rdiv! for TriangularRFP

[dmbates]

I just noticed that we have an ldiv! method for TriangularRFP but not an rdiv! method. Of course, the one I need right now is rdiv!. I will prepare a PR to add it.

[dmbates]

Shouldn't the definition of the ldiv! method be for LinearAlgebra.ldiv! like the methods for LinearAlgebra.inv! or is ldiv! only intended for use in the next method definition for (\)?

[andreasnoack]

Yeah. I think that is probably a mistake.

[dmbates]

I think I have a PR working but ran into problems setting up tests for the rdiv! methods. I used code similar to the tests for rmul! and lmul! with modifications to ensure that the Matrix versions had A as LowerTriangular or UpperTriangular as warranted. I got errors of the form

  MethodError: no method matching rdiv!(::Vector{ComplexF32}, ::UpperTriangular{ComplexF32, Matrix{ComplexF32}})
  The function `rdiv!` exists, but no method is defined for this combination of argument types.

Is there a reason that these methods for rdiv! don't exist?

[andreasnoack]

I guess it is because the vector is considered a column so it would only work when the triangular matrix is 1x1

[dmbates]

It was a good thing that this test did cause an error because my test was wrong. 😑

[andreasnoack]

Fixed in #19