Sparse Matrix multiplied by variable vector with Julia v1.0.5 and 1.3

[kaarthiksundar]

The code that worked with Julia 1.0.5 does not work with Julia 1.3.

# in julia 1.0.5
julia> using JuMP, GLPK, SparseArrays
julia> glpk_optimizer = JuMP.with_optimizer(GLPK.Optimizer, tm_lim = 100.0, msg_lev = GLPK.OFF)
julia> m = JuMP.Model(glpk_optimizer)
julia> num_variables = 8; 
julia> JuMP.@variable(m, y[i=1:num_variables])
julia> I = [1, 2, 1, 2, 1, 2, 1, 2, 1, 2]; 
julia> J = [1, 2, 3, 3, 4, 4, 5, 5, 6, 6]; 
julia> V = [1, 1, -0.333333, -1.0, -0.666667, 0.0, -1.0, -1.0, -1.0, -1.0];
julia> A = sparse(I, J, V, 8, 2)
julia> A * y
2-element Array{GenericAffExpr{Float64,VariableRef},1}:
 y[1] - 0.33333333333333337 y[3] - 0.6666666666666666 y[4] - y[5] - y[6]
 y[2] - y[3] - y[5] - y[6]   

# in julia 1.3
julia> A * y
ERROR: MethodError: no method matching one(::Type{GenericAffExpr{C,VariableRef} where C})
Closest candidates are:
  one(::Type{Missing}) at missing.jl:103
  one(::BitArray{2}) at bitarray.jl:400
  one(::Missing) at missing.jl:100
  ...
Stacktrace:
 [1] *(::SparseMatrixCSC{Real,Int64}, ::Array{VariableRef,1}) at /Users/julia/buildbot/worker/package_macos64/build/usr/share/julia/stdlib/v1.3/SparseArrays/src/linalg.jl:54
 [2] top-level scope at REPL[23]:1

Any help is appreciated.

[odow]

You likely need to do A * y in a macro, i.e.,

@expression(model, A * y)

I can't reproduce this on Julia 1.3 and JuMP#master.

julia> model = Model()
A JuMP Model
Feasibility problem with:
Variables: 0
@aModel mode: AUTOMATIC
CachingOptimizer state: NO_OPTIMIZER
Solver name: No optimizer attached.

julia> @variable(model, y[i=1:8])
8-element Array{VariableRef,1}:
 y[1]
 y[2]
 y[3]
 y[4]
 y[5]
 y[6]
 y[7]
 y[8]

julia> I = [1, 2, 1, 2, 1, 2, 1, 2, 1, 2];

julia> J = [1, 2, 3, 3, 4, 4, 5, 5, 6, 6];

julia> V = [1, 1, -0.333333, -1.0, -0.666667, 0.0, -1.0, -1.0, -1.0, -1.0];  

julia> A = sparse(I, J, V, 2, 8)  # Is your code correct for this line? I reversed 8, 2.
2×8 SparseMatrixCSC{Float64,Int64} with 10 stored entries:
  [1, 1]  =  1.0
  [2, 2]  =  1.0
  [1, 3]  =  -0.333333
  [2, 3]  =  -1.0
  [1, 4]  =  -0.666667
  [2, 4]  =  0.0
  [1, 5]  =  -1.0
  [2, 5]  =  -1.0
  [1, 6]  =  -1.0
  [2, 6]  =  -1.0

julia> A * y
2-element Array{GenericAffExpr{Float64,VariableRef},1}:
 y[1] - 0.333333 y[3] - 0.666667 y[4] - y[5] - y[6]
 y[2] - y[3] - y[5] - y[6]                         

Much of the default sparse fallbacks don't work well with JuMP because we don't define one() and zero() for GenericAffExpr.

See: #1151

[blegat]

This might be because the type of the sparse matrix is SparseMatrixCSC{Real,Int64} which is not concrete, try SparseMatrixCSC{Float64,Int64}.

[odow]

Still can't reproduce this on JuMP 0.21.2, even with V = Real[ ... ].

julia> A * y
2-element Array{GenericAffExpr{Float64,VariableRef},1}:
 y[1] - 0.333333 y[3] - 0.666667 y[4] - y[5] - y[6]
 y[2] - y[3] - y[5] - y[6]                         

julia> typeof(A)
SparseMatrixCSC{Real,Int64}

julia> typeof(y)
Array{VariableRef,1}

Can you reproduce? If not, I will close.

[odow]

Closing as not reproducible. Please re-open if you can find a reproducible example on JuMP#master.