Type constraints in Distributions and StatsFuns

Tooling
12

Alright, so I went ahead and implemented the requested feature, which works on master branch

julia> Reduce.@subtype FakeReal <: Real

julia> FakeReal(R"x+1") + FakeReal(R"y")
y + 1 + x

the implementation looks like this at the moment

macro subtype(x)
    x.head ≠ :<: && throw(error("$x is not a subtype expression"))
    name = x.args[1]
    Expr(:struct,false,x,Expr(:block,Expr(:(::), :r, :RExpr))) |> eval
    @eval begin
        export $name
        show(io::IO, r::$name) = show(io,r.r)
        extract(r::$name) = r.r
        Algebra.init_subtype($name)
    end
    nothing
end

and

function init_subtype(name)
    Expr(:block,[:(Base.$i(r::$name...)=$i(extract.(r)...)|>$name) for i ∈ [alg;iops]]...) |> eval
    Expr(:block,[:($i(r::$name...)=$i(extract.(r)...)|>$name) for i ∈ [calculus;cnan;cmat]]...) |> eval
    Expr(:block,[:(Base.$i(r::$name)=$i(extract(r))|>$name) for i ∈ [sbas;[:length]]]...) |> eval
    Expr(:block,[:($i(r::$name)=$i(extract(r))|>$name) for i ∈ [sfun;snan;snum;scom;sint;sran;smat]]...) |> eval
end

this provides all of the basic functions from the Reduce.Algebra module, but outside that, the new subtypes generated do not have the same treatment ast RExpr at the moment