Jishnu Bhattacharya

As suggested on [discourse](https://discourse.julialang.org/t/parametrically-constrained-arguments-not-working/52975/5), the solution might be to add a function `localindices` to `Base`, and have both `SharedArrays` and `DistributedArrays` add methods to that function.

Here is the line: https://github.com/JuliaParallel/DistributedArrays.jl/blob/ffb128bd4623cf6108594585be0d95b35d5773a6/src/darray.jl#L824

Tbh the name Neumann is a bit ambiguous here, due to the commonly used Neumann boundary condition in differential equations. Perhaps something like `NeumannSymmetry` might be less ambiguous?

Related: ```julia julia> F = FastTransforms.plan_chebyshevtransform(rand(BigFloat, 4)) FastTransforms.ChebyshevTransformPlan{BigFloat, 1, Nothing, false, 1, UnitRange{Int64}}(#undef) julia> F.plan ERROR: UndefRefError: access to undefined reference Stacktrace: [1] getproperty(x::FastTransforms.ChebyshevTransformPlan{BigFloat, 1, Nothing, false, 1, UnitRange{Int64}}, f::Symbol)...

Not directly, but this leads to downstream test failures if a package checks for zero ambiguities without specifically excluding dependencies. This isn't a big issue as such, but would be...

Slightly unrelated to the original post, but why does an inferred type differ between Cthulhu and `@code_warntype`? As an example: ```julia julia> using ApproxFunOrthogonalPolynomials julia> using Cthulhu julia> @descend_code_warntype TensorSpace((Chebyshev(0..1),...

Oops. Let's try again with a new version of `ApproxFunBase.jl` that includes the bugfix.

Are the curve tests supposed to be this resource intensive?

@stevengj @ararslan Could you have a look at this?

To me this seems to be a quality-of-life improvement, as it makes the output of `@code_warntype` significantly more readable on terminals with word wrap. Without this, the constant may span...