Fix 0.6 depwarnings (#18)

Fix 0.6 depwarnings
JuliaLinearAlgebra · Feb 4, 2017 · a69f9c7 · a69f9c7
1 parent 16d93e5
commit a69f9c7
Show file tree

Hide file tree

Showing 4 changed files with 70 additions and 72 deletions.
diff --git a/src/SymWoodburyMatrices.jl b/src/SymWoodburyMatrices.jl
@@ -6,8 +6,8 @@ using Base.LinAlg.BLAS:gemm!,gemm,axpy!
 Represents a matrix of the form A + BDBᵀ.
 """
 type SymWoodbury{T,AType, BType, DType} <: AbstractMatrix{T}
-  A::AType; 
-  B::BType; 
+  A::AType;
+  B::BType;
   D::DType;
 end
 
@@ -28,7 +28,7 @@ end
 function SymWoodbury{T}(A, B::AbstractVector{T}, D::T)
     n = size(A, 1)
     k = 1
-    if size(A, 2) != n || length(B) != n 
+    if size(A, 2) != n || length(B) != n
         throw(DimensionMismatch("Sizes of B ($(size(B))) and/or D ($(size(D))) are inconsistent with A ($(size(A)))"))
     end
     SymWoodbury{T,typeof(A),typeof(B),typeof(D)}(A,B,D)
@@ -47,14 +47,14 @@ function calc_inv(A, B, D)
   SymWoodbury(W,X,Z);
 end
 
-Base.inv{T<:Any, AType<:Any, BType<:AbstractVector, DType<:Real}(O::SymWoodbury{T,AType,BType,DType}) = 
+Base.inv{T<:Any, AType<:Any, BType<:AbstractVector, DType<:Real}(O::SymWoodbury{T,AType,BType,DType}) =
   calc_inv(O.A, O.B, O.D)
 
-Base.inv{T<:Any, AType<:Any, BType<:Any, DType<:AbstractMatrix}(O::SymWoodbury{T,AType,BType,DType}) = 
+Base.inv{T<:Any, AType<:Any, BType<:Any, DType<:AbstractMatrix}(O::SymWoodbury{T,AType,BType,DType}) =
   calc_inv(O.A, O.B, O.D)
 
-# D is typically small, so this is acceptable. 
-Base.inv{T<:Any, AType<:Any, BType<:Any, DType<:SparseMatrixCSC}(O::SymWoodbury{T,AType,BType,DType}) = 
+# D is typically small, so this is acceptable.
+Base.inv{T<:Any, AType<:Any, BType<:Any, DType<:SparseMatrixCSC}(O::SymWoodbury{T,AType,BType,DType}) =
   calc_inv(O.A, O.B, full(O.D));
 
 \(W::SymWoodbury, R::StridedVecOrMat) = inv(W)*R
@@ -100,7 +100,7 @@ on evaluation, i.e. `liftFactor(A)(x)` is the same as `inv(A)*x`.
 """
 liftFactor(O::SymWoodbury) = liftFactorVars(O.A,O.B,O.D)
 
-function *{T}(O::SymWoodbury{T}, x::Union{Matrix,Vector,SubArray}) 
+function *{T}(O::SymWoodbury{T}, x::Union{Matrix,Vector,SubArray})
   o = O.A*x;
   plusBDBtx!(o, O.B, O.D, x)
   return o
@@ -112,7 +112,7 @@ end
 
 plusBDBtx!(o, B::AbstractVector, D, x) = plusBDBtx!(o, reshape(B,size(B,1),1),D,x)
 
-# Optimization - use specialized BLAS package 
+# Optimization - use specialized BLAS package
 function plusBDBtx!(o, B::Array{Float64,2}, D, x::Array{Float64,2})
   w = D*gemm('T','N',B,x);
   gemm!('N','N',1.,B,w,1., o)
@@ -124,12 +124,12 @@ function plusBDBtx!(o, B::Array{Float64,1}, d::Real, x::Union{Array{Float64,2},
     axpy!(vecdot(B,x)*d, B, o)
   else
     w = d*gemm('T', 'N' ,reshape(B, size(B,1), 1),x);
-    gemm!('N','N',1.,B,w,1., o)    
+    gemm!('N','N',1.,B,w,1., o)
   end
 end
 
 Base.Ac_mul_B{T}(O1::SymWoodbury{T}, x::AbstractVector{T}) = O1*x
-Base.Ac_mul_B{T}(O1::SymWoodbury{T}, x::AbstractMatrix{T}) = O1*x
+Base.Ac_mul_B(O1::SymWoodbury, x::AbstractMatrix) = O1*x
 
 +(O::SymWoodbury, M::SymWoodbury)    = SymWoodbury(O.A + M.A, [O.B M.B],
                                                    cat([1,2],O.D,M.D) );
@@ -148,7 +148,7 @@ function square(O::SymWoodbury)
   AB = O.A*O.B
   Z  = [(AB + O.B) (AB - O.B)]
   R  = O.D*(O.B'*O.B)*O.D/4
-  D  = [ O.D/2 + R  -R 
+  D  = [ O.D/2 + R  -R
         -R          -O.D/2 + R ]
   SymWoodbury(A, Z, D)
 end
@@ -159,11 +159,11 @@ except when they are the same, i.e. the user writes A'A or A*A' or A*A.
 
 Z(A + B*D*Bᵀ) = ZA + ZB*D*Bᵀ
 
-This package will not support support left multiplication by a generic 
+This package will not support support left multiplication by a generic
 matrix, to keep return types consistent.
 """
 function *(O1::SymWoodbury, O2::SymWoodbury)
-  if (O1 === O2) 
+  if (O1 === O2)
     return square(O1)
   else
     if O1.A == O2.A && O1.B == O2.B && O1.D == O2.D
@@ -174,7 +174,6 @@ function *(O1::SymWoodbury, O2::SymWoodbury)
   end
 end
 
-Base.Ac_mul_B(O1::SymWoodbury, O2::SymWoodbury) = O1*O2
 Base.A_mul_Bc(O1::SymWoodbury, O2::SymWoodbury) = O1*O2
 
 conjm(O::SymWoodbury, M) = SymWoodbury(M*O.A*M', M*O.B, O.D);
@@ -187,6 +186,6 @@ Base.sparse(O::SymWoodbury) = sparse(full(O))
 
 # returns a pointer to the original matrix, this is consistent with the
 # behavior of Symmetric in Base.
-Base.ctranspose(O::SymWoodbury) = O 
+Base.ctranspose(O::SymWoodbury) = O
 
-Base.det(W::SymWoodbury) = det(convert(Woodbury, W))
+Base.det(W::SymWoodbury) = det(convert(Woodbury, W))
diff --git a/src/WoodburyMatrices.jl b/src/WoodburyMatrices.jl
@@ -38,15 +38,16 @@ function Woodbury{T}(A, U::AbstractMatrix{T}, C, V::AbstractMatrix{T})
     end
     Cp = inv(inv(C) + V*(A\U))
     # temporary space for allocation-free solver
-    tmpN1 = Array(T, N)
-    tmpN2 = Array(T, N)
-    tmpk1 = Array(T, k)
-    tmpk2 = Array(T, k)
+    tmpN1 = Array{T,1}(N)
+    tmpN2 = Array{T,1}(N)
+    tmpk1 = Array{T,1}(k)
+    tmpk2 = Array{T,1}(k)
     # don't copy A, it could be huge
     Woodbury{T,typeof(A),typeof(U),typeof(V),typeof(C),typeof(Cp)}(A, copy(U), copy(C), Cp, copy(V), tmpN1, tmpN2, tmpk1, tmpk2)
 end
 
 Woodbury{T}(A, U::Vector{T}, C, V::Matrix{T}) = Woodbury(A, reshape(U, length(U), 1), C, V)
+@static if isdefined(:RowVector)    Woodbury(A, U::AbstractVector, C, V::RowVector) = Woodbury(A, U, C, Matrix(V)) end
 
 size(W::Woodbury) = size(W.A)
 

diff --git a/test/runtests.jl b/test/runtests.jl
@@ -45,14 +45,14 @@ for elty in (Float32, Float64, Complex64, Complex128, Int)
     # Woodbury
     W = Woodbury(T, U, C, V)
     F = full(W)
-    @test_approx_eq W*v F*v
+    @test W*v ≈ F*v
     iFv = F\v
     @test norm(W\v - iFv)/norm(iFv) <= n*cond(F)*ε # Revisit. Condition number is wrong
     @test abs((det(W) - det(F))/det(F)) <= n*cond(F)*ε # Revisit. Condition number is wrong
     iWv = similar(iFv)
     if elty != Int
         iWv = A_ldiv_B!(W, copy(v))
-        @test_approx_eq iWv iFv
+        @test iWv ≈ iFv
     end
 end
 
@@ -110,7 +110,7 @@ for i in 1:5  # repeat 5 times
         by_hand[i, j] = v
     end
 
-    @test maxabs(out - by_hand) == 0.0
+    @test maximum(abs,out - by_hand) == 0.0
 end
 
 include("runtests_sym.jl")
diff --git a/test/runtests_sym.jl b/test/runtests_sym.jl
@@ -25,53 +25,51 @@ for elty in (Float32, Float64, Complex64, Complex128, Int), AMat in (diagm,)
 
     ε = eps(abs2(float(one(elty))))
     A = AMat(a)
-    
+
     # Woodbury
     for W in (SymWoodbury(A, B, D), SymWoodbury(A, B[:,1][:], 2.))
 
         F = full(W)
-
-        @test_approx_eq (2*W)*v 2*(W*v)
-        @test_approx_eq W'*v W*v
-        @test_approx_eq (W'W)*v full(W)*(full(W)*v)
-        @test_approx_eq (W*W)*v full(W)*(full(W)*v)
-        @test_approx_eq (W*W')*v full(W)*(full(W)*v)
-        @test_approx_eq W[1:3,1:3]*v[1:3] full(W)[1:3,1:3]*v[1:3]
-        @test_approx_eq sparse(W) full(W)
+        @test (2*W)*v ≈ 2*(W*v)
+        @test W'*v ≈ W*v
+        @test (W'W)*v ≈ full(W)*(full(W)*v)
+        @test (W*W)*v ≈ full(W)*(full(W)*v)
+        @test (W*W')*v ≈ full(W)*(full(W)*v)
+        @test W[1:3,1:3]*v[1:3] ≈ full(W)[1:3,1:3]*v[1:3]
+        @test sparse(W) ≈ full(W)
         @test W === W'
-        @test_approx_eq W*eye(n) full(W)
-        @test_approx_eq W'*eye(n) full(W)
+        @test W*eye(n) ≈ full(W)
+        @test W'*eye(n) ≈ full(W)
 
         Z = randn(n,n)
-        @test_approx_eq full(W*Z) full(W)*Z
+        @test full(W*Z) ≈ full(W)*Z
 
         R = rand(n,n)
 
         for v = (rand(n, 1), view(rand(n,1), 1:n), view(rand(n,2),1:n,1:2))
-            @test_approx_eq (2*W)*v 2*(W*v)
-            @test_approx_eq (W*2)*v 2*(W*v)
-            @test_approx_eq (W'W)*v full(W)*(full(W)*v)
-            @test_approx_eq (W*W)*v full(W)*(full(W)*v)
-            @test_approx_eq (W*W')*v full(W)*(full(W)*v)
-            @test_approx_eq W[1:3,1:3]*v[1:3] full(W)[1:3,1:3]*v[1:3]
-            @test_approx_eq full(WoodburyMatrices.conjm(W, R)) R*full(W)*R'
-            @test_approx_eq full((copy(W)'W)*v) full(W)*(full(W)*v)
-            @test_approx_eq full(W + A) full(W)+full(A)
-            @test_approx_eq full(A + W) full(W)+full(A)
+            @test (2*W)*v ≈ 2*(W*v)
+            @test (W*2)*v ≈ 2*(W*v)
+            @test (W'W)*v ≈ full(W)*(full(W)*v)
+            @test (W*W)*v ≈ full(W)*(full(W)*v)
+            @test (W*W')*v ≈ full(W)*(full(W)*v)
+            @test W[1:3,1:3]*v[1:3] ≈ full(W)[1:3,1:3]*v[1:3]
+            @test full(WoodburyMatrices.conjm(W, R)) ≈ R*full(W)*R'
+            @test full((copy(W)'W)*v) ≈ full(W)*(full(W)*v)
+            @test full(W + A) ≈ full(W)+full(A)
+            @test full(A + W) ≈ full(W)+full(A)
         end
 
         v = rand(n,1)
-
         W2 = convert(Woodbury, W)
-        @test_approx_eq full(W2) full(W)
+        @test full(W2) ≈ full(W)
 
         if elty != Int
-            @test_approx_eq inv(W)*v inv(full(W))*v
-            @test_approx_eq W\v inv(full(W))*v        
-            @test_approx_eq liftFactor(W)(v) inv(W)*v
-            @test_approx_eq WoodburyMatrices.partialInv(W)[1] inv(W).B
-            @test_approx_eq WoodburyMatrices.partialInv(W)[2] inv(W).D
-            @test_approx_eq det(W) det(full(W))
+            @test inv(W)*v ≈ inv(full(W))*v
+            @test W\v ≈ inv(full(W))*v
+            @test liftFactor(W)(v) ≈ inv(W)*v
+            @test WoodburyMatrices.partialInv(W)[1] ≈ inv(W).B
+            @test WoodburyMatrices.partialInv(W)[2] ≈ inv(W).D
+            @test det(W) ≈ det(full(W))
         end
 
     end
@@ -84,11 +82,11 @@ for elty in (Float32, Float64, Complex64, Complex128, Int)
     elty = Float64
 
     a1 = rand(n); B1 = rand(n,2); D1 = rand(2,2); v = rand(n)
-    a2 = rand(n); B2 = rand(n,2); D2 = rand(2,2); 
+    a2 = rand(n); B2 = rand(n,2); D2 = rand(2,2);
 
     if elty == Int
         v = rand(1:100, n)
-        
+
         a1 = rand(1:100, n)
         B1 = rand(1:100, 2, n)
         D1 = rand(1:100, 2, 2)
@@ -109,7 +107,7 @@ for elty in (Float32, Float64, Complex64, Complex128, Int)
     end
 
     ε = eps(abs2(float(one(elty))))
-    
+
     # Woodbury
     A1 = diagm(a1)
     A2 = diagm(a2)
@@ -121,9 +119,9 @@ for elty in (Float32, Float64, Complex64, Complex128, Int)
     W2r = SymWoodbury(A2, B2[:,1][:], 3.)
 
     for (W1, W2) = ((W1,W2), (W1r, W2), (W1, W2r), (W1r,W2r))
-        @test_approx_eq (W1 + W2)*v (full(W1) + full(W2))*v
-        @test_approx_eq (full(W1) + W2)*v (full(W1) + full(W2))*v
-        @test_approx_eq (W1 + 2*diagm(a1))*v (full(W1) + full(2*diagm(a1)))*v
+        @test (W1 + W2)*v ≈ (full(W1) + full(W2))*v
+        @test (full(W1) + W2)*v ≈ (full(W1) + full(W2))*v
+        @test (W1 + 2*diagm(a1))*v ≈ (full(W1) + full(2*diagm(a1)))*v
         @test_throws MethodError W1*W2
     end
 
@@ -142,18 +140,18 @@ V = randn(n,1)
 @test size(W,1) == n
 @test size(W,2) == n
 
-@test_approx_eq inv(W)*v inv(full(W))*v
-@test_approx_eq (2*W)*v 2*(W*v)
-@test_approx_eq (W'W)*v full(W)*(full(W)*v)
-@test_approx_eq (W*W)*v full(W)*(full(W)*v)
-@test_approx_eq (W*W')*v full(W)*(full(W)*v)
-@test_approx_eq liftFactor(W)(v) inv(W)*v
-
-@test_approx_eq inv(W)*V inv(full(W))*V
-@test_approx_eq (2*W)*V 2*(W*V)
-@test_approx_eq (W'W)*V full(W)*(full(W)*V)
-@test_approx_eq (W*W)*V full(W)*(full(W)*V)
-@test_approx_eq (W*W')*V full(W)*(full(W)*V)
+@test inv(W)*v ≈ inv(full(W))*v
+@test (2*W)*v ≈ 2*(W*v)
+@test (W'W)*v ≈ full(W)*(full(W)*v)
+@test (W*W)*v ≈ full(W)*(full(W)*v)
+@test (W*W')*v ≈ full(W)*(full(W)*v)
+@test liftFactor(W)(v) ≈ inv(W)*v
+
+@test inv(W)*V ≈ inv(full(W))*V
+@test (2*W)*V ≈ 2*(W*V)
+@test (W'W)*V ≈ full(W)*(full(W)*V)
+@test (W*W)*V ≈ full(W)*(full(W)*V)
+@test (W*W')*V ≈ full(W)*(full(W)*V)
 
 # Mismatched sizes
 @test_throws DimensionMismatch SymWoodbury(rand(5,5),rand(5,2),rand(2,3))