My thinking was that this would save allocating an array here. Although for findall in particular, it might not actually be faster?
julia> @btime findfirst(x->!x, r) setup=(r=rand(Bool, 10^6)); # same as findfirst(!, r)
2.283 ns (0 allocations: 0 bytes)
julia> @btime findfirst(.!r) setup=(r=rand(Bool, 10^6));
137.891 μs (4 allocations: 126.42 KiB)
julia> @btime count(x->!x, r) setup=(r=rand(Bool, 10^6));
80.810 μs (0 allocations: 0 bytes)
julia> @btime count(.!r) setup=(r=rand(Bool, 10^6));
142.010 μs (4 allocations: 126.42 KiB)
julia> @btime findall(x->!x, r) setup=(r=rand(Bool, 10^6));
671.744 μs (8 allocations: 3.92 MiB)
julia> @btime findall(.!r) setup=(r=rand(Bool, 10^6));
668.053 μs (6 allocations: 3.93 MiB)
Edit – that’s because the second-last line explicitly calls the last here, findall(testf::Function, A::AbstractArray) = findall(testf.(A)), so there can’t be a difference. Sorry about the noise!
