Firedrake meeting 2023 11 22 - firedrakeproject/firedrake GitHub Wiki

Date and time 2023-11-22 1600 GMT (1600 UTC)

Action Items

  1. Pick Chair and Minuter (RNH to pick minuter)
  2. ALL: (ongoing) triage the open issues and confirm if they are indeed still open (and perhaps provide labels)
  3. JB: Move pyop3 and FInAT to firedrakeproject
  4. ALL: do things with SV's branches
  5. DH: Email to Andreas to have 2 (+ others!!!) loopy PRs merged TODO: FIND OUT WHICH PRS THESE ARE
  6. DH: Get Firedrake a docker open source account (link here)
  7. DH: Talk to GregVernon about PR#2116.
  8. DH: Revisit PR#2484.
  9. DH: Order more Firedrake stickers
  10. ALL/ANY: Drop libsupermesh (minutes)? JB: Waiting for https://github.com/Toblerity/rtree/pull/292
  11. NB: Look for changes needed to manual to account for dual spaces

Minutes

Present: CW (minuter), DD, JB, KS, DH, RNH, PB, NB, JHC, IM, FA

Apologies:

JB: Pyadjoint Firedrake tests erroring

#3247

  • JB: PETSc errors are getting raised in PetscFinalize in the Firedrake pyadjoint tests.
  • JB: Solution should be able to catch this in future.
  • DD: Issue is somehow related to pytest-xdist (i.e. pytest -n ???).
  • Action point DD: Investigate further.

TB: What guarantees does Firedrake provide that checkpoints will survive firedrake-update?

  • DH: No guarantees are made for the ordering between checkpoint files and meshes. To make comparisons one needs to load the mesh from the checkpoint.
  • DH: Could possibly use interpolation to get this to work. One can take a spanning DG space and interpolate into it.
  • DH: If FE definitions changes (e.g. PB changing Lagrange) then these guarantees do not hold. One could add a check here where we do an interpolation and make sure that the interpolation matrix is the identity.

JB: Separate Matplotlib

This is PR 3117 below. CW suggests this is worth wider discussion.

  • CW: This is currently an API change since plotting functions would no longer be in the from firedrake import *.
  • DH: We should replace the moved functions in the global namespace with functions that raise a hard error and point users to the right function to call.
  • DH: We could follow the same approach for making VTK optional.
  • DH: Could have a policy that external dependencies should be truly optional.

JB: Public interface to GC

This is PR 3152 below. CW suggests this is worth wider discussion. JB thinks this should really only be a semi-public interface, excessive calls to this functionality will be disastrous for performance!

  • DH: Fine for GC routines to be out of the public namespace.
  • DH: garbage_cleanup should crash if a communicator is not found.
  • DH: Let's keep the atexit comm cleanup stuff. No real reason to get rid of it.

CW: Docs currently failing, what's the status of the fix?

See below.

DD: Users are having errors on installing Firedrake with Xcode15

and apparently, Xcode15 has sorted this issue.

See the PR 3249

DD: Documentation building is now working with the PR 3239.

JB and I were trying to move the documentation for numpydoc format.

Merged, build is fixed. The transition to numpydoc remains outstanding.

Merge PRs

Date of next meeting

1600 GMT (1600 UTC) 2023-11-29