[Hawkular-dev] [Metrics] Python client and perspective from client developing..

[Michael Burman]

Hi,

As we need clients to actually push data to our system (and that way to 
get users & community), I've written one for the Python. I've tried to 
follow what a normal contributor would do and just read the 
documentation that's available. Short story - there isn't enough 
documentation.

I don't think we reached a consensus yet where to push the client source 
codes, so I haven't created a pull request for this and for that reason 
haven't renamed the repository either. It is available from 
https://github.com/burmanm/rhq-metrics-python-client with documentation. 
I'll gladly take ideas, my Python most likely looks like some other 
language written in Python syntax - but it works.

There are tests that trigger against the real hawkular-metrics server 
(0.30.0-SNAPSHOT). Usually the Python client developers target a mocked 
endpoint, but I decided to go after the real server as I believe our API 
might still change and the documentation isn't a reliable reference 
point. I tried to follow the documentation on 
https://github.com/hawkular/hawkular-metrics/tree/master/api/metrics-api-jaxrs 
but I had to revert in some cases to reading the source code on what was 
really wanted, especially after receiving 400 / 500 error codes while 
doing the testing (when the json looked like in the examples). I also 
only created abilities that are defined on that page, as I have no idea 
if rest of the undocumented REST APIs are actually there to stay or to 
be removed.

Which brings me to the question, what parts of the API are stable at the 
moment? There's counters and such which are not documented at all. We 
have huge selection of endpoints for any client developer to implement, 
perhaps too many of them. I can try to improve documentation, but even 
then I would want to know what's going on. I heard alerts is getting 
swagger-annotations, which is nice.

   -  Micke