Add Python 3.11 to GH Actions #1341
Conversation
|
Just FYI there are a bunch of issues with the docs requirements that need updating. I can update all the packages to 3.11-compliant versions and hope that old bugs have been fixed that requires the version pin? Or we could leave the GH action to build docs with python 3.9 like it does now. But someone wouldn't be able to build them locally if they were using 3.11. Thoughts @john-science? |
|
We can just continue to build the docs on GitHub with Python 3.9, and get the tests and downstream simulations up to 3.11. Right? And transfer the docs later? They are fairly independent. |
Yes I think it's fine to do that on CI. Someone building the docs locally will need to have a separate 3.9 venv, though. I've actually never built ARMI's locally I'm ashamed to say. |
…uirements-docs.txt later on
|
@john-science I have a few questions above. I can't test mpi4py update with our internal code for a week or three but I'm worried about pinned version dependency clashes, so I'm proposing removing the version pin altogether (or could update downstream and test it on python 3.9). |
|
I'm going to mark this as draft for now, until we resolve some stuff internally. I don't want to break things with new version pins. |
|
@opotowsky @ntouran Arrielle, it would be nice to allow people to check out ARMI and run it on Python 3.11. Can we make this two PRs?
|
Among others, I reverted the |
| @@ -1,7 +1,4 @@ | |||
| numpy>=1.21,<=1.23.5 | |||
| ipython>=7.34.0,<=8.12.0 | |||
|
|
|||
| # docutils 0.17 introduced a bug that prevents bullets from rendering | |||
| # see https://github.com/terrapower/armi/issues/274 | |||
| docutils <0.17 | |||
There was a problem hiding this comment.
This version pin is incompatible with python 3.11. It could stay removed and I could instead update the requirement in requirements-docs.txt (which, oddly, it has docutils <0.18 instead)
Whatcha think @john-science?
There was a problem hiding this comment.
That's just for docs, right?
Or is it used outside the docs?
There was a problem hiding this comment.
It's used in utils:
But I don't think bullet point rendering matters there ;-)
It needs to be unpinned here/in setup.py to be python 3.11 compatible.
Actually we have mpi4py version pinned in-house, so removing the version pin here won't change anything. |
|
@john-science @ntouran we have py311 success! |
…id-refactor-pre-1278 * terrapower/main: Move max assembly number out of global scope (terrapower#1383) Ensuring we throw an error on bad YAML files (terrapower#1358) Removing unused settings (terrapower#1393) Fixing some miscellaneous TODOs (terrapower#1392) Removing global plugin flags (terrapower#1388) Update reactivity coefficient parameters (terrapower#1355) Fixing ruff formatting of doc gallery examples (terrapower#1389) Fixing broken link (terrapower#1390) Removing unreachable code block (terrapower#1391) Removing unnecessary f-strings (terrapower#1387) Updating documentation dependencies for Python 3.11 (terrapower#1378) Adding a little code coverage to the CLIs (terrapower#1371) Remove coveragepy-lcov dependency during CI. (terrapower#1382) Reconnect logger after disconnecting in test. (terrapower#1381) Add Python 3.11 to GH Actions (terrapower#1341) Removing global variable from runLog.py (terrapower#1370) Moving the First-Time Contributors guide to the docs (terrapower#1368)
What is the change?
Updating ARMI to run CI with python 3.11.
Note, however, the docs GH action is remaining behind with python 3.9.
Why is the change being made?
Python 3.11 offers a 10-40% speedup! Let's get with the program.
Resolves Issue #957, but Opens Issues #1344 and #1377
Checklist
Tests have been added/updated to verify that the new/changed code works.doc/release/0.X.rst) are up-to-date with any important changes.docfolder.setup.py.