Implementation of MultiFieldStyle in TransientMultiFieldFESpaces

[oriolcg]

Hi @JordiManyer and @amartinhuertas, I'm trying to extend the latest developments for MultiFieldFESpaces, e.g. MultiFieldStyle, to the transient case. I created a new branch for this development: https://github.com/gridap/Gridap.jl/tree/Transient_MultifieldStyle

When running the dedicated tests it seems that the matrices are well assembled, but the vector is not. I get a test fail in this line:

Gridap.jl/test/ODEsTests/TransientFEsTests/TransientBlockMultiFieldStyleTests.jl

Line 61 in 817cac1

@test b1 ≈ b1_blocks # Failing

I haven't gone through all the details of BlockMultiFieldStyle and I'm struggling to figure out what can be going wrong. If you have some time, could you take a look at this error and point me to the right direction on how to fix this?

[JordiManyer]

Hi @oriolcg , I am a bit swamped atm, but make some time next week to have a look. Hopefully it's nothing too serious.

[JordiManyer]

Hi @oriolcg , I've pushed some fixes for your branch. Two things were failing:

• First, there were dispatching issues. The assembler was not being created block-wise. This was due to the MultiFieldStyle not being correctly propagated from the transient FESpace to the single-time-point FESpaces spawned when calling Xb(t).

• You also had some errors in the tests, where the integration data was not being created with the block-wise FESpaces. The FESpaces with which you integrate determine the assembly indices, which are different for BlockMultiFieldStyle (since we do not need to add offsets to the sub-space dof ids).

Everything should be working now. Let me know if you encounter any more issues.

[oriolcg]

Hi @JordiManyer, now it works. Thanks! Do you think we can merge these changes as they are?

[JordiManyer]

Hi @oriolcg , let's do a PR. We'll have to edit the NEWS.md and clean up the tests, but overall everything look ok to me.

[oriolcg]

Done in #946