mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-24 04:56:55 +01:00
replace unused _download_webpage_handle with _download_webpage
This commit is contained in:
parent
6a41a12d29
commit
48058d82dc
@ -51,7 +51,7 @@ class WeiboIE(InfoExtractor):
|
|||||||
})
|
})
|
||||||
|
|
||||||
genvisitor_url = 'https://passport.weibo.com/visitor/genvisitor'
|
genvisitor_url = 'https://passport.weibo.com/visitor/genvisitor'
|
||||||
webpage, _ = self._download_webpage_handle(genvisitor_url, video_id, data=data, headers=headers, note="gen visitor")
|
webpage = self._download_webpage(genvisitor_url, video_id, data=data, headers=headers, note="gen visitor")
|
||||||
|
|
||||||
p = strip_jsonp(webpage)
|
p = strip_jsonp(webpage)
|
||||||
i1 = p.find('{')
|
i1 = p.find('{')
|
||||||
@ -71,9 +71,9 @@ class WeiboIE(InfoExtractor):
|
|||||||
'_rand': random.random()
|
'_rand': random.random()
|
||||||
}
|
}
|
||||||
gencallback_url = "https://passport.weibo.com/visitor/visitor"
|
gencallback_url = "https://passport.weibo.com/visitor/visitor"
|
||||||
self._download_webpage_handle(gencallback_url, video_id, note="gen callback", query=query)
|
self._download_webpage(gencallback_url, video_id, note="gen callback", query=query)
|
||||||
|
|
||||||
webpage, _ = self._download_webpage_handle(url, video_id, note="retry to visit the page")
|
webpage = self._download_webpage(url, video_id, note="retry to visit the page")
|
||||||
|
|
||||||
title = self._html_search_regex(r'<title>(.+?)</title>', webpage, 'title')
|
title = self._html_search_regex(r'<title>(.+?)</title>', webpage, 'title')
|
||||||
|
|
||||||
@ -118,7 +118,7 @@ class WeiboMobileIE(InfoExtractor):
|
|||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
# to get Referer url for genvisitor
|
# to get Referer url for genvisitor
|
||||||
webpage, _ = self._download_webpage_handle(url, video_id, note="visit the page")
|
webpage = self._download_webpage(url, video_id, note="visit the page")
|
||||||
js_code = self._search_regex(r'var\s+\$render_data\s*=\s*\[({.*})\]\[0\] \|\| {};', webpage, 'js_code', flags=re.DOTALL)
|
js_code = self._search_regex(r'var\s+\$render_data\s*=\s*\[({.*})\]\[0\] \|\| {};', webpage, 'js_code', flags=re.DOTALL)
|
||||||
weibo_info = self._parse_json(js_code, video_id, transform_source=js_to_json)
|
weibo_info = self._parse_json(js_code, video_id, transform_source=js_to_json)
|
||||||
page_info = weibo_info['status']['page_info']
|
page_info = weibo_info['status']['page_info']
|
||||||
|
Loading…
Reference in New Issue
Block a user