Vector-Matrix product again

[goszlanyi]

While waiting for JuliaLang/LinearAlgebra.jl#360 , I have found a suspicious method.

line 92 of matmul.jl is:
(*)(A::AbstractVector, B::AbstractMatrix) = reshape(A,length(A),1)*B

I think that this line is the reason why v*mx multiplication does not work at all.
The error message is the following:

julia> ones(3)*ones(3,3)
ERROR: DimensionMismatch("A has dimensions (3,1) but B has dimensions (3,3)")
 in gemm_wrapper!(::Array{Float64,2}, ::Char, ::Char, ::Array{Float64,2}, ::Array{Float64,2}) at .\linalg\matmul.jl:308
 in *(::Array{Float64,1}, ::Array{Float64,2}) at .\linalg\matmul.jl:86

If I redefine the method of line 92 in my workspace as:

import Base.*
(*)(A::AbstractVector, B::AbstractMatrix) = reshape(A,1,length(A))*B

then v*mx works and returns a 1xn matrix.

@andreasnoack @StefanKarpinski
Note, that this is not the same issue as JuliaLang/LinearAlgebra.jl#360.
Here I report something that should already work,
while there I would like the result to be a real 1-dimensional vector.

[andreasnoack]

This is the method that lets you compute x*x'. As mentioned the vector discussion is JuliaLang/LinearAlgebra.jl#42 so I'll close here.

[goszlanyi]

This is the method that lets you compute x*x'.

Thank you. Then this is even more serious than I have originally thought.
Based on the (*)(A::AbstractVector, B::AbstractMatrix) signature
we can only have a single method, which is only good for either v*v' or v*mx.

The second usage is perfectly valid mathematics,
but only if it also returns a 1-dimensional vector.

I will try to summarize, and as you have suggested, move it to JuliaLang/LinearAlgebra.jl#42.