OAuth 2 Tests depend on dict ordering about oauthlib HOT 6 CLOSED

Would it not be better to explicitly sort the keyword arguments?

--- a/oauthlib/oauth2/draft25/parameters.py
+++ b/oauthlib/oauth2/draft25/parameters.py
@@ -64,7 +64,7 @@
     if state:
         params.append(('state', state))

-    for k in kwargs:
+    for k in sorted(kwargs):
         if kwargs[k]:
             params.append((unicode_type(k), kwargs[k]))

@@ -98,7 +98,7 @@
     if 'scope' in kwargs:
         kwargs['scope'] = scope_to_string(kwargs['scope'])

-    for k in kwargs:
+    for k in sorted(kwargs):
         if kwargs[k]:
             params.append((unicode_type(k), kwargs[k]))

--- a/tests/oauth2/draft25/test_client.py
+++ b/tests/oauth2/draft25/test_client.py
@@ -294,7 +294,7 @@
     password = "bar"
     body = "not=empty"

-    body_up = "not=empty&grant_type=password&username=%s&password=%s" % (username, password)
+    body_up = "not=empty&grant_type=password&password=%s&username=%s" % (password, username)
     body_kwargs = body_up + "&some=providers&require=extra+arguments"

     token_json = ('{   "access_token":"2YotnFZFEjr1zCsicMWpAA",'
--- a/tests/oauth2/draft25/test_parameters.py
+++ b/tests/oauth2/draft25/test_parameters.py
@@ -56,7 +56,7 @@
         'username': 'johndoe',
         'password': 'A3ddj3w'
     }
-    password_body = 'grant_type=password&username=johndoe&password=A3ddj3w'
+    password_body = 'grant_type=password&password=A3ddj3w&username=johndoe'

     cred_grant = {'grant_type': 'client_credentials'}
     cred_body = 'grant_type=client_credentials'

from oauthlib.

Sorry, that patch isn't sufficient. I've uploaded a patch to Ubuntu and will do a pull request.

from oauthlib.

@warsaw - Thanks for the patch & #63!

OAuthLib aims to be very strictly spec compliant* and while sorting the parameters seem rather innocent it's not the expected behaviour. Furthermore I much rather improve the tests and make them robust than tweak the code to fit the tests. Adding an assert_url_equals method to TestCase is rather trivial (using urlparse for the heavy lifting).

For the two tuples I think TestCase.assertItemsEqual should be sufficient.

• if you find places where this could be improved upon please point it out =)

from oauthlib.

@warsaw I was not sure if you found any time to work more on the PR or not and since I found a free spot in my calendar and sorted this out. I've added you to AUTHORS, let me know if you'd prefer not to be listed.

Hopefully I covered most of it but will keep an eye out for flakyness.

I've pushed this to PYPI as 0.3.2.

Heard from @christofferholmstedt that you are working a bit with making OAuth requests and thought I'd give a shoutout to the awesome requests library that has OAuthLib powered OAuth support built in =)

Here's an example on doing HMAC-SHA1 signed requests but RSA and plaintext are of course also supported.
http://docs.python-requests.org/en/latest/user/quickstart/#oauth-authentication

from oauthlib.

On Nov 08, 2012, at 12:05 PM, Ib Lundgren wrote:

@warsaw I was not sure if you found any time to work more on the PR or not
and since I found a free spot in my calendar and sorted this out. I've added
you to AUTHORS, let me know if you'd prefer not to be listed.

Thanks so much! I was almost literally just getting back to this after UDS-R
and post-Copenhagen travel. You probably beat me by just a couple of hours. :)

Hopefully I covered most of it but will keep an eye out for flakyness.

I've pushed this to PYPI as 0.3.2.

Fantastic, thanks. I'll test the PyPI version. If it looks good, I'll go
ahead and upload to Ubuntu 13.04, otherwise will file bugs if I find any other
problems.

Heard from @christofferholmstedt that you are working a bit with making OAuth
requests and thought I'd give a shoutout to the awesome requests library that
has OAuthLib powered OAuth support built in =)

Indeed! We've upgraded urllib3 and requests in 13.04 to the latest PyPI
versions. One change we're making in the packaging though, is to use
chardet2, urllib3, and oauthlib from the Debian packages instead of the
bundled versions in requests (I wonder why they do that :/).

It all needs a bit of testing, but after I get everything uploaded properly,
I'll start converting some of the dependent packages, and continue testing.

Here are an example on doing HMAC-SHA1 signed requests but RSA and plaintext
is of course also supported.
http://docs.python-requests.org/en/latest/user/quickstart/#oauth-authentication

Thanks!

from oauthlib.

The reason requests vendors OAuthLib is because we are still in a flux with quite frequent changes. @kennethreitz is a very responsive and friendly guy so if the vendoring is a concern just ping him.

Although we are unlikely to break anything related to OAuth 1 clients most of the OAuth 2 API will likely undergo quite a bit of changes. One day (probably quite a few months from now) I envision hitting 1.0 at which point the API will be stable.

Feel free to send bugs, PRs and just about anything our way and we'll take a look at it =)

from oauthlib.