advertise ccolor in writing generic code

[johnnychen94]

Functions in JuliaImages should accept both numerical array AbstractArray{<:Number} and colorant array Abstract{<:Colorant} if possible.

It is not uncommon to infer the return eltype in the beginning of the algorithm so as to pre-allocate array. I previously have two versions of codes to achieve this. The first version is:

function foo(img::AbstractArray{T}) where T<:Number
    CT = RGB{floattype(T)}
    ...
end

function foo(img::AbstractArray{T}) where T<:Colorant
    CT = RGB{floattype(eltype(T))}
    ...
end

Then one day I found out that eltype(Float32) == Float32, which make it possible to merge it into one method:

function foo(img::AbstractArray{T}) where T<:Number
    CT = RGB{floattype(eltype(T))}
    ...
end

As @kimikage pointed it out in JuliaGraphics/ColorTypes.jl#201, ccolor does a similar thing more reliably (I assume so?):

function foo(img::AbstractArray{T}) where T<:Number
    CT = ccolor(RGB, floattype(T))
    ...
end

I feel this worth documenting somewhere in the tutorials; otherwise, it's like a missing spell and people repeatedly reinvent it.

[kimikage]

The ccolor is a convenient and powerful function for such a purpose. However, ccolor is not so "generic".

For example:

julia> ccolor(HSV, Gray{Float32})
HSV{Float32}

julia> ccolor(HSV, Float32) 
ERROR: in ccolor, no automatic conversion from Float32 and HSV

(This is not a bug.)

However, at least for gray and rgb colors, it should work as expected. So, I also feel ccolor is worth documenting.