This and py4cl2

Question

This and py4cl2

digikar99 opened this issue 4 years ago · comments

In reference to https://github.com/digikar99/py4cl2, how do we go about the upgradation - or do we keep things separate?

I hope to find time to backport some more of rest of the things worked on in the past few months - those ones should be non-breaking.

However, as far as I understand, argument imports and writing to stdout from separate threads could break a good amount of things with users using existing py4cl.

Do we still procede with deprecating the symbols and then push users to upgrade, or any other suggestions?

Ben Dudson · Answer 1 · Wed Jul 15 2020 18:28:38 GMT+0800 (China Standard Time)

You've made some really nice improvements, and put a lot of good work into pycl2. I haven't had a lot of time to work on this recently, unfortunately. I'm not sure what the best thing to do is, but I think my priorities are to:

Avoid duplicating or fragmenting effort; the community is already small enough
Avoid breaking people's code

The easiest way I can see to achieve this is to:

Focus effort on pushing forward py4cl2; I'm happy to contribute to that for future updates
Keep the current py4cl interface, but make it clear that this repo is in maintenance mode: I could add a notice pointing to py4cl2, and saying that I'll accept bug fixes and non-breaking minor updates into py4cl, but that people looking to help should base their work on py4cl2.
Does that sound reasonable? It seems that you might have more time/motivation for maintaining this than I currently do.

Shubhamkar Ayare · Answer 2 · Thu Jul 16 2020 23:22:51 GMT+0800 (China Standard Time)

Okay with that; I'll get back some time this month. Also, zach suggested to use a different name - do we keep it as it is (py4cl2)?

Ben Dudson · Answer 3 · Thu Jul 16 2020 23:47:16 GMT+0800 (China Standard Time)

I think keeping it as py4cl2 would be good, as a major version change from py4cl.

Shubhamkar Ayare · Answer 4 · Sat Jul 25 2020 04:27:38 GMT+0800 (China Standard Time)

Currently on the list include:

python-process-startup-error
test and if required fix large numpy array transfer bug
embedding py4cl.py into the lisp image: requires removing all the single-quotes in "py4cl.py"
look into importing submodules and may be provide a continue-ignoring-errors option as in py4cl2

Another change is about preserving the array types between python (numpy) and common lisp - not sure if this can break people's code; I think no.

Shubhamkar Ayare · Answer 5 · Sat Jul 25 2020 04:36:04 GMT+0800 (China Standard Time)

Also, could you elaborate on async functions? I mean, one could simply use a separate thread, no?

rpgoldman · Answer 6 · Wed Jan 20 2021 23:47:35 GMT+0800 (China Standard Time)

As far as I can tell, only Py4CL, and not Py4CL2, are installable with Quicklisp.

Shubhamkar Ayare · Answer 7 · Thu Jan 21 2021 00:03:10 GMT+0800 (China Standard Time)

Erm, both should be available in recent quicklisp dists. Did you try quickloading after (ql:update-dist "quicklisp")?

rpgoldman · Answer 8 · Thu Jan 21 2021 02:12:54 GMT+0800 (China Standard Time)

@digikar99 Got it, thanks! Forgot to update.

Jin · Answer 9 · Mon Nov 14 2022 21:15:16 GMT+0800 (China Standard Time)

I wonder what the current status is. Will we merge py4cl2 back into py4cl in near future?

Ben Dudson · Answer 10 · Tue Nov 15 2022 00:25:15 GMT+0800 (China Standard Time)

Hi @jcguu95 My understanding is that py4cl2 is unlikely to be merged back into py4cl:
py4cl2 is a major version that changes some of the API.
In the interests of stability py4cl is in "maintenance mode". I'm happy to merge bug fixes, but encourage anyone interested in contributing to please base their improvements on py4cl2.