Define size for SVD

[mohamed82008]

Closes #28438. And first PR!

[ararslan]

Welcome to Julia! Given @fredrikekre's comment in the linked issue, is this pull request still relevant? If so, it would be great if you could add a test.

[mohamed82008]

My motivation for this PR was:

julia> using LinearAlgebra

julia> ldiv!(zeros(3), svd(rand(3,3)), rand(3))
ERROR: MethodError: no method matching size(::SVD{Float64,Float64,Array{Float64,2}}, ::Int64)
Closest candidates are:
  size(::Any, ::Integer, ::Integer) at deprecated.jl:53
  size(::Any, ::Integer, ::Integer, ::Integer...) at deprecated.jl:53
  size(::BitArray{1}, ::Any) at bitarray.jl:74
  ...
Stacktrace:
 [1] ldiv!(::Array{Float64,1}, ::SVD{Float64,Float64,Array{Float64,2}}, ::Array{Float64,1}) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\stdlib\v0.7\LinearAlgebra\src\factorization.jl:101
 [2] top-level scope at none:0

The above works after defining size for SVD. So I think it's relevant yes. size is defined for QR and Cholesky factorizations but not SVD.

[Keno]

One of the linear algebra folks will have to comment if this makes sense, but it would be good to add the motivating failure as a test case.

[andreasnoack]

This is right. For factorizations, size is the size of the factorized matrix. Iteration is just to be able to deconstruct the factorization and shouldn't generally be used.

It looks like there is some whitespace somewhere in the diff which is causing the tests to fail.

[fisiognomico]

The indentation here is making CI fail, as it is superfluous. Extra white spaces are not allowed as they make diffs harder to read.

[ararslan]

Should this be included in 0.7?

[andreasnoack]

Should this be included in 0.7?

Yes, I've added the backport label.