[Yue-Zhengyuan]

This PR exports module-specific symmetrize_operator functions to symmetrize custom operators.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/bosonoperators.jl	95.65% <100.00%> (+0.09%)	⬆️
src/fermionoperators.jl	95.12% <100.00%> (+0.12%)	⬆️
src/hubbardoperators.jl	98.88% <100.00%> (-0.01%)	⬇️
src/quantumgates.jl	100.00% <100.00%> (ø)
src/spinoperators.jl	91.75% <100.00%> (+0.17%)	⬆️
src/tjoperators.jl	97.95% <100.00%> (-0.03%)	⬇️
src/utils.jl	93.18% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lkdvos]

Trying to unpack this here, mostly to see if I'm correctly understanding the goals etc.

Am I correct in saying that this is effectively doing two things:

1. exposing Module._symmetrize_operator
2. accepting AbstractArray directly

For the first part, sounds good, although I might advocate to keep the name to be symmetrize_operator instead of custom, specifically for an exported function, and just making that exported.

I'm slightly less certain about the second part, as I guess that is more of a convenience layer on top and I tend to like steering people slightly into TensorKit, since realistically they will have to at some point anyways. If you really think this is necessary, I'd just use dispatch and define a small helper function that does TensorMap(A, *(ComplexSpace.(size.(Ref(A), 1:div(ndims(A), 2))...), *(ComplexSpace.(size.(Ref(A), div(ndims(A), 2)+1:ndims(A))...)), so both versions are supported?

This is also more or less why I don't really see much benefit in asserting equal sizes of the input array, as we probably error right after that in the symmetrization code anyways?

Otherwise definitely looks good, thanks for the addition!

[Yue-Zhengyuan]

Now I directly export symmetrize_operator, accepting an AbstractTensorMap as input. For fermionic systems the input should already have FermionParity grading (instead of being completely desymmetrized).

[lkdvos]

Except for removing the extra tests, looks great to me!
You might want to also immediately bump the version for a patch release?

[Yue-Zhengyuan]

I still want to take a closer look at the t-J case with slave_fermion = true and see if _maybe_slave_fermion can be optimized away.

[Yue-Zhengyuan]

Let's not release 0.3.1 immediately. I also want to export the function to add charges in a separate PR.

[lkdvos]

Apologies for the second round of review comments, feel free to ignore if you disagree, had some time to kill before a flight ;)

[Yue-Zhengyuan]

@lkdvos I skipped CI in the last commit since it affects documentation only. Can you help directly merge, or force start the tests?