(self._html_search_regex( - r'<duration>([0-9\.]+)', configpage, 'duration', fatal=False)) + video_url = xpath_text(config, './file', 'video url', fatal=True) + title = xpath_text(config, './title', 'title', fatal=True) + thumbnail = xpath_text(config, './image', 'thumbnail') + duration = float_or_none(xpath_text(config, './duration', 'duration')) + video_id = xpath_text(config, './mediaid', 'video id') or video_id return { 'id': video_id, + 'display_id': display_id, 'url': video_url, 'title': title, + 'thumbnail': thumbnail, 'duration': duration }