[leburgel]

The Dynkin labels are unbounded, so findindex and getindex now pass TensorKitSectors.Unbounded() sizes to the Manhattan indexing (QuantumKitHub/TensorKitSectors.jl#117, in 0.3.11), which leaves out the bound corrections at compile time. Same indices; resolves the TODO in sector.jl.

	findindex before	after	getindex before	after
SU(3)	7–8 ns	6 ns	172–177 ns	165–182 ns
SU(4)	26–27 ns	7 ns	303–315 ns	170–178 ns
SU(5)	61–69 ns	11 ns	411–428 ns	170–183 ns

Reproducer

using SUNRepresentations, TensorKitSectors, BenchmarkTools
fi(s) = TensorKitSectors.findindex(values(typeof(s)), s)
gi(I, i) = values(I)[i]
for s in (Irrep[SU₃]((7, 2, 0)), Irrep[SU₄]((6, 4, 3, 0)), Irrep[SU₅]((4, 3, 3, 1, 0)))
    @btime fi($s)
    @btime gi($(typeof(s)), $(fi(s)))
end

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/sector.jl	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[lkdvos]

Is this something we can add to tensorkitsectors such that all infinite combinations immediately get this?

[leburgel]

I tried this first, but it didn't seem to work. I didn't check in detail, but I think the conclusion was that the current approach in TensorKitSectors has a loop of fixed length which the compilers makes really fast, while porting this approach there gave an indefinite loop length which actually made it slower.

[lkdvos]

I wonder if it is just a matter of replacing the typemax of the integer with something static so the compiler can unroll again. If it doesn't immediately work I'm fine with this here as well, in principle these functions aren't marked public in TKS so not depending on that isn't so bad

[leburgel]

I wonder if it is just a matter of replacing the typemax of the integer with something static so the compiler can unroll again.

Apparently yes, this seems to work perfectly. Even better than the closed form here for some reason. I'll clean up the quick test and file this in TensorKitSectors.

[leburgel]

Done in QuantumKitHub/TensorKitSectors.jl#117, which also covers getindex. Once that's released, this PR reduces to passing TensorKitSectors.Unbounded() (a local commit replacing the closed form is ready); converted to draft until then.

[leburgel]

Updated after QuantumKitHub/TensorKitSectors.jl#117 (released in 0.3.11): the closed form is replaced by passing TensorKitSectors.Unbounded() sizes, which also speeds up getindex. Ready for review.