Summary

• Fix _maybe_SArray functions in static_arrays.jl: The original code was trying to construct SVector{N}(x) and SArray{Tuple{Sz...}}(x) without element type parameters and without converting to Tuple, which caused MethodErrors. Now properly constructs SVector{N, T}(Tuple(x)) and SArray{Tuple{Sz...}, T, N, prod(Sz)}(Tuple(x)).
• Add JET.jl tests: Added JET.jl as a test dependency and created test/jet_tests.jl to test type stability of core ComponentArray operations (getdata, getaxes, getindex).

Background

Ran JET.jl analysis on the package which identified several issues. The most actionable one was the _maybe_SArray functions used by @static_unpack which had incorrect type signatures for constructing SVector and SArray types. The macro now correctly returns SVector and SMatrix types.

Before fix:

julia> @static_unpack a, b = ComponentArray(a = 5.0, b = [1.0, 2.0])
ERROR: MethodError: no method matching (SArray{Tuple{2}})(::SubArray{Float64, 1, Vector{Float64}, Tuple{UnitRange{Int64}}, true})

After fix:

julia> @static_unpack a, b = ComponentArray(a = 5.0, b = [1.0, 2.0])
julia> typeof(b)
SVector{2, Float64}

Test plan

• All existing tests pass (Pkg.test())
• New JET tests pass
• Manually tested @static_unpack with vectors and matrices

cc @ChrisRackauckas

🤖 Generated with Claude Code