Firedrake meeting 2022 11 09

Date and time 2022-11-09 16:00UTC

Action Items

Pick Chair and Minuter (CW to pick minuter)
ALL: (ongoing) triage the open issues and confirm if they are indeed still open (and perhaps provide labels)
DH: Organise Firedrake 2022 (Abstracts open)
ALL: Rename split (.split -> .subfunctions) (ongoing)
JB: A Firedrake manual (ongoing)
JB: Python 3.11 (ongoing)
All: Go through TSFC PRs
RNH: check if issue #2516 is still a problem, if yes, see if #2509 Try installing pre-release of VTK from PyPI (supporting Python 3.10) fixes

Agenda

Present: RK RNH KS CW JB DH

Apologies:

DH: Firedrake 22

Please submit abstracts and register.

Actionables:

Firedrake 22: Please submit abstracts and register.
.split -> .subfunctions to become an issue.
Python 3.11: Still need macos VTK wheels, Koki to (hopefully) make intel macos wheel
- Mac mini test machines will be up and running soon, just need appropriate admin rights

JB: Black

(Low priority) We have had the conversation before in the past: Should we use black to autoformat code? Some formatting is inconsistent/bad. I have used black in a recent PR to cleanup utility_meshes.py.

We will see if we can make it work, including adding "has this been run through black" to the linter

JB: PyOP2

(Low priority) This PR changes testing from Python 3.6 to Python 3.7-3.11, is the matrix worth it? Should PyOP2 use our Firedrake runners? Should we move PyOP2 to Firedrake organisation?

Python change could cause issues with ubuntu's default python: we should choose not to care! We will instead follow python's support (which is 5 years, i.e. back to 3.7)
- We will keep f-strings out of the install script to make sure users of old python get appropriate warnings
Yes PyOP2 should use our runners
Fork PyOP2 and FInAT into firedrake organisation and ~~tombstone the old ones~~ (will need to update the documentation)
- Move them instead AS LONG AS urls resolve to the new location (JB to check before moving)