Avoid row promotion in heterogeneous arrays

[pabloferz]

Fixes #23994

This is the shameless obvious fix to the issue. I'll leave making hvcat more efficient (including optimizing based on inference when possible) for another occasion.

[pabloferz]

Of course the obvious fix doesn't work, so I went the second obvious fix.

Now the problem is that if I leave the methods definitions within the @testset it segfaults during testing. @JeffBezanson Is there a problem using the eltype of the Vararg arguments as in here?

EDIT: I was guessing it wouldn't be the case as I don't think some one is going to hvcat a large number of arguments by hand.

[pabloferz]

@nanosoldier runbenchmarks(ALL, vs=":master")

[mbauman]

I believe this will fail with

julia> [[1 2]; 3 4]
2×2 Array{Int64,2}:
 1  2
 3  4

I'm not as familiar with this code; can we just do:

hvcat(rows::Tuple{Vararg{Int}}, xs::Number...) = typed_hvcat(promote_typeof(xs...), rows, xs...)
hvcat(rows::Tuple{Vararg{Int}}, as...) = typed_hvcat(promote_eltypeof(as...), rows, as...)

[pabloferz]

can we just do:

Yes, I believe something along those lines should work. Let me try again.

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[pabloferz]

@nanosoldier runbenchmarks(ALL, vs=":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[pabloferz]

Nanosoldier regressions seem noise, but just for kicks: @nanosoldier runbenchmarks("array" || "linalg", vs = ":master")

[nanosoldier]

Something went wrong when running your job:

NanosoldierError: failed to run benchmarks against primary commit: stored type BenchmarkTools.ParametersPreV006 does not match currently loaded type

Logs and partial data can be found here
cc @ararslan

[kshyatt]

Looks like nanosoldier is down for a bit - do we want to merge anyway?

[ararslan]

Someone could manually verify any slowdowns that Nanosoldier noted on its last successful run. Nanosoldier is indeed down for the count due to an outstanding issue with JLD.

[pabloferz]

@nanosoldier runbenchmarks("array" || "linalg", vs = ":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[pabloferz]

I believe the regressions are the same noisy ones observed all over the place. This should be good to go.