Skip to content

RFC: Refactor LinearAlgebra to use isa instead of method dispatch to avoid ambiguity hell #33072

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 2 commits into from
Closed

RFC: Refactor LinearAlgebra to use isa instead of method dispatch to avoid ambiguity hell #33072

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
138aeb4
Use `A isa (Adjoint|Transpose)AbsVec` instead of method
tkf Aug 26, 2019
c7fadec
Remove obsolete ambiguity resolutions
tkf Aug 26, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 0 additions & 4 deletions stdlib/LinearAlgebra/src/adjtrans.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,10 +251,6 @@ function *(u::TransposeAbsVec, v::AbstractVector)
end
# vector * Adjoint/Transpose-vector
*(u::AbstractVector, v::AdjOrTransAbsVec) = broadcast(*, u, v)
# Adjoint/Transpose-vector * Adjoint/Transpose-vector
# (necessary for disambiguation with fallback methods in linalg/matmul)
*(u::AdjointAbsVec, v::AdjointAbsVec) = throw(MethodError(*, (u, v)))
*(u::TransposeAbsVec, v::TransposeAbsVec) = throw(MethodError(*, (u, v)))

# AdjOrTransAbsVec{<:Any,<:AdjOrTransAbsVec} is a lazy conj vectors
# We need to expand the combinations to avoid ambiguities
Expand Down
9 changes: 5 additions & 4 deletions stdlib/LinearAlgebra/src/matmul.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -129,10 +129,6 @@ end
return generic_matvecmul!(y, 'C', A, x, MulAddMul(alpha, beta))
end

# Vector-Matrix multiplication
(*)(x::AdjointAbsVec, A::AbstractMatrix) = (A'*x')'
(*)(x::TransposeAbsVec, A::AbstractMatrix) = transpose(transpose(A)*transpose(x))

# Matrix-matrix multiplication

"""
Expand All @@ -149,6 +145,11 @@ julia> [1 1; 0 1] * [1 0; 1 1]
```
"""
function (*)(A::AbstractMatrix, B::AbstractMatrix)
if A isa AdjointAbsVec
return (A'*x')'
elseif A isa TransposeAbsVec
return transpose(transpose(A)*transpose(x))
end
TS = promote_op(matprod, eltype(A), eltype(B))
mul!(similar(B, TS, (size(A,1), size(B,2))), A, B)
end
Expand Down
1 change: 0 additions & 1 deletion stdlib/LinearAlgebra/src/qr.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -725,7 +725,6 @@ function *(A::StridedMatrix, adjB::Adjoint{<:Any,<:AbstractQ})
throw(DimensionMismatch("matrix A has dimensions $(size(A)) but matrix B has dimensions $(size(B))"))
end
end
*(u::AdjointAbsVec, A::Adjoint{<:Any,<:AbstractQ}) = adjoint(A.parent * u.parent)


### AcQ/AcQc
Expand Down
5 changes: 0 additions & 5 deletions stdlib/LinearAlgebra/src/triangular.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2020,11 +2020,6 @@ function *(A::AbstractMatrix, transB::Transpose{<:Any,<:AbstractTriangular})
copyto!(AA, A)
rmul!(AA, transpose(convert(AbstractArray{TAB}, B)))
end
# ambiguity resolution with definitions in linalg/rowvector.jl
*(v::AdjointAbsVec, A::AbstractTriangular) = adjoint(adjoint(A) * v.parent)
*(v::TransposeAbsVec, A::AbstractTriangular) = transpose(transpose(A) * v.parent)
*(v::AdjointAbsVec, A::Adjoint{<:Any,<:AbstractTriangular}) = adjoint(A.parent * v.parent)
*(v::TransposeAbsVec, A::Transpose{<:Any,<:AbstractTriangular}) = transpose(A.parent * v.parent)


# If these are not defined, they will fallback to the versions in matmul.jl
Expand Down