array fixes

src/FDDerivatives.jl

@@ -3,14 +3,14 @@ module FDDerivatives
 "Create Jacobian function"
 function makejacobian(f::Function, h::Float64=sqrt(eps(Float64)))
 	function jacobian(x::Vector)
-		xphs = Array(Array{Float64, 1}, length(x) + 1)
+		xphs = Array{Array{Float64, 1}}(length(x) + 1)
 		for i = 1:length(x)
 			xphs[i] = copy(x)
 			xphs[i][i] += h
 		end
 		xphs[end] = copy(x)
 		ys = pmap(f, xphs)
-		J = Array(eltype(ys[1]), length(ys[1]), length(x))
+		J = Array{eltype(ys[1])}(length(ys[1]), length(x))
 		for i = 1:length(x)
 			J[:, i] = ys[i] - ys[end]
 		end
@@ -22,14 +22,14 @@ end
 "Create Gradient function"
 function makegradient(f::Function, h::Float64=sqrt(eps(Float64)))
 	function gradient(x::Vector)
-		xphs = Array(Array{Float64, 1}, length(x) + 1)
+		xphs = Array{Array{Float64, 1}}(length(x) + 1)
 		for i = 1:length(x)
 			xphs[i] = copy(x)
 			xphs[i][i] += h
 		end
 		xphs[end] = copy(x)
 		ys = pmap(f, xphs)
-		grad = Array(eltype(ys), length(x))
+		grad = Array{eltype(ys)}(length(x))
 		for i = 1:length(x)
 			grad[i] = ys[i] - ys[end]
 		end

src/FFTRF.jl

@@ -6,12 +6,12 @@ import Base.Cartesian
 "Reduce k"
 @generated function reducek(k, dimensionvaltype)
 	if dimensionvaltype == Type{Val{2}}
-		setupfinalk = :(finalk = Array(Float64, div(size(k, 2), 2), div(size(k, 1), 2)))
+		setupfinalk = :(finalk = Array{Float64}(div(size(k, 2), 2), div(size(k, 1), 2)))
 		innerloop = quote
 			finalk[j, i] = real(k[i, j])
 		end
 	elseif dimensionvaltype == Type{Val{3}}
-		setupfinalk = :(finalk = Array(Float64, div(size(k, 2), 2), div(size(k, 1), 2), div(size(k, 3), 2)))
+		setupfinalk = :(finalk = Array{Float64}(div(size(k, 2), 2), div(size(k, 1), 2), div(size(k, 3), 2)))
 		innerloop = quote
 			for h = 1:div(size(k, 3), 2)
 				finalk[j, i, h] = real(k[i, j, h])
@@ -69,7 +69,7 @@ end
 
 function mulbyphi(S)
 	phi = randn(size(S))
-	result = Array(Complex{eltype(S)}, size(S))
+	result = Array{Complex{eltype(S)}}(size(S))
 	for i = 1:length(phi)
 		result[i] = S[i] * Complex{eltype(S)}(cospi(2 * phi[i]), sinpi(2 * phi[i]))
 	end
@@ -79,7 +79,7 @@ end
 function powerlaw_structuredgrid(Ns, k0, dk, beta)
 	originalNs = Ns
 	Ns = 2 * Ns
-	fouriercoords = Array(Array{Float64, 1}, length(Ns))
+	fouriercoords = Array{Array{Float64, 1}}(length(Ns))
 	for i = 1:length(Ns)
 		fouriercoords[i] = vcat(collect(0:originalNs[i]), -1 * collect((originalNs[i] - 1):-1:1))
 	end
@@ -113,7 +113,7 @@ end
 		@Base.Cartesian.nexprs $ndims j->begin minx_j, maxx_j = extrema(points[j, :]) end
 		@Base.Cartesian.nexprs $ndims j->begin x_j = range(minx_j, (maxx_j - minx_j) / (size(structuredvals, j) - 1), size(structuredvals, j)) end
 		valinterp = Grid.CoordInterpGrid($t, structuredvals, Grid.BCnil, Grid.InterpLinear)
-		result = Array(Float64, size(points, 2))
+		result = Array{Float64}(size(points, 2))
 		for i = 1:length(result)
 			result[i] = valinterp[points[:, i]...]
 		end

src/GeostatInversion.jl

@@ -25,7 +25,7 @@ function getxis(::Type{Val{:iwantfields}}, samplefield::Function, numfields::Int
 	fields = RobustPmap.rpmap(i->samplefield(), 1:numfields; t=Array{Float64, 1})
 	lrcm = LowRankCovMatrix(fields)
 	Z = randsvdwithseed(lrcm, numxis, p, q, seed)
-	xis = Array(Array{Float64, 1}, numxis)
+	xis = Array{Array{Float64, 1}}(numxis)
 	for i = 1:numxis
 		xis[i] = Z[:, i]
 	end
@@ -56,7 +56,7 @@ function getxis(samplefield::Function, numfields::Int, numxis::Int, p::Int, q::I
 end
 
 function getxis(Q::Matrix, numxis::Int, p::Int, q::Int=3, seed=nothing)#numxis is the number of xis, p is oversampling for randsvd accuracy, q is the number of power iterations -- see review paper by Halko et al
-	xis = Array(Array{Float64, 1}, numxis)
+	xis = Array{Array{Float64, 1}}(numxis)
 	Z = randsvdwithseed(Q, numxis, p, q, seed)
 	for i = 1:numxis
 		xis[i] = Z[:, i]

src/RandMatFact.jl

@@ -2,7 +2,7 @@
 module RandMatFact
 
 function colnorms(Y)
-	norms = Array(Float64, size(Y, 2))
+	norms = Array{Float64}(size(Y, 2))
 	for i = 1:size(Y, 2)
 		norms[i] = norm(Y[:, i])
 	end
@@ -15,11 +15,11 @@ function rangefinder(A; epsilon=1e-8, r=10)#implements algorithm 4.2 in halko et
 	Yfull = zeros(Float64, n, r + min(n, m))
 	Y = view(Yfull, :, 1:r)
 	BLAS.gemm!('N', 'N', 1., A, randn(n, r), 0., Y)
-	omega = Array(Float64, n)
+	omega = Array{Float64}(n)
 	j = 0
 	Qfull = zeros(Float64, m, min(n, m))
-	tempvec = Array(Float64, m)
-	Aomega = Array(Float64, m)
+	tempvec = Array{Float64}(m)
+	Aomega = Array{Float64}(m)
 	while maximum(colnorms(view(Yfull, :, j+1:j+r))) > epsilon / sqrt(200 / pi)
 		j = j + 1
 		Yj = view(Yfull, :, j)

src/direct.jl

@@ -19,7 +19,7 @@ Arguments:
 - callback : a function of the form `(params, observations)->...` that is called during each iteration
 """
 function pcgadirect(forwardmodel::Function, s0::Vector, X::Vector, xis::Array{Array{Float64, 1}, 1}, R, y::Vector; maxiters::Int=5, delta::Float64=sqrt(eps(Float64)), xtol::Float64=1e-6, callback=(s, obs_cal)->nothing)
-	HQH = Array(Float64, length(y), length(y))
+	HQH = Array{Float64}(length(y), length(y))
 	converged = false
 	s = s0
 	itercount = 0
@@ -36,14 +36,14 @@ end
 
 function pcgadirectiteration!(HQH::Matrix, forwardmodel::Function, s::Vector, X::Vector, xis::Array{Array{Float64, 1}, 1}, R, y::Vector, delta::Float64, callback)
 	p = 1#p = 1 because X is a vector rather than a full matrix
-	paramstorun = Array(Array{Float64, 1}, length(xis) + 3)
+	paramstorun = Array{Array{Float64, 1}}(length(xis) + 3)
 	for i = 1:length(xis)
 		paramstorun[i] = s + delta * xis[i]
 	end
 	paramstorun[length(xis) + 1] = s + delta * X
 	paramstorun[length(xis) + 2] = s + delta * s
 	paramstorun[length(xis) + 3] = s
-	results = pmap(forwardmodel, paramstorun) 
+	results = pmap(forwardmodel, paramstorun)
 	callback(s, results[length(xis) + 3])
 	hs = results[length(xis) + 3]
 	fill!(HQH, 0.)

src/lowrank.jl

@@ -10,7 +10,7 @@ import Base.transpose
 type LowRankCovMatrix
 	samples::Array{Array{Float64, 1}, 1}
 	function LowRankCovMatrix(samples::Array{Array{Float64, 1}, 1})
-		zeromeansamples = Array(Array{Float64, 1}, length(samples))
+		zeromeansamples = Array{Array{Float64, 1}}(length(samples))
 		means = zeros(Float64, length(samples[1]))
 		for i = 1:length(samples)
 			for j = 1:length(means)
@@ -92,7 +92,7 @@ function At_mul_B!(v::Vector, A::Union{LowRankCovMatrix, PCGALowRankMatrix}, x::
 end
 
 function *(A::PCGALowRankMatrix, x::Vector)
-	result = Array(Float64, size(A, 1))
+	result = Array{Float64}(size(A, 1))
 	A_mul_B!(result, A, x)
 	return result
 end

src/lsqr.jl

@@ -34,7 +34,7 @@ end
 
 function pcgalsqriteration(forwardmodel::Function, s::Vector, X::Vector, xis::Array{Array{Float64, 1}, 1}, R, y::Vector, delta::Float64)
 	p = 1#p = 1 because X is a vector rather than a full matrix
-	paramstorun = Array(Array{Float64, 1}, length(xis) + 3)
+	paramstorun = Array{Array{Float64, 1}}(length(xis) + 3)
 	for i = 1:length(xis)
 		paramstorun[i] = s + delta * xis[i]
 	end
@@ -43,7 +43,7 @@ function pcgalsqriteration(forwardmodel::Function, s::Vector, X::Vector, xis::Ar
 	paramstorun[length(xis) + 3] = s
 	results = pmap(forwardmodel, paramstorun)
 	hs = results[length(xis) + 3]
-	etas = Array(Array{Float64, 1}, length(xis))
+	etas = Array{Array{Float64, 1}}(length(xis))
 	for i = 1:length(xis)
 		etas[i] = (results[i] - hs) / delta
 	end

test/testrpcga.jl

@@ -7,7 +7,7 @@ function simplepcgalowranktest(numetas=10, numobs=20)
 	noiselevels = [1e16, 0.]#none, nuge
 	etagenerators = [zeros, randn]
 	HXgenerators = [zeros, randn]
-	etas = Array(Array{Float64, 1}, numetas)
+	etas = Array{Array{Float64, 1}}(numetas)
 	for noiselevel in noiselevels
 		for etagenerator in etagenerators
 			for HXgenerator in HXgenerators
@@ -49,9 +49,9 @@ function lowrankcovconsistencytest()
 	const M = 100
 	sqrtcovmatrix = randn(M, M)
 	covmatrix = sqrtcovmatrix * sqrtcovmatrix'
-	samples = Array(Array{Float64, 1}, N)
-	onesamples = Array(Float64, N)
-	twosamples = Array(Float64, N)
+	samples = Array{Array{Float64, 1}}(N)
+	onesamples = Array{Float64}(N)
+	twosamples = Array{Float64}(N)
 	for i = 1:N
 		samples[i] = sqrtcovmatrix * randn(M)
 		onesamples[i] = samples[i][1]