On 21/10/2020 10.42, Tomasz Urbaszek wrote:
Yesterday I took a deeper look at the api. And if I correctly
understand, currently we are generating open API spec from python
code.
If we would like to preserve it, I would like to suggest considering
the FastAPI framework https://fastapi.tiangolo.com . It has many
advantages including using native type hints from python 3 for data
validation as well as generating OpenAPI spec.
mhmm, I was even hoping to go as far as py3.9 for annotations to help
"documenting as you code" :) I am very much for a type hint solution,
especially as we could use mypy for tests then.
Tomek
On Mon, Oct 19, 2020 at 9:02 PM Tomasz Urbaszek <[email protected]> wrote:
Thanks for the replies!
I'm not too concerned about breaking API, we _could_ make the next one a
version 2 with a version 2 API, a clean break from version 1
My main point in preserving the compatibiliy is to be sure that the ui
will work as expected. This will also limit the scope of the changes.
Whichever framework we pick, I would prefer if we could auto-generate the
JSON/GraphQL API from the source.
+1 for automating as much as possible, probably I would lean to first
do the JSON as it is now and the consider adding GraphQL.
If there is no objections to merging scanners with server I will do
this as a first step. Here is a Github issue so we can manage all the
work:
https://github.com/apache/kibble/issues/64
As per the ideas - I created issues to keep it for future reference :)
Bests,
Tomek
On Mon, Oct 19, 2020 at 7:54 PM Sharan Foga <[email protected]> wrote:
Hi Tomek
Thanks for the initiative and ideas!
Anything that makes things simpler or easier is OK with me :-) and
consolidating things seems like a good idea. I am guessing that a lot of the
changes would probably be behind the scenes rather than at the user end - right?
On the wishlist - I saw a talk by Myrle at ApacheCon@Home and she had done some
initial work linking contributions to affiliation and I mentioned seeing if
that could be something we could look at including into Kibble. I think we
tried to do it with Meta Pony Factor but dont think it is complete. Anyway lots
of ideas to keep us going!
Thanks
Sharan
On 2020/10/18 14:28:33, Tomasz Urbaszek <[email protected]> wrote:
Hello all,
I would like to propose a few things that I think are worth considering:
1. Merge all three repositories (kibble, kibble-scanners,
kibble-docker) into one. In this way I think we may simplify
development, dependency management, and distribution. Our project is
rather small and increasing places needing management probably won't
help.
2. Once we have the API server and scanner in one repo we can make
Kibble a python package. This will not only simplify installation and
setup but also may help us create a nice CLI to manage all components
(setup, server, scanners). By doing this users will be able to run
something like `kibble server` or `kibble scanners ponymail`.
3. Rewrite kibble API server to use a framework (for example FastAPI
which supports OpenAPI specs). This point is probably the hardest one
to implement as it will require adding tests to preserve compatibility
with kibble ui.
What do you think? Do you have anything that you would love to see in
Apache Kibble? :)
Cheers,
Tomek
--
Tomasz Urbaszek
Polidea | Software Engineer
M: +48 505 628 493
E: [email protected]
Unique Tech
Check out our projects!
