youtube-dl/youtube_dl/extractor/yam.py

124 lines
4.1 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..compat import compat_urlparse
from ..utils import (
float_or_none,
month_by_abbreviation,
ExtractorError,
get_element_by_attribute,
)
class YamIE(InfoExtractor):
IE_DESC = '蕃薯藤yam天空部落'
_VALID_URL = r'https?://mymedia\.yam\.com/m/(?P<id>\d+)'
_TESTS = [{
# An audio hosted on Yam
'url': 'http://mymedia.yam.com/m/2283921',
'md5': 'c011b8e262a52d5473d9c2e3c9963b9c',
'info_dict': {
'id': '2283921',
'ext': 'mp3',
'title': '發現 - 趙薇 京華煙雲主題曲',
'description': '發現 - 趙薇 京華煙雲主題曲',
'uploader_id': 'princekt',
'upload_date': '20080807',
'duration': 313.0,
}
}, {
# An external video hosted on YouTube
'url': 'http://mymedia.yam.com/m/3599430',
'md5': '03127cf10d8f35d120a9e8e52e3b17c6',
'info_dict': {
'id': 'CNpEoQlrIgA',
'ext': 'mp4',
'upload_date': '20150306',
'uploader': '新莊社大瑜伽社',
'description': 'md5:11e2e405311633ace874f2e6226c8b17',
'uploader_id': '2323agoy',
'title': '20090412陽明山二子坪-1',
},
'skip': 'Video does not exist',
}, {
'url': 'http://mymedia.yam.com/m/3598173',
'info_dict': {
'id': '3598173',
'ext': 'mp4',
},
'skip': 'cause Yam system error',
}, {
'url': 'http://mymedia.yam.com/m/3599437',
'info_dict': {
'id': '3599437',
'ext': 'mp4',
},
'skip': 'invalid YouTube URL',
}, {
'url': 'http://mymedia.yam.com/m/2373534',
'md5': '7ff74b91b7a817269d83796f8c5890b1',
'info_dict': {
'id': '2373534',
'ext': 'mp3',
'title': '林俊傑&蔡卓妍-小酒窩',
'description': 'md5:904003395a0fcce6cfb25028ff468420',
'upload_date': '20080928',
'uploader_id': 'onliner2',
}
}]
def _real_extract(self, url):
video_id = self._match_id(url)
page = self._download_webpage(url, video_id)
# Check for errors
system_msg = self._html_search_regex(
r'系統訊息(?:<br>|\n|\r)*([^<>]+)<br>', page, 'system message',
default=None)
if system_msg:
raise ExtractorError(system_msg, expected=True)
# Is it hosted externally on YouTube?
youtube_url = self._html_search_regex(
r'<embed src="(http://www.youtube.com/[^"]+)"',
page, 'YouTube url', default=None)
if youtube_url:
return self.url_result(youtube_url, 'Youtube')
title = self._html_search_regex(
r'<h1[^>]+class="heading"[^>]*>\s*(.+)\s*</h1>', page, 'title')
api_page = self._download_webpage(
'http://mymedia.yam.com/api/a/?pID=' + video_id, video_id,
note='Downloading API page')
api_result_obj = compat_urlparse.parse_qs(api_page)
info_table = get_element_by_attribute('class', 'info', page)
uploader_id = self._html_search_regex(
r'<!-- 發表作者 -->[\n ]+<a href="/([a-z0-9]+)"',
info_table, 'uploader id', fatal=False)
mobj = re.search(r'<!-- 發表於 -->(?P<mon>[A-Z][a-z]{2})\s+' +
r'(?P<day>\d{1,2}), (?P<year>\d{4})', page)
if mobj:
upload_date = '%s%02d%02d' % (
mobj.group('year'),
month_by_abbreviation(mobj.group('mon')),
int(mobj.group('day')))
else:
upload_date = None
duration = float_or_none(api_result_obj['totaltime'][0], scale=1000)
return {
'id': video_id,
'url': api_result_obj['mp3file'][0],
'title': title,
'description': self._html_search_meta('description', page),
'duration': duration,
'uploader_id': uploader_id,
'upload_date': upload_date,
}