_fill_dot support general vectors (#229)

* Update fillalgebra.jl

* promote_op

* add breaking test

* add breaking test

* fix

* accept round-off errors

* Update test/runtests.jl

Co-authored-by: Sheehan Olver <[email protected]>

* update

* support inf and nan

* fix 1.6

* Update fillalgebra.jl

* Update fillalgebra.jl

* trying to fix Julia 1.6

* comments

* Update runtests.jl

* add @inferred

---------

Co-authored-by: Sheehan Olver <[email protected]>

Author: putianyi889

src/fillalgebra.jl

@@ -159,38 +159,22 @@ function *(a::Transpose{T, <:AbstractVector{T}}, b::ZerosVector{T}) where T<:Rea
 end
 *(a::Transpose{T, <:AbstractMatrix{T}}, b::ZerosVector{T}) where T<:Real = mult_zeros(a, b)
 
-# treat zero separately to support ∞-vectors
-function _zero_dot(a, b)
-    axes(a) == axes(b) || throw(DimensionMismatch("dot product arguments have lengths $(length(a)) and $(length(b))"))
-    zero(promote_type(eltype(a),eltype(b)))
-end
-
-_fill_dot(a::Zeros, b::Zeros) = _zero_dot(a, b)
-_fill_dot(a::Zeros, b) = _zero_dot(a, b)
-_fill_dot(a, b::Zeros) = _zero_dot(a, b)
-_fill_dot(a::Zeros, b::AbstractFill) = _zero_dot(a, b)
-_fill_dot(a::AbstractFill, b::Zeros) = _zero_dot(a, b)
-
-function _fill_dot(a::AbstractFill, b::AbstractFill)
-    axes(a) == axes(b) || throw(DimensionMismatch("dot product arguments have lengths $(length(a)) and $(length(b))"))
-    getindex_value(a)getindex_value(b)*length(b)
-end
-
 # support types with fast sum
-function _fill_dot(a::AbstractFill, b)
+# infinite cases should be supported in InfiniteArrays.jl
+# type issues of Bool dot are ignored at present.
+function _fill_dot(a::AbstractFillVector{T}, b::AbstractVector{V}) where {T,V}
     axes(a) == axes(b) || throw(DimensionMismatch("dot product arguments have lengths $(length(a)) and $(length(b))"))
-    getindex_value(a)sum(b)
+    dot(getindex_value(a), sum(b))
 end
 
-function _fill_dot(a, b::AbstractFill)
+function _fill_dot_rev(a::AbstractVector{T}, b::AbstractFillVector{V}) where {T,V}
     axes(a) == axes(b) || throw(DimensionMismatch("dot product arguments have lengths $(length(a)) and $(length(b))"))
-    sum(a)getindex_value(b)
+    dot(sum(a), getindex_value(b))
 end
 
-
 dot(a::AbstractFillVector, b::AbstractFillVector) = _fill_dot(a, b)
 dot(a::AbstractFillVector, b::AbstractVector) = _fill_dot(a, b)
-dot(a::AbstractVector, b::AbstractFillVector) = _fill_dot(a, b)
+dot(a::AbstractVector, b::AbstractFillVector) = _fill_dot_rev(a, b)
 
 function dot(u::AbstractVector, E::Eye, v::AbstractVector)
     length(u) == size(E,1) && length(v) == size(E,2) ||

test/runtests.jl

@@ -329,20 +329,32 @@ end
 # type, and produce numerically correct results.
 as_array(x::AbstractArray) = Array(x)
 as_array(x::UniformScaling) = x
-function test_addition_and_subtraction(As, Bs, Tout::Type)
+equal_or_undef(a::Number, b::Number) = (a == b) || isequal(a, b)
+equal_or_undef(a, b) = all(equal_or_undef.(a, b))
+function test_addition_subtraction_dot(As, Bs, Tout::Type)
     for A in As, B in Bs
-        @testset "$(typeof(A)) ± $(typeof(B))" begin
+        @testset "$(typeof(A)) and $(typeof(B))" begin
             @test A + B isa Tout{promote_type(eltype(A), eltype(B))}
-            @test as_array(A + B) == as_array(A) + as_array(B)
+            @test equal_or_undef(as_array(A + B), as_array(A) + as_array(B))
 
             @test A - B isa Tout{promote_type(eltype(A), eltype(B))}
-            @test as_array(A - B) == as_array(A) - as_array(B)
+            @test equal_or_undef(as_array(A - B), as_array(A) - as_array(B))
 
             @test B + A isa Tout{promote_type(eltype(B), eltype(A))}
-            @test as_array(B + A) == as_array(B) + as_array(A)
+            @test equal_or_undef(as_array(B + A), as_array(B) + as_array(A))
 
             @test B - A isa Tout{promote_type(eltype(B), eltype(A))}
-            @test as_array(B - A) == as_array(B) - as_array(A)
+            @test equal_or_undef(as_array(B - A), as_array(B) - as_array(A))
+            
+            # Julia 1.6 doesn't support dot(UniformScaling)
+            if VERSION < v"1.6.0" || VERSION >= v"1.8.0"
+                d1 = dot(A, B)
+                d2 = dot(as_array(A), as_array(B))
+                d3 = dot(B, A)
+                d4 = dot(as_array(B), as_array(A))
+                @test d1 ≈ d2 || d1 ≡ d2
+                @test d3 ≈ d4 || d3 ≡ d4
+            end
         end
     end
 end
@@ -372,37 +384,37 @@ end
     @test -A_fill === Fill(-A_fill.value, 5)
 
     # FillArray +/- FillArray should construct a new FillArray.
-    test_addition_and_subtraction((A_fill, B_fill), (A_fill, B_fill), Fill)
+    test_addition_subtraction_dot((A_fill, B_fill), (A_fill, B_fill), Fill)
     test_addition_and_subtraction_dim_mismatch(A_fill, Fill(randn(rng), 5, 2))
 
     # FillArray + Array (etc) should construct a new Array using `getindex`.
-    A_dense, B_dense = randn(rng, 5), [5, 4, 3, 2, 1]
-    test_addition_and_subtraction((A_fill, B_fill), (A_dense, B_dense), Array)
+    B_dense = (randn(rng, 5), [5, 4, 3, 2, 1], fill(Inf, 5), fill(NaN, 5))
+    test_addition_subtraction_dot((A_fill, B_fill), B_dense, Array)
     test_addition_and_subtraction_dim_mismatch(A_fill, randn(rng, 5, 2))
 
     # FillArray + StepLenRange / UnitRange (etc) should yield an AbstractRange.
     A_ur, B_ur = 1.0:5.0, 6:10
-    test_addition_and_subtraction((A_fill, B_fill), (A_ur, B_ur), AbstractRange)
+    test_addition_subtraction_dot((A_fill, B_fill), (A_ur, B_ur), AbstractRange)
     test_addition_and_subtraction_dim_mismatch(A_fill, 1.0:6.0)
     test_addition_and_subtraction_dim_mismatch(A_fill, 5:10)
 
     # FillArray + UniformScaling should yield a Matrix in general
     As_fill_square = (Fill(randn(rng, Float64), 3, 3), Fill(5, 4, 4))
     Bs_us = (UniformScaling(2.3), UniformScaling(3))
-    test_addition_and_subtraction(As_fill_square, Bs_us, Matrix)
+    test_addition_subtraction_dot(As_fill_square, Bs_us, Matrix)
     As_fill_nonsquare = (Fill(randn(rng, Float64), 3, 2), Fill(5, 3, 4))
     for A in As_fill_nonsquare, B in Bs_us
         test_addition_and_subtraction_dim_mismatch(A, B)
     end
 
     # FillArray + StaticArray should not have ambiguities
     A_svec, B_svec = SVector{5}(rand(5)), SVector(1, 2, 3, 4, 5)
-    test_addition_and_subtraction((A_fill, B_fill, Zeros(5)), (A_svec, B_svec), SVector{5})
+    test_addition_subtraction_dot((A_fill, B_fill, Zeros(5)), (A_svec, B_svec), SVector{5})
 
     # Issue #224
     A_matmat, B_matmat = Fill(rand(3,3),5), [rand(3,3) for n=1:5]
-    test_addition_and_subtraction((A_matmat,), (A_matmat,), Fill)
-    test_addition_and_subtraction((B_matmat,), (A_matmat,), Vector)
+    test_addition_subtraction_dot((A_matmat,), (A_matmat,), Fill)
+    test_addition_subtraction_dot((B_matmat,), (A_matmat,), Vector)
 
     # Optimizations for Zeros and RectOrDiagonal{<:Any, <:AbstractFill}
     As_special_square = (
@@ -412,7 +424,7 @@ end
         RectDiagonal(Fill(randn(rng, Float64), 3), 3, 3), RectDiagonal(Fill(3, 4), 4, 4)
     )
     DiagonalAbstractFill{T} = Diagonal{T, <:AbstractFill{T, 1}}
-    test_addition_and_subtraction(As_special_square, Bs_us, DiagonalAbstractFill)
+    test_addition_subtraction_dot(As_special_square, Bs_us, DiagonalAbstractFill)
     As_special_nonsquare = (
         Zeros(3, 2), Zeros{Int}(3, 4),
         Eye(3, 2), Eye{Int}(3, 4),
@@ -537,7 +549,7 @@ end
         @test [SVector(1,2)', SVector(2,3)', SVector(3,4)']' * Zeros{Int}(3) === SVector(0,0)
         @test_throws DimensionMismatch randn(4)' * Zeros(3)
         @test Zeros(5)' * randn(5,3) ≡ Zeros(5)'*Zeros(5,3) ≡ Zeros(5)'*Ones(5,3) ≡ Zeros(3)'
-        @test Zeros(5)' * randn(5) ≡ Zeros(5)' * Zeros(5) ≡ Zeros(5)' * Ones(5) ≡ 0.0
+        @test abs(Zeros(5)' * randn(5)) ≡ abs(Zeros(5)' * Zeros(5)) ≡ abs(Zeros(5)' * Ones(5)) ≡ 0.0
         @test Zeros(5) * Zeros(6)' ≡ Zeros(5,1) * Zeros(6)' ≡ Zeros(5,6)
         @test randn(5) * Zeros(6)' ≡ randn(5,1) * Zeros(6)' ≡ Zeros(5,6)
         @test Zeros(5) * randn(6)' ≡ Zeros(5,6)
@@ -552,7 +564,7 @@ end
         @test transpose([1, 2, 3]) * Zeros{Int}(3) === zero(Int)
         @test_throws DimensionMismatch transpose(randn(4)) * Zeros(3)
         @test transpose(Zeros(5)) * randn(5,3) ≡ transpose(Zeros(5))*Zeros(5,3) ≡ transpose(Zeros(5))*Ones(5,3) ≡ transpose(Zeros(3))
-        @test transpose(Zeros(5)) * randn(5) ≡ transpose(Zeros(5)) * Zeros(5) ≡ transpose(Zeros(5)) * Ones(5) ≡ 0.0
+        @test abs(transpose(Zeros(5)) * randn(5)) ≡ abs(transpose(Zeros(5)) * Zeros(5)) ≡ abs(transpose(Zeros(5)) * Ones(5)) ≡ 0.0
         @test randn(5) * transpose(Zeros(6)) ≡ randn(5,1) * transpose(Zeros(6)) ≡ Zeros(5,6)
         @test Zeros(5) * transpose(randn(6)) ≡ Zeros(5,6)
         @test transpose(randn(5)) * Zeros(5) ≡ 0.0
@@ -571,13 +583,13 @@ end
         @test +(z1) === z1
         @test -(z1) === z1
 
-        test_addition_and_subtraction((z1, z2), (z1, z2), Zeros)
+        test_addition_subtraction_dot((z1, z2), (z1, z2), Zeros)
         test_addition_and_subtraction_dim_mismatch(z1, Zeros{Float64}(4, 2))
     end
 
     # `Zeros` +/- `Fill`s should yield `Fills`.
     fill1, fill2 = Fill(5.0, 4), Fill(5, 4)
-    test_addition_and_subtraction((z1, z2), (fill1, fill2), Fill)
+    test_addition_subtraction_dot((z1, z2), (fill1, fill2), Fill)
     test_addition_and_subtraction_dim_mismatch(z1, Fill(5, 5))
 
     X = randn(3, 5)
@@ -1326,17 +1338,19 @@ end
     Random.seed!(5)
     u = rand(n)
     v = rand(n)
+    c = rand(ComplexF16, n)
 
     @test dot(u, D, v) == dot(u, v)
     @test dot(u, 2D, v) == 2dot(u, v)
     @test dot(u, Z, v) == 0
 
-    @test dot(Zeros(5), Zeros{ComplexF16}(5)) ≡ zero(ComplexF64)
-    @test dot(Zeros(5), Ones{ComplexF16}(5)) ≡ zero(ComplexF64)
-    @test dot(Ones{ComplexF16}(5), Zeros(5)) ≡ zero(ComplexF64)
-    @test dot(randn(5), Zeros{ComplexF16}(5)) ≡ dot(Zeros{ComplexF16}(5), randn(5)) ≡ zero(ComplexF64)
+    @test @inferred(dot(Zeros(5), Zeros{ComplexF16}(5))) ≡ zero(ComplexF64)
+    @test @inferred(dot(Zeros(5), Ones{ComplexF16}(5))) ≡ zero(ComplexF64)
+    @test abs(@inferred(dot(Ones{ComplexF16}(5), Zeros(5)))) ≡ abs(@inferred(dot(randn(5), Zeros{ComplexF16}(5)))) ≡ abs(@inferred(dot(Zeros{ComplexF16}(5), randn(5)))) ≡ zero(Float64) # 0.0 !≡ -0.0
+    @test @inferred(dot(c, Fill(1 + im, 15))) ≡ (@inferred(dot(Fill(1 + im, 15), c)))' ≈ @inferred(dot(c, fill(1 + im, 15)))
 
-    @test dot(Fill(1,5), Fill(2.0,5)) ≡ 10.0
+    @test @inferred(dot(Fill(1,5), Fill(2.0,5))) ≡ 10.0
+    @test_skip dot(Fill(true,5), Fill(Int8(1),5)) isa Int8 # not working at present
 
     let N = 2^big(1000) # fast dot for fast sum
         @test dot(Fill(2,N),1:N) == dot(Fill(2,N),1:N) == dot(1:N,Fill(2,N)) == 2*sum(1:N)