Revert "Compile AgML under python 3"

[klendathu2k]

Reverts #646

This remains a draft PR. It modifies AgML to use python 3 syntax which potentially introduces breaking changes if it is used with python 2.7.

[genevb]

I'm confused. This PR says it "modifies AgML to use python 3 syntax", implying it is intended to allow AgML to work with python 3. This PR also states that it "reverts #646 ", itself a PR which says it is "for compiling AgML using python 3". Is this PR then an alternative approach to getting AgML to work with python 3 in some way?

I understand that python does allow for version-checking at run-time, e.g. sys.version_info >= (3,0) or something close to that. So unless there is a syntax change that prevents the compilation before running, one could perhaps leverage that for version dependencies?

[klendathu2k]

I did not do any version checking to handle the 2.7 --> 3.0 changes. So once we merge PR #646, AgML is guaranteed to run correctly only under python 3+. I cannot guarantee that the geometries will be correct if the code in PR #646 is run with python 2.7. This is why it is a draft PR.

I'm surprised that (1) people found / reviewed the code, (2) the code doesn't crash in the CI tests and (3) that github even lets us merge a draft.

This PR will undo the change. The code will revert back to the python 2.7 syntax, and will be consistent with the build system on SL7.

Alternatively, we could switch to python 3 in our production environment. This might be a better option.

[genevb]

Just a note on why I looked at this PR: github sent me this note at 4:32pm: ***@***.*** requested your review on: #726 Revert "Compile AgML under python 3" as a code owner.'

…

-Gene

On Jan 22, 2025, at 7:08 PM, klendathu2k ***@***.***> wrote: I'm surprised that (1) people found / reviewed the code,

[genevb]

Notes on the available python versions I see at SDCC:

SL7 interactive nodes:
2.7.18 <= Will allow us to re-compile existing libraries, and currently to what "python" soft-links
3.6.8 <= Will allow us to move forward with py3 on SL7
3.8.13 <= Currently to what "python3" soft-links

SL7 container on Alma9 interactive nodes:
2.7.5 <= Will allow us to re-compile existing libraries, and currently to what "python" soft-links
3.6.8 <= Will allow us to move forward with py3 on SL7, and currently to what "python3" soft links

Alma9 interactive nodes:
3.9.16 <= Once we know py3 works on SL7, we can begin looking at compiling natively on Alma9

Hopefully no surprises between the sub-versions of py2 and py3...

-Gene

[klendathu2k]

None are really expected. Looking back at the release notes for versions 3.7 to 3.9, I don't see any backwards incompatible changes listed that worry me. That said, I'll place it in my TODO pipeline to run agml with the -W flag, which warns of any deprecated features.

…

On 2025-01-23 15:49, Gene Van Buren wrote: Notes on the available python versions I see at SDCC: SL7 interactive nodes: 2.7.18 <= Will allow us to re-compile existing libraries, and currently to what "python" soft-links 3.6.8 <= Will allow us to move forward with py3 on SL7 3.8.13 <= Currently to what "python3" soft-links SL7 container on Alma9 interactive nodes: 2.7.5 <= Will allow us to re-compile existing libraries, and currently to what "python" soft-links 3.6.8 <= Will allow us to move forward with py3 on SL7, and currently to what "python3" soft links Alma 9 interactive nodes: 3.9.16 <= Once we know py3 works on SL7, we can begin looking at compiling natively on Alma9 Hopefully no surprises between the sub-versions of py2 and py3... -Gene -- Reply to this email directly, view it on GitHub [1], or unsubscribe [2]. You are receiving this because you authored the thread.Message ID: ***@***.***> Links: ------ [1] #726 (comment) [2] https://github.com/notifications/unsubscribe-auth/ANL4LVHKI4P57NDXSDV2FWL2MFIWNAVCNFSM6AAAAABVV4X366VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMJQHE4DOOBUHE