[southpark:de] Add test for non-ASCII in URLs

[youtube-dl] / youtube_dl / utils.py

diff --git a/youtube_dl/utils.py b/youtube_dl/utils.py
index ae813099dded05c15ed57d27e51be7704e778520..c7db75f80b5a544a0893fb11a371a682fee98e41 100644 (file)
--- a/youtube_dl/utils.py
+++ b/youtube_dl/utils.py
@@ -139,21 +139,24 @@ def write_json_file(obj, fn):
 
 
 if sys.version_info >= (2, 7):
 
 
 if sys.version_info >= (2, 7):

-    def find_xpath_attr(node, xpath, key, val):
+    def find_xpath_attr(node, xpath, key, val=None):

         """ Find the xpath xpath[@key=val] """
         assert re.match(r'^[a-zA-Z-]+$', key)
         """ Find the xpath xpath[@key=val] """
         assert re.match(r'^[a-zA-Z-]+$', key)

-        assert re.match(r'^[a-zA-Z0-9@\s:._-]*$', val)
-        expr = xpath + "[@%s='%s']" % (key, val)
+        if val:
+            assert re.match(r'^[a-zA-Z0-9@\s:._-]*$', val)
+        expr = xpath + ('[@%s]' % key if val is None else "[@%s='%s']" % (key, val))

         return node.find(expr)
 else:
         return node.find(expr)
 else:

-    def find_xpath_attr(node, xpath, key, val):
+    def find_xpath_attr(node, xpath, key, val=None):

         # Here comes the crazy part: In 2.6, if the xpath is a unicode,
         # .//node does not match if a node is a direct child of . !
         if isinstance(xpath, compat_str):
             xpath = xpath.encode('ascii')
 
         for f in node.findall(xpath):
         # Here comes the crazy part: In 2.6, if the xpath is a unicode,
         # .//node does not match if a node is a direct child of . !
         if isinstance(xpath, compat_str):
             xpath = xpath.encode('ascii')
 
         for f in node.findall(xpath):

-            if f.attrib.get(key) == val:
+            if key not in f.attrib:
+                continue
+            if val is None or f.attrib.get(key) == val:

                 return f
         return None
 
                 return f
         return None
 


@@ -576,11 +579,9 @@ class ContentTooShortError(Exception):
     download is too small for what the server announced first, indicating
     the connection was probably interrupted.
     """
     download is too small for what the server announced first, indicating
     the connection was probably interrupted.
     """

-    # Both in bytes
-    downloaded = None
-    expected = None

 
     def __init__(self, downloaded, expected):
 
     def __init__(self, downloaded, expected):

+        # Both in bytes

         self.downloaded = downloaded
         self.expected = expected
 
         self.downloaded = downloaded
         self.expected = expected
 


@@ -650,6 +651,26 @@ class YoutubeDLHandler(compat_urllib_request.HTTPHandler):
         return ret
 
     def http_request(self, req):
         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
         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