Specialization of _map_getindex

[emmt]

In order to allow for accouting for uniform arrays (arrays whose entries all have the same values as provided by StructuredArrays) in mapreduce, I have specialized the _map_getindex method as follows:

using StructuredArrays, CUDA
@inline @Base.propagate_inbounds CUDA._map_getindex(args::Tuple{AbstractUniformArray,Vararg}, I) =
    (value(args[1]), _map_getindex(Base.tail(args), I)...)
@inline @Base.propagate_inbounds CUDA._map_getindex(args::Tuple{AbstractUniformArray}, I) =
    (value(args[1]),)

Using uniform arrays is a mean to spare allocating large arrays whose entries are all the same in contexts where broadcasting rules cannot be applied.

Although this hack seems to work, I am not confortable at specializing such a private function. Is there a better way to achieve this?

edit: Sorry, I forget to indicate that value(A) returns the value of any entry of a uniform array A.

[emmt]

I am a dummy, I have made my tests without the above specialization and it turns out that the specialization of getindex for uniform arrays is sufficient. Hence, the above hack was, in fact, unnecessary. I was trying to fix a problem that does not exist. Sorry for the disturbance.