Re: [netconf] Trouble with RFC 8040 (Restconf) fields Query Parameter

[Henning Rogge <[e-mail]>]

On Fri, May 8, 2020 at 3:11 PM Douglas Hubler <[e-mail]> wrote:
>
> I've been very interested in restconf test cases and glad to see some activity here.
>
> So is the idea restconf server implementations stand up a server that loads each config/state one-by-one then a utility will also read in same test cases, exercise queries and assert the expected results?

Yes.

> I'd definitely like to hear more about how you're using this currently and if you have a working model.

I use python with two threads... one thread starts the server and
loads the "before" state into the datamodel.
The other one waits 100ms, sends the HTTP request, parses the answer,
check the answers body and headers and then checks the (possibly)
changed internal state of the server.
Both before and after uses the standard Restconf JSON encoding.
After this the server thread is stopped and the client closes the connection.

This is iterated for each test.

At the moment the "/restconf" part of the URL is hardcoded... I will
definitely change this when I switch to the "unified data store".
I think I also have some tests where "before.json" and "after.json"
are the same, so I could replace them with a single "state.json" file
(easier to read).

The trickiest part was to write a compare function for the servers
response body and for the final server state (and setup and teardown
of the second thread ^^).

My tests use "pytest", so no special things involved. I am currently
trying to get permission to share the code of the server under LGPLv3,
same licence as the yangson library I use.
I have also written a couple of extensions for the yangson library
(e.g. XML input/output support), but I am trying to get them pulled
into the upstream implementation.

In theory the "write initial state" and "read final state" could be
done with Restconf HTTP commands, but I would like to keep each test
using only a single HTTP request.

Henning Rogge