getcellset(::SubDofHandler)

[KnutAM]

Since we already have a function for getting the cellset from the grid,
I find it logical to also provide getcellset(::SubDofHandler), such that users
don't have to use direct field-access to obtain the SubDofHandler's cellset.

[codecov-commenter]

Codecov Report

All modified lines are covered by tests ✅

Comparison is base (32fbd6b) 92.77% compared to head (d5d0952) 92.77%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #809   +/-   ##
=======================================
  Coverage   92.77%   92.77%           
=======================================
  Files          33       33           
  Lines        4952     4953    +1     
=======================================
+ Hits         4594     4595    +1     
  Misses        358      358           

Files	Coverage Δ
ext/FerriteMetis.jl	100.00% <100.00%> (ø)
src/Dofs/ConstraintHandler.jl	95.98% <100.00%> (ø)
src/Dofs/DofHandler.jl	95.10% <100.00%> (+0.01%)	⬆️
src/Dofs/DofRenumbering.jl	91.30% <100.00%> (ø)
src/Dofs/apply_analytical.jl	100.00% <100.00%> (ø)
src/Dofs/sparsity_pattern.jl	100.00% <100.00%> (ø)
src/PointEval/PointEvalHandler.jl	92.81% <100.00%> (ø)
src/iterators.jl	94.69% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kimauth]

Any particular reason why you don't want to access the field itself? Since we don't have a "subdofhandler interface" I don't really see why this is problematic.
In case it's just because the field is technically internal, we could alternatively document that SubDofHandler has a cellset field, I don't think this is likely to change (similar to how the fields of CellCache are documented).

[KnutAM]

Documenting the field is also an option.

I find that an access function for use outside Ferrite.jl is more user-friendly, e.g. when writing an assembly loop for one subdofhandler. And since we already have that function, I find such a dispatch logical. But for sure this is just a matter of taste.

[termi-official]

This is a thing I would really like to discuss at FerriteCon, because not having an interface here causes some trouble downstream.

[termi-official]

The name suggests that a set is returned. Do we want to enforce this? I played a bit around with using ranges instead of sets in #625 which might be an efficient alternative for the case that subdofhandlers model physical subdomains.