Header keys should be native strings.

This commit follows a discussion on IRC. For more information, see the
Pull Request associated with it.

diff --git a/requests/models.py b/requests/models.py
index 2439153..62cd0f5 100644 (file)
--- a/requests/models.py
+++ b/requests/models.py
@@ -23,7 +23,7 @@ from .exceptions import HTTPError, RequestException, MissingSchema, InvalidURL
 from .utils import (
     guess_filename, get_auth_from_url, requote_uri,
     stream_decode_response_unicode, to_key_val_list, parse_header_links,
-    iter_slices, guess_json_utf, super_len)
+    iter_slices, guess_json_utf, super_len, to_native_string)
 from .compat import (
     cookielib, urlunparse, urlsplit, urlencode, str, bytes, StringIO,
     is_py2, chardet, json, builtin_str, basestring)
@@ -346,7 +346,7 @@ class PreparedRequest(RequestEncodingMixin, RequestHooksMixin):
         """Prepares the given HTTP headers."""
 
         if headers:
-            headers = dict((name.encode('ascii'), value) for name, value in headers.items())
+            headers = dict((to_native_string(name), value) for name, value in headers.items())
             self.headers = CaseInsensitiveDict(headers)
         else:
             self.headers = CaseInsensitiveDict()

diff --git a/requests/utils.py b/requests/utils.py
index 37aa19e..e592c8b 100644 (file)
--- a/requests/utils.py
+++ b/requests/utils.py
@@ -21,8 +21,8 @@ from netrc import netrc, NetrcParseError
 from . import __version__
 from . import certs
 from .compat import parse_http_list as _parse_list_header
-from .compat import quote, urlparse, bytes, str, OrderedDict, urlunparse
-from .compat import getproxies, proxy_bypass
+from .compat import (quote, urlparse, bytes, str, OrderedDict, urlunparse,
+                     is_py2, is_py3, builtin_str, getproxies, proxy_bypass)
 from .cookies import RequestsCookieJar, cookiejar_from_dict
 from .structures import CaseInsensitiveDict
 
@@ -393,18 +393,18 @@ def get_environ_proxies(url):
     # we're getting isn't in the no_proxy list.
     no_proxy = get_proxy('no_proxy')
     netloc = urlparse(url).netloc
-    
+
     if no_proxy:
         # We need to check whether we match here. We need to see if we match
         # the end of the netloc, both with and without the port.
         no_proxy = no_proxy.split(',')
-        
+
         for host in no_proxy:
             if netloc.endswith(host) or netloc.split(':')[0].endswith(host):
                 # The URL does match something in no_proxy, so we don't want
                 # to apply the proxies on this URL.
                 return {}
-                
+
     # If the system proxy settings indicate that this URL should be bypassed,
     # don't proxy.
     if proxy_bypass(netloc):
@@ -414,7 +414,7 @@ def get_environ_proxies(url):
     # anywhere that no_proxy applies to, and the system settings don't require
     # bypassing the proxy for the current URL.
     return getproxies()
-    
+
 
 def default_user_agent():
     """Return a string representing the default user agent."""
@@ -546,3 +546,22 @@ def get_auth_from_url(url):
         return (parsed.username, parsed.password)
     else:
         return ('', '')
+
+
+def to_native_string(string, encoding='ascii'):
+    """
+    Given a string object, regardless of type, returns a representation of that
+    string in the native string type, encoding and decoding where necessary.
+    This assumes ASCII unless told otherwise.
+    """
+    out = None
+
+    if isinstance(string, builtin_str):
+        out = string
+    else:
+        if is_py2:
+            out = string.encode(encoding)
+        else:
+            out = string.decode(encoding)
+
+    return out

diff --git a/test_requests.py b/test_requests.py
index 27d5e67..ca05f4a 100755 (executable)
--- a/test_requests.py
+++ b/test_requests.py
@@ -199,13 +199,13 @@ class RequestsTestCase(unittest.TestCase):
         assert r.json()['cookies']['foo'] == 'bar'
         # Make sure the session cj is still the custom one
         assert s.cookies is cj
-    
+
     def test_requests_in_history_are_not_overridden(self):
         resp = requests.get(httpbin('redirect/3'))
         urls = [r.url for r in resp.history]
         req_urls = [r.request.url for r in resp.history]
         self.assertEquals(urls, req_urls)
-        
+
     def test_user_agent_transfers(self):
 
         heads = {
@@ -263,7 +263,7 @@ class RequestsTestCase(unittest.TestCase):
         self.assertEqual(r.status_code, 401)
 
         s = requests.session()
-        
+
         # Should use netrc and work.
         r = s.get(url)
         self.assertEqual(r.status_code, 200)
@@ -640,6 +640,16 @@ class RequestsTestCase(unittest.TestCase):
         r = requests.Request('GET', url).prepare()
         self.assertEqual(r.url, url)
 
+    def test_header_keys_are_native(self):
+        headers = {u'unicode': 'blah', 'byte'.encode('ascii'): 'blah'}
+        r = requests.Request('GET', httpbin('get'), headers=headers)
+        p = r.prepare()
+
+        # This is testing that they are builtin strings. A bit weird, but there
+        # we go.
+        self.assertTrue('unicode' in p.headers.keys())
+        self.assertTrue('byte' in p.headers.keys())
+
 
 class TestCaseInsensitiveDict(unittest.TestCase):