JuliaLang / julia

The Julia Programming Language
https://julialang.org/
MIT License
45.9k stars 5.49k forks source link

Diagonal has no ldiv! #32648

Closed ChrisRackauckas closed 5 years ago

ChrisRackauckas commented 5 years ago

Seems silly, but since factorize(::Diagonal) returns a ::Diagonal, in order for generic non-allocating codes with factorize to work well with Diagonal the missing method needs to be added:

A = Diagonal(rand(3))
pA = factorize(A)
out = rand(3)
ldiv!(out,A,rand(3))
MethodError: no method matching ldiv!(::Array{Float64,1}, ::Diagonal{Float64,Array{Float64,1}}, ::Array{Float64,1})
Closest candidates are:
  ldiv!(::Union{DenseArray{T<:Union{Complex{Float64}, Float64},1}, DenseArray{T<:Union{Complex{Float64}, Float64},2}, ReinterpretArray{T<:Union{Complex{Float64}, Float64},1,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReinterpretArray{T<:Union{Complex{Float64}, Float64},2,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReshapedArray{T<:Union{Complex{Float64}, Float64},1,A,MI} where MI<:Tuple{Vararg{SignedMultiplicativeInverse{Int64},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReshapedArray{T<:Union{Complex{Float64}, Float64},2,A,MI} where MI<:Tuple{Vararg{SignedMultiplicativeInverse{Int64},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, SubArray{T<:Union{Complex{Float64}, Float64},1,A,I,L} where L where I<:Tuple{Vararg{Union{Int64, AbstractRange{Int64}, AbstractCartesianIndex},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, ReshapedArray{T,N,A,MI} where MI<:Tuple{Vararg{SignedMultiplicativeInverse{Int64},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, DenseArray}, SubArray{T<:Union{Complex{Float64}, Float64},2,A,I,L} where L where I<:Tuple{Vararg{Union{Int64, AbstractRange{Int64}, AbstractCartesianIndex},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, ReshapedArray{T,N,A,MI} where MI<:Tuple{Vararg{SignedMultiplicativeInverse{Int64},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, DenseArray}}, !Matched::SuiteSparse.UMFPACK.UmfpackLU{T<:Union{Complex{Float64}, Float64},Ti} where Ti<:Union{Int32, Int64}, ::Union{DenseArray{T<:Union{Complex{Float64}, Float64},1}, DenseArray{T<:Union{Complex{Float64}, Float64},2}, ReinterpretArray{T<:Union{Complex{Float64}, Float64},1,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReinterpretArray{T<:Union{Complex{Float64}, Float64},2,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReshapedArray{T<:Union{Complex{Float64}, Float64},1,A,MI} where MI<:Tuple{Vararg{SignedMultiplicativeInverse{Int64},N} where N} where A<:Union{ReinterpretArray{T,N,S,A} where S where A<:Union{SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray} where N where T, SubArray{T,N,A,I,true} where I<:Union{Tuple{Vararg{Real,N} where N}, Tuple{AbstractUnitRange,Vararg{Any,N} where N}} where A<:DenseArray where N where T, DenseArray}, ReshapedArray{T...

of course the fix is:

LinearAlgebra.ldiv!(x,A::Diagonal,b) = (x .= b ./ A.diag)
A = Diagonal(rand(2))
b = ones(2)
out = rand(2)
ldiv!(out,A,b)
antoine-levitt commented 5 years ago

Dup of https://github.com/JuliaLang/LinearAlgebra.jl/issues/615 I think the simplest way to solve this is to add a fallback ldiv!(Y::AbstractArray, A::AbstractArray, B::AbstractArray) = (copy!(Y,B); ldiv!(A,Y)).

dkarrasch commented 5 years ago

The proposed fix is not the right one, e.g., for non-commutative types. It should be D.diag .\ b, as in line 512 of diagonal.jl. To have a decent fallback would be good, but in the diagonal case, one can solve it in one loop, whereas the fallback has two.

andreasnoack commented 5 years ago

Closing as dup of JuliaLang/LinearAlgebra.jl#615