The manifest in the example is gone:

http://mozilla.github.com/MarketplaceClientExample/manifest.webapp

I used http://pasjans-online.pl/manifest.webapp but I guess it needs to be replaced by one we own

Our automated tests, which make use of Marketplace.Python to create marketplace apps via the API, have started failing due to an error in the oauth2 package.

The error can be seen at https://webqa-ci.mozilla.com/view/Buildmaster/job/marketplace.stage.developer_hub.saucelabs/423/HTML_Report/

One solution to this would be to pin to an older version of oauth2, but the next previous version on PyPI is very old. In any case, something should be done to address this issue to allow for the use of the library.

Here is the docs:

https://zamboni.readthedocs.org/en/latest/topics/api.html#enabling-an-app

https://github.com/mozilla/Marketplace.Python/blob/master/marketplace/client.py#L5 points to 404.

The readme doesn't point out that by default the example will hit the dev server, not the stage server. I think it would be nice if we pointed this out, users might have to do:

export MARKETPLACE_DOMAIN=marketplace.allizom.org

If they'd like to hit stage.

I think the method should first get current data and then display them one by one and providing existing data as defaults
empty string - accept default, enter some data - override default.

Some one was trying to use the MarketplaceClientExample code and got a 401 error back. The 401 errors now contain the information about the error. It would be nice if we had a verbose mode option that showed the full request and response. So we don't have to do hack's like this:

http://amckay.pastebin.mozilla.org/1853517

Optionally, if the response is not a 200, just show the response.

I don't understand why we are splitting the scheme, location and port here.

I need to use urlparse currently to get the values I want:

url = urlparse.urlparse(self.root)
# and then url.hostname, url.port, url.scheme

I think it's easier to create an urlstring than to parse the bits we need out of one urlstring we already have.
Maybe I'm missing some use case here ? What do you think about changing the signature of this class to accept an urlstring?

I have been using this client to automate some work I want to do with marketplace, but I'm a bit surprised to have requests objects in response of my function calls.

For instance, the requests library changed recently and when I want to have a json version of the request, I need to do response.json() rather than response.json previously.

My point is that things can break easily on thelibrary we use and that we probably shouldn't expose this on the users of our library.

For instance, I would prefer to have the results of the APIs returned directly (as a python dict), and to have exceptions from this library if we don't have the behaviour we want on the server.

What do you think? I would be happy to help you work on the APIs we would like to have for this library.

When creating an app (via the command line), response.content seems to be a string, not a dict.

$ python main.py create 45a6aa470f524f6eb72d64f1fc5bb5a3                                  
Traceback (most recent call last):
  File "main.py", line 31, in <module>
    result = commands[args.method](auth, *args.attrs)
  File "/Users/fred/dev/MarketplaceClientExample/python/app/commands.py", line 38, in create
    'id: %d, slug: %s') % (response.content['id'],
TypeError: string indices must be integers, not str

Although its an example, there's useful stuff in that lib we could re-use. It would be much easier if that lib was re-usable and a setup.py, then I could install it.