new version of the plugin (1.5) #8
Conversation
With the previous version of the plugin you created an API client object by passing provider_key AND the corresponding application keys. Now you only pass the provider_key and the application keys are passed when the method is invokated (authrep, authorize or report). That's how the Ruby or Java plugins work. It is backwards compatible with the older version. You can check it by running the old test suite (tests.py) over the new plugin version. Other changes: - new suite of tests to cover the new way of using the plugin (tests-v2.py) - added test coverage for the user_key auth pattern - new examples and extended documentation
| @@ -85,11 +88,10 @@ def testAuthRepWithInvalidMetric(self): | |||
| """test authrep API with invalid metric""" | |||
| authrep = self.ThreeScaleAuthRep(self.provider_key, self.app_id, self.app_key) | |||
|
|
|||
| self.assertFalse(authrep.authrep({"invalid_metric":1})) | |||
| self.assertFalse(authrep.authrep(usage={"invalid_metric":1})) #expects (None, None, {...}) or (usage={...}) | |||
There was a problem hiding this comment.
won't this break compatibility when upgrading? can't we do some type checking in the authrep method?
There was a problem hiding this comment.
you are right, but at that moment using named parameters was the only way I came up with to be able to pass the app keys in the authrep call (not in the client object) while maintaining compatibility with the old way. I will document that point a bit more. If it is ok, I will correct with smth more elegant it in the new update.
There was a problem hiding this comment.
there should be huge warning when upgrading that it breaks calls when reporting metrics in authrep call (and maybe others)
isn't there in python something like "key".isString? or checking for dict type or checking for number of params (1 - usage, 2 - app_key, app_id, but then what about user_key?)
There was a problem hiding this comment.
I'm trying now doing this check in authrep and seems to work so far:
if isinstance(app_id, dict):
usage = app_id
app_id = ""
|
good work @vdel26 ! |
With the previous version of the plugin, you created an API client object
by passing provider_key AND the corresponding application keys.
Now you only pass the provider_key when creating the API client object. Application authorization keys are passed
when the method is called (authrep, authorize or report). Therefore you don't need to create a new client for each incoming request. Also, that's how the Ruby or Java plugins work.
It is backwards compatible with the older version. You can check it by
running the old test suite (tests.py) over the new plugin version.
Other changes:
(tests-v2.py)