From 51f267d9d4d26c3cd67f318a2040513946f2b4d3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sergey=20M=E2=80=A4?= Date: Thu, 6 Aug 2015 22:01:01 +0600 Subject: [PATCH] [YoutubeDL:utils] Move percent encode non-ASCII URLs workaround to http_request and simplify (Closes #6457) --- youtube_dl/YoutubeDL.py | 21 --------------------- youtube_dl/utils.py | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 21 deletions(-) diff --git a/youtube_dl/YoutubeDL.py b/youtube_dl/YoutubeDL.py index 1446b3254..079d42ce8 100755 --- a/youtube_dl/YoutubeDL.py +++ b/youtube_dl/YoutubeDL.py @@ -1860,27 +1860,6 @@ class YoutubeDL(object): def urlopen(self, req): """ Start an HTTP download """ - - # According to RFC 3986, URLs can not contain non-ASCII characters, however this is not - # always respected by websites, some tend to give out URLs with non percent-encoded - # non-ASCII characters (see telemb.py, ard.py [#3412]) - # urllib chokes on URLs with non-ASCII characters (see http://bugs.python.org/issue3991) - # To work around aforementioned issue we will replace request's original URL with - # percent-encoded one - req_is_string = isinstance(req, compat_basestring) - url = req if req_is_string else req.get_full_url() - url_escaped = escape_url(url) - - # Substitute URL if any change after escaping - if url != url_escaped: - if req_is_string: - req = url_escaped - else: - req_type = HEADRequest if req.get_method() == 'HEAD' else compat_urllib_request.Request - req = req_type( - url_escaped, data=req.data, headers=req.headers, - origin_req_host=req.origin_req_host, unverifiable=req.unverifiable) - return self._opener.open(req, timeout=self._socket_timeout) def print_debug_header(self): diff --git a/youtube_dl/utils.py b/youtube_dl/utils.py index 78dc2b449..c7db75f80 100644 --- a/youtube_dl/utils.py +++ b/youtube_dl/utils.py @@ -651,6 +651,26 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler): return ret def http_request(self, req): + # According to RFC 3986, URLs can not contain non-ASCII characters, however this is not + # always respected by websites, some tend to give out URLs with non percent-encoded + # non-ASCII characters (see telemb.py, ard.py [#3412]) + # urllib chokes on URLs with non-ASCII characters (see http://bugs.python.org/issue3991) + # To work around aforementioned issue we will replace request's original URL with + # percent-encoded one + # Since redirects are also affected (e.g. http://www.southpark.de/alle-episoden/s18e09) + # the code of this workaround has been moved here from YoutubeDL.urlopen() + url = req.get_full_url() + url_escaped = escape_url(url) + + # Substitute URL if any change after escaping + if url != url_escaped: + req_type = HEADRequest if req.get_method() == 'HEAD' else compat_urllib_request.Request + new_req = req_type( + url_escaped, data=req.data, headers=req.headers, + origin_req_host=req.origin_req_host, unverifiable=req.unverifiable) + new_req.timeout = req.timeout + req = new_req + for h, v in std_headers.items(): # Capitalize is needed because of Python bug 2275: http://bugs.python.org/issue2275 # The dict keys are capitalized because of this bug by urllib