ConvertEltype(Gray) not working as expect anymore

[Evizero]

It seems that for some reason or another, leaving the inner type open for Gray now causes fatal issues with inference. Will either have to force user to be specific with ConvertEltype(Gray{INNER}) or see if I can fix this.

[Evizero]

Seems to only be an issue if followed by a SplitChannel(). I suspect the issue is somewhere upstream with channelview

[johnnychen94]

Augmentor.jl/src/operations/convert.jl

Lines 54 to 60 in 6055db7

	function applyeager(op::ConvertEltype{T}, img::AbstractArray, param) where T
	maybe_copy(convert(AbstractArray{T}, img))
	end
	function applylazy(op::ConvertEltype{T}, img::AbstractArray, param) where T
	mappedarray(c->convert(T,c), img)
	end

A simple decision could be: if storage type T here is Any, then use eltype(eltype(img)) as the output type. Of course we need to deal with cases for color types that don't support N0f8 or Bool (e.g., Lab)

Update:

This would cause type-instability so it doesn't work.

[johnnychen94]

causes fatal issues with inference

I guess it's no longer an issue now. @Evizero does this look good to you?

julia> @inferred(augment(img, ConvertEltype(Gray))) |> summary # Gray{Any}
"300×400 Array{Gray{Any},2} with eltype Gray"

julia> pl = SplitChannels() |> ConvertEltype(Gray)
2-step Augmentor.ImmutablePipeline:
 1.) Split colorant into its color channels
 2.) Convert eltype to Gray{Any}

julia> @inferred(augment(img, pl)) |> summary # Gray{N0f8}
"4×300×400 Array{Gray{N0f8},3} with eltype Gray{Normed{UInt8,8}}"

[timholy]

Just a brief comment, instead of mappedarray(c->convert(T,c), img) it's slightly better to use of_eltype(T, img) because it also defines the inverse operation and thus supports setindex!. https://github.com/JuliaArrays/MappedArrays.jl#of_eltype

[barucden]

I confirm that the issue is gone. See #95, which implements a few test cases for this.