extending methods in a loop over functions results in misleading method table printing

[tkelman]

This seems to be just a printing issue, but there might be something else going on here. Looks more or less the same on latest master too.

  | | |_| | | | (_| |  |  Version 0.3.6 (2015-01-08 22:33 UTC)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org release
|__/                   |  x86_64-linux-gnu

julia> type newtype
       end
       for f in (cos, sin, tan)
           f(x::newtype) = println("$f")
       end

julia> methods(cos)
#12 methods for generic function "cos":
cos(a::Complex{Float16}) at float16.jl:141
cos(z::Complex{T<:Real}) at complex.jl:528
cos(x::Float64) at math.jl:122
cos(x::Float32) at math.jl:123
cos(a::Float16) at float16.jl:140
cos(x::BigFloat) at mpfr.jl:488
cos(x::Real) at math.jl:124
cos{Tv}(A::SparseMatrixCSC{Tv,Ti<:Integer}) at sparse/sparsematrix.jl:512
cos{T<:Number}(::AbstractArray{T<:Number,1}) at operators.jl:359
cos{T<:Number}(::AbstractArray{T<:Number,2}) at operators.jl:360
cos{T<:Number}(::AbstractArray{T<:Number,N}) at operators.jl:362
tan(x::newtype) at none:2

julia> methods(sin)
#12 methods for generic function "sin":
sin(a::Complex{Float16}) at float16.jl:141
sin(z::Complex{T<:Real}) at complex.jl:518
sin(x::Float64) at math.jl:122
sin(x::Float32) at math.jl:123
sin(a::Float16) at float16.jl:140
sin(x::BigFloat) at mpfr.jl:488
sin(x::Real) at math.jl:124
sin{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}) at sparse/sparsematrix.jl:448
sin{T<:Number}(::AbstractArray{T<:Number,1}) at operators.jl:359
sin{T<:Number}(::AbstractArray{T<:Number,2}) at operators.jl:360
sin{T<:Number}(::AbstractArray{T<:Number,N}) at operators.jl:362
tan(x::newtype) at none:2

julia> methods(tan)
#12 methods for generic function "tan":
tan(a::Complex{Float16}) at float16.jl:141
tan(z::Complex{T<:Real}) at complex.jl:544
tan(x::Float64) at math.jl:122
tan(x::Float32) at math.jl:123
tan(a::Float16) at float16.jl:140
tan(x::BigFloat) at mpfr.jl:488
tan(x::Real) at math.jl:124
tan{Tv,Ti}(A::SparseMatrixCSC{Tv,Ti}) at sparse/sparsematrix.jl:448
tan{T<:Number}(::AbstractArray{T<:Number,1}) at operators.jl:359
tan{T<:Number}(::AbstractArray{T<:Number,2}) at operators.jl:360
tan{T<:Number}(::AbstractArray{T<:Number,N}) at operators.jl:362
tan(x::newtype) at none:2

julia> sin(newtype())
sin

julia> cos(newtype())
cos

julia> tan(newtype())
tan

[JeffBezanson]

I'm not sure this code pattern should be allowed. This is happening because the same code (with different closure environments) is being added to multiple generic functions.

[tkelman]

Outside of a loop at the repl I get an error cannot define function f; it already has a value. I don't mind sprinkling eval in various places and using symbols, but I have an application where it's convenient to be saving the function instead of its name in the values of a Dict for evaluation purposes: functiondict[idx](x), and elsewhere define methods for a type by looping over values(functiondict).

[JeffBezanson]

...yikes. Ok I guess I'll just fix this for now.