Like method for default zeros, ones, rand methods

[aminya]

Like method is added for manually defined methods, but still needs to be added for default zeros, ones, rand methods

Different tasks of this to do plan:

1. Ask in Discourse forum or Slack group for ideas

Achievement check:

having both:

zerosM(Int32, 2, 2)
zerosM(2, 2, like=ones(Int32,2,2) )

[aminya]

Here's what I do for copying all the methods

 zerosM(args...) = zeros(args...) 

Other than editing Julia Base, one way which seems that may work is to exclude args to not be a type.
Do we have something like !:T?
This is my crazy prototype (doesn't work obviously):

zerosM(args::T...; like::AbstractArray) where {T!::Type{T1}} where {T1} = zeros(eltype(like), args...) 

[aminya]

julia> fun(x::Integer) = 1
fun (generic function with 1 method)

julia> fun(x::Integer; y::Array) = 1
fun (generic function with 1 method)

julia> fun(x::Integer, y::Array) = 1
fun (generic function with 2 methods)

The 2nd definition doesn't create another method
I know that I can make keyword optional by

julia> fun(x::Integer; y::Array= nothing) = 1
fun (generic function with 1 method)

But in my situation, I am expanding another function's methods, and the function has fun(x::Integer) = 1 originally, and I don't want to overwrite that method.

However, the keyword arguments never participate in dispatch, and so there is no way other than redefining the original methods.

JuliaLang/julia#9498

https://discourse.julialang.org/t/rules-for-method-dispatch-with-keyword-arguments/23626/3

[aminya]

For other cases, probably it is better to not use a keyword argument and simply pass the Array as the first argument to the function.

I will redefine the methods using metaprogramming.