Closed grigi closed 1 year ago
@grigi thanks for this contribution, really appreciate it, left a couple comments, happy to discuss
Hi, I removed green, used coverage directly instead.
The reason I committed tests/requirements.txt
was so that we won't get unexpected test failures during development, but only when on explicitly updates the test dependencies. Since you didn't want it I removed it and changed make up
to simulate a clean install and update packages eagerly. This way you should be able to keep the dev environment a little cleaner.
Turns out I was wrong, and redis==5.0.0
does break something. I backed off on allowing it to upgrade for now.
There is a few janitorial things I would like to discuss:
[dev]
dependencies into tests/requirements.txt
because it's not ideal to allow users to install library dev dependencies.format.sh
and run_checks.sh
as they can be replaced with relevant Makefile commands.
Hi, I've recently discovered this project and I really like it's focus and simplicity. Apologies if this description is a little too wordy.
This is an initial project organising PR.
What I've done here is:
Makefile
as a way to manage the project from the command line, currently only manages test dependency pinning, linting & testing. I find doingmake up
,make format
,make ci
etc... quite convenient.make up
will try to do a clean-install of dependencies.mypi.ini
intosetup.cfg
and also handle test/coverage settings in that fileredis==5.0.0
is not working perfectly.I was considering changing
format.sh
andrun_checks.sh
, but decided against it as I want your feedback first.Anything I did here is up for change.
Rambling
This is purely out of selfish reasons where I need a simple and reliable queue manager that can run asyncio code. All the established ones are sync (Celery/rq), or in a weird limbo (arq). This seems to fit my requirements quite well.
It would need a few more things such as :
readthedocs.io
, Sphinx (as you have relatively thorough docstrings) and piccolo-themecloser-to-exactly-once
orat-least-once-with-idempotency
(e.g. for better reliability, if you can declare a task as idempotent, you can er on the chance of accidentally running multiple times, but if not idempotent you really shouldn't)Less important, but useful:
I'll likely be working on a few of the above points in any case.