Open Stack

Hi all,

So firstly, if you're an XML guru, I apologize, the questions below are
probably really basic, I always prefer JSON or YAML, because every time I
deal with XML, I get a week-long-headache ;)

So I'm writing Tempest API tests for the keystone OS-TRUST extension, as
was previously requested - it's going pretty well, & I'm finding some real
bugs.  Here's a WIP review:

https://review.openstack.org/#/c/54810/

However, I've got a bit stuck formatting the POST body for the trust create
in XML (all works fine via JSON).  The json body looks like:

{
    "trust": {
        "expires_at": "2013-02-27T18:30:59.999999Z",
        "impersonation": true,
        "project_id": "ddef321",
        "roles": [
            {
                "name": "member"
            }
        ],
        "trustee_user_id": "86c0d5",
        "trustor_user_id": "a0fdfd"
    }
}

And looking at other XML tests which format a POST body, they do something
like:

post_body = Element("trust",
                    xmlns=XMLNS,
                    trustor_user_id=trustor_user_id,
                    trustee_user_id=trustee_user_id,
                    project_id=project_id,
                    impersonation=impersonation,
                    expires_at=expires_at)

This gives me a post body which looks weird (but seems to work):

<trust impersonation="True"
xmlns="http://docs.openstack.org/identity/api/v3"
trustor_user_id="efc6504105c54fbe95928a51459d06c9" expires_at=""
trustee_user_id="f55efd1d617e4367891d202a811d7728"
project_id="b5d498f9631244b59912ce2a0025cf8d"/>


So my questions are:
1. Why do we create a single element like this, instead of appending
subelements so the XML body looks more like the JSON request?

2. If any elements have a None value, they are encoded as a zero-length
string, is that expected?

3. How do I go about encoding the list of roles, as in the sample request
(it's a list of dicts, where each dict has one key called "name")

Any help, review feedback or pointers to docs/examples would be hugely
appreciated!

Thanks,

Steve