Merge pull request #1349 from JuliaOpt/bl/fix0.7

Fix various Julia v0.7 warnings

src/Derivatives/coloring.jl

@@ -1,5 +1,7 @@
 module Coloring
 
+using Compat
+
 import DataStructures
 
 include("topological_sort.jl")
@@ -82,16 +84,16 @@ function gen_adjlist(I,J,nel)
         adjcount[i] += 1
         adjcount[j] += 1
     end
-    offsets = Array{Int}(nel+1)
+    offsets = Array{Int}(undef, nel+1)
     offsets[1] = 1
     for k in 1:nel
         offsets[k+1] = offsets[k] + adjcount[k]
     end
     fill!(adjcount,0)
 
-    edges = Array{MyPair{Int}}(n_edges)
-    adjlist = Array{Int}(offsets[nel+1]-1)
-    edgeindex = Array{Int}(length(adjlist))
+    edges = Array{MyPair{Int}}(undef, n_edges)
+    adjlist = Array{Int}(undef, offsets[nel+1]-1)
+    edgeindex = Array{Int}(undef, length(adjlist))
     edge_count = 0
 
     for k in 1:length(I)
@@ -176,7 +178,7 @@ function acyclic_coloring(g::UndirectedGraph)
     end
     num_colors = 0
     forbiddenColors = Int[]
-    firstNeighbor = Array{Edge}(0)
+    firstNeighbor = Edge[]
     firstVisitToTree = fill(Edge(0,0,0),num_edges(g))
     color = fill(0, num_vertices(g))
     # disjoint set forest of edges in the graph
@@ -270,7 +272,7 @@ function recovery_preprocess(g::UndirectedGraph,color,num_colors, local_indices)
     twocolorindex = zeros(Int32,num_colors, num_colors)
     seen_twocolors = 0
     # count of edges in each subgraph
-    edge_count = Array{Int}(0)
+    edge_count = Int[]
     for k in 1:length(g.edges)
         e = g.edges[k]
         u = e.first
@@ -286,9 +288,9 @@ function recovery_preprocess(g::UndirectedGraph,color,num_colors, local_indices)
         edge_count[idx] += 1
     end
     # edges sorted by twocolor subgraph
-    sorted_edges = Array{Vector{MyPair{Int}}}(seen_twocolors)
+    sorted_edges = Array{Vector{MyPair{Int}}}(undef, seen_twocolors)
     for idx in 1:seen_twocolors
-        sorted_edges[idx] = Array{MyPair{Int}}(0)
+        sorted_edges[idx] = MyPair{Int}[]
         sizehint!(sorted_edges[idx],edge_count[idx])
     end
 
@@ -303,10 +305,10 @@ function recovery_preprocess(g::UndirectedGraph,color,num_colors, local_indices)
     end
 
     # list of unique vertices in each twocolor subgraph
-    vertexmap = Array{Vector{Int}}(seen_twocolors)
+    vertexmap = Array{Vector{Int}}(undef, seen_twocolors)
 
-    postorder = Array{Vector{Int}}(seen_twocolors)
-    parents = Array{Vector{Int}}(seen_twocolors)
+    postorder = Array{Vector{Int}}(undef, seen_twocolors)
+    parents = Array{Vector{Int}}(undef, seen_twocolors)
 
     # temporary lookup map from global index to subgraph index
     revmap = zeros(Int,num_vertices(g))
@@ -342,7 +344,7 @@ function recovery_preprocess(g::UndirectedGraph,color,num_colors, local_indices)
         end
 
         # set up offsets for adjlist
-        offset = Array{Int}(length(vlist)+1)
+        offset = Array{Int}(undef, length(vlist)+1)
         offset[1] = 1
         for k in 1:length(vlist)
             offset[k+1] = offset[k] + adjcount[vlist[k]]
@@ -351,7 +353,7 @@ function recovery_preprocess(g::UndirectedGraph,color,num_colors, local_indices)
         # adjlist for node u in twocolor idx starts at
         # vec[offset[u]]
         # u has global index vlist[u]
-        vec = Array{Int}(offset[length(vlist)+1]-1)
+        vec = Array{Int}(undef, offset[length(vlist)+1]-1)
 
         # now fill in
         for k in 1:length(my_edges)
@@ -471,7 +473,7 @@ end
 export hessian_color_preprocess
 
 # allocate a seed matrix
-seed_matrix(rinfo::RecoveryInfo) = Array{Float64}(length(rinfo.local_indices),rinfo.num_colors)
+seed_matrix(rinfo::RecoveryInfo) = Array{Float64}(undef,length(rinfo.local_indices),rinfo.num_colors)
 
 export seed_matrix
 
@@ -512,7 +514,7 @@ function recover_from_matmat!(V, R, rinfo::RecoveryInfo, stored_values)
         vmap = rinfo.vertexmap[t]
         order = rinfo.postorder[t]
         parent = rinfo.parents[t]
-        stored_values[1:length(order)] = 0.0
+        stored_values[1:length(order)] .= 0.0
 
         @inbounds for z in 1:length(order)
             v = order[z]

src/Derivatives/conversion.jl

@@ -66,8 +66,8 @@ export expr_to_nodedata
 # edges leaving any node (i.e., the children)
 function adjmat(nd::Vector{NodeData})
     len = length(nd)
-    I = Array{Int}(len)
-    J = Array{Int}(len)
+    I = Vector{Int}(undef, len)
+    J = Vector{Int}(undef, len)
     realnz = 0
     for nz in 1:len
         par = nd[nz].parent

src/Derivatives/linearity.jl

@@ -8,7 +8,7 @@ export CONSTANT, LINEAR, PIECEWISE_LINEAR, NONLINEAR
 
 function classify_linearity(nd::Vector{NodeData},adj,subexpression_linearity)
 
-    linearity = Array{Linearity}(length(nd))
+    linearity = Array{Linearity}(undef, length(nd))
 
     # do a forward pass through the graph, which is reverse order of nd
 

src/Derivatives/subexpressions.jl

@@ -55,7 +55,7 @@ function order_subexpressions(main_expressions::Vector{Vector{NodeData}},subexpr
     end
     N = nsub+length(main_expressions)
     sp = sparse(I,J,ones(length(I)),N,N)
-    cmap = Array{Int}(N)
+    cmap = Vector{Int}(undef, N)
     order = reverse(Coloring.reverse_topological_sort_by_dfs(sp.rowval,sp.colptr,N,cmap)[1])
     # remove the subexpressions which never appear anywhere
     # and the indices of the main expressions

src/Derivatives/topological_sort.jl

@@ -31,7 +31,7 @@ mutable struct TopologicalSortVisitor
     parents::Vector{Int}
 
     function TopologicalSortVisitor(n::Int)
-        vs = Array{Int}(0)
+        vs = Int[]
         sizehint!(vs, n)
         new(vs, zeros(Int,n))
     end

src/Derivatives/types.jl

@@ -70,7 +70,7 @@ mutable struct UserOperatorRegistry
     univariate_operator_fprimeprime::Vector{Any}
 end
 
-UserOperatorRegistry() = UserOperatorRegistry(Dict{Symbol,Int}(),Vector{MOI.AbstractNLPEvaluator}(0),Dict{Symbol,Int}(),[],[],[])
+UserOperatorRegistry() = UserOperatorRegistry(Dict{Symbol,Int}(),MOI.AbstractNLPEvaluator[],Dict{Symbol,Int}(),[],[],[])
 
 # we use the MathOptInterface NLPEvaluator interface, where the
 # operator takes the place of the objective function.

src/JuMPArray.jl

@@ -108,7 +108,7 @@ Base.isassigned(A::JuMPArray{T,N}, idx...) where {T,N} = length(idx) == N && all
 # For ambiguity
 Base.isassigned(A::JuMPArray{T,N}, idx::Int...) where {T,N} = length(idx) == N && all(t -> haskey(A.lookup[t[1]], t[2]), enumerate(idx))
 
-Base.eachindex(A::JuMPArray) = CartesianRange(size(A.data))
+Base.eachindex(A::JuMPArray) = CartesianIndices(size(A.data))
 
 # TODO: similar
 
@@ -160,7 +160,7 @@ function Base.show_nd(io::IO, a::JuMPArray, print_matrix::Function, label_slices
     end
     tailinds = Base.tail(Base.tail(indices(a.data)))
     nd = ndims(a)-2
-    for I in CartesianRange(tailinds)
+    for I in CartesianIndices(tailinds)
         idxs = I.I
         if limit
             for i = 1:nd

src/macros.jl

@@ -21,13 +21,13 @@ function buildrefsets(expr::Expr, cname)
     # Creating an indexed set of refs
     refcall = Expr(:ref, cname)
     if isexpr(c, :typed_vcat) || isexpr(c, :ref)
-        shift!(c.args)
+        popfirst!(c.args)
     end
     condition = :()
     if isexpr(c, :vcat) || isexpr(c, :typed_vcat)
         if isexpr(c.args[1], :parameters)
             @assert length(c.args[1].args) == 1
-            condition = shift!(c.args).args[1]
+            condition = popfirst!(c.args).args[1]
         else
             condition = pop!(c.args)
         end
@@ -1046,7 +1046,7 @@ macro variable(args...)
         x = gensym()
         anon_singleton = true
     else
-        x = shift!(extra)
+        x = popfirst!(extra)
         if x in [:Int,:Bin,:PSD]
             _error("Ambiguous variable name $x detected. Use the \"category\" keyword argument to specify a category for an anonymous variable.")
         end

src/variables.jl

@@ -112,7 +112,7 @@ MOI.isvalid(m::Model, v::VariableRef) = (v.m === m) && MOI.isvalid(m.moibackend,
 # The default hash is slow. It's important for the performance of AffExpr to
 # define our own.
 # https://github.com/JuliaOpt/MathOptInterface.jl/issues/234#issuecomment-366868878
-Base.hash(v::VariableRef, h::UInt) = hash(object_id(v.m), hash(v.index.value, h))
+Base.hash(v::VariableRef, h::UInt) = hash(objectid(v.m), hash(v.index.value, h))
 Base.isequal(v1::VariableRef, v2::VariableRef) = v1.m === v2.m && v1.index == v2.index
 
 

test/containers.jl

@@ -25,7 +25,7 @@ function containermatches(c1::AbstractArray,c2::AbstractArray)
 end
 
 function containermatches(c1::JuMPArray,c2::JuMPArray)
-    return typeof(c1) == typeof(c2) && indices(c1) == indices(c2)
+    return typeof(c1) == typeof(c2) && Compat.axes(c1) == Compat.axes(c2)
 end
 
 containermatches(c1::Dict, c2::Dict) = (eltype(c1) == eltype(c2))
@@ -79,7 +79,7 @@ containermatches(c1, c2) = false
     @test_throws ErrorException @dummycontainer([i=1:10; iseven(i)], JuMPArray)
     @test containermatches(@dummycontainer([i=1:10; iseven(i)], Dict), Dict{Any,Bool}())
 
-    # Dependent indices
+    # Dependent axes
     @test containermatches(@dummycontainer([i=1:10, j=1:i], Auto), Dict{Any,Bool}())
     @test_throws ErrorException @dummycontainer([i=1:10, j=1:i], Array)
     @test_throws ErrorException @dummycontainer([i=1:10, j=1:i], JuMPArray)
@@ -99,7 +99,7 @@ end
     @test A[3,1,1,1,1] == 2.0
     @test isassigned(A, 2)
     @test !isassigned(A, 1)
-    @test length.(indices(A)) == (2,)
+    @test length.(Compat.axes(A)) == (2,)
     B = plus1.(A)
     @test B[2] == 2.0
     @test B[3] == 3.0
@@ -113,7 +113,7 @@ And data, a 2-element Array{Float64,1}:
     A = @inferred JuMPArray([1.0,2.0], s2)
     @test @inferred A[:a] == 1.0
     @test A[:b] == 2.0
-    @test length.(indices(A)) == (2,)
+    @test length.(Compat.axes(A)) == (2,)
     B = plus1.(A)
     @test B[:a] == 2.0
     @test B[:b] == 3.0
@@ -125,7 +125,7 @@ And data, a 2-element Array{Float64,1}:
  3.0"""
 
     A = @inferred JuMPArray([1 2; 3 4], s1, s2)
-    @test length.(indices(A)) == (2,2)
+    @test length.(Compat.axes(A)) == (2,2)
     @test @inferred A[2,:a] == 1
     @test A[3,:a] == 3
     @test A[2,:b] == 2

test/derivatives.jl

@@ -201,7 +201,7 @@ function test_linearity(ex,testval,IJ = [],indices=[])
     linearity = classify_linearity(nd,adj,[])
     @test linearity[1] == testval
     idxset = Coloring.IndexedSet(100)
-    edgelist = compute_hessian_sparsity(nd,adj,linearity,idxset,Array{Set{Tuple{Int,Int}}}(0), Array{Vector{Int}}(0))
+    edgelist = compute_hessian_sparsity(nd,adj,linearity,idxset,Array{Set{Tuple{Int,Int}}}(undef, 0), Array{Vector{Int}}(undef, 0))
     if linearity[1] != NONLINEAR
         @test length(edgelist) == 0
     elseif length(IJ) > 0