Skip to content

Commit

Permalink
Refactor to use _c_column in MOI_wrapper.jl (#265)
Browse files Browse the repository at this point in the history
  • Loading branch information
@odow
odow authored Nov 15, 2023
1 parent 36872c4 commit 35d88fc
Showing 1 changed file with 30 additions and 27 deletions.
57 changes: 30 additions & 27 deletions src/MOI_wrapper.jl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,13 @@
# Use of this source code is governed by an MIT-style license that can be found
# in the LICENSE.md file or at https://opensource.org/licenses/MIT.

"""
_c_column(x::MOI.VariableIndex) --> Cint

Return the 0-indexed `Cint` corressponding to the column of `x`.
"""
_c_column(x::MOI.VariableIndex) = Cint(x.value - 1)

const _SETS = Union{
MOI.LessThan{Float64},
MOI.GreaterThan{Float64},
Expand All @@ -28,7 +35,7 @@ function _canonical_quadratic_reduction(f::MOI.ScalarQuadraticFunction)
end

function _canonical_linear_reduction(terms::Vector{<:MOI.ScalarAffineTerm})
columns = Cint[term.variable.value - 1 for term in terms]
columns = Cint[_c_column(term.variable) for term in terms]
coefficients = Cdouble[term.coefficient for term in terms]
return columns, coefficients
end
Expand All @@ -45,7 +52,7 @@ function _canonical_vector_affine_reduction(f::MOI.VectorAffineFunction)
I, J, V = Cint[], Cint[], Cdouble[]
for t in f.terms
push!(I, t.output_index - 1)
push!(J, t.scalar_term.variable.value - 1)
push!(J, _c_column(t.scalar_term.variable))
push!(V, t.scalar_term.coefficient)
end
return I, J, V
Expand Down Expand Up @@ -384,7 +391,7 @@ function MOI.set(
)
MOI.throw_if_not_valid(model, x)
start = something(value, 0.0)
KN_set_var_primal_init_value(model.inner, x.value - 1, Cdouble(start))
KN_set_var_primal_init_value(model.inner, _c_column(x), Cdouble(start))
return
end

Expand Down Expand Up @@ -463,8 +470,7 @@ function MOI.add_constraint(
end
ub = _check_value(lt.upper)
model.variable_info[x.value].has_upper_bound = true
# By construction, MOI's indexing is the same as KNITRO's indexing.
KN_set_var_upbnd(model.inner, x.value - 1, ub)
KN_set_var_upbnd(model.inner, _c_column(x), ub)
ci = MOI.ConstraintIndex{MOI.VariableIndex,MOI.LessThan{Float64}}(x.value)
model.constraint_mapping[ci] = convert(Cint, x.value)
return ci
Expand Down Expand Up @@ -499,8 +505,7 @@ function MOI.add_constraint(
end
lb = _check_value(gt.lower)
model.variable_info[x.value].has_lower_bound = true
# We assume that MOI's indexing is the same as KNITRO's indexing.
KN_set_var_lobnd(model.inner, x.value - 1, lb)
KN_set_var_lobnd(model.inner, _c_column(x), lb)
ci = MOI.ConstraintIndex{MOI.VariableIndex,MOI.GreaterThan{Float64}}(x.value)
model.constraint_mapping[ci] = convert(Cint, x.value)
return ci
Expand Down Expand Up @@ -538,9 +543,8 @@ function MOI.add_constraint(
ub = _check_value(set.upper)
model.variable_info[x.value].has_lower_bound = true
model.variable_info[x.value].has_upper_bound = true
# We assume that MOI's indexing is the same as KNITRO's indexing.
KN_set_var_lobnd(model.inner, x.value - 1, lb)
KN_set_var_upbnd(model.inner, x.value - 1, ub)
KN_set_var_lobnd(model.inner, _c_column(x), lb)
KN_set_var_upbnd(model.inner, _c_column(x), ub)
ci = MOI.ConstraintIndex{MOI.VariableIndex,MOI.Interval{Float64}}(x.value)
model.constraint_mapping[ci] = convert(Cint, x.value)
return ci
Expand Down Expand Up @@ -578,8 +582,7 @@ function MOI.add_constraint(
end
eqv = _check_value(eq.value)
model.variable_info[x.value].is_fixed = true
# We assume that MOI's indexing is the same as KNITRO's indexing.
KN_set_var_fxbnd(model.inner, x.value - 1, eqv)
KN_set_var_fxbnd(model.inner, _c_column(x), eqv)
ci = MOI.ConstraintIndex{MOI.VariableIndex,MOI.EqualTo{Float64}}(x.value)
model.constraint_mapping[ci] = convert(Cint, x.value)
return ci
Expand Down Expand Up @@ -619,24 +622,23 @@ end
###

function MOI.add_constraint(model::Optimizer, x::MOI.VariableIndex, ::MOI.ZeroOne)
indv = x.value - 1
MOI.throw_if_not_valid(model, x)
model.number_zeroone_constraints += 1
lb, ub = nothing, nothing
if model.variable_info[x.value].has_lower_bound
lb = max(0.0, KN_get_var_lobnd(model.inner, indv))
lb = max(0.0, KN_get_var_lobnd(model.inner, _c_column(x)))
end
if model.variable_info[x.value].has_upper_bound
ub = min(1.0, KN_get_var_upbnd(model.inner, indv))
ub = min(1.0, KN_get_var_upbnd(model.inner, _c_column(x)))
end
KN_set_var_type(model.inner, x.value - 1, KN_VARTYPE_BINARY)
KN_set_var_type(model.inner, _c_column(x), KN_VARTYPE_BINARY)
# Calling `set_var_type` resets variable bounds in KNITRO. To fix, we need
# to restore them after calling `set_var_type`.
if lb !== nothing
KN_set_var_lobnd(model.inner, indv, lb)
KN_set_var_lobnd(model.inner, _c_column(x), lb)
end
if ub !== nothing
KN_set_var_upbnd(model.inner, indv, ub)
KN_set_var_upbnd(model.inner, _c_column(x), ub)
end
return MOI.ConstraintIndex{MOI.VariableIndex,MOI.ZeroOne}(x.value)
end
Expand All @@ -649,7 +651,7 @@ function MOI.add_constraint(model::Optimizer, x::MOI.VariableIndex, set::MOI.Int
_throw_if_solved(model, x, set)
MOI.throw_if_not_valid(model, x)
model.number_integer_constraints += 1
KN_set_var_type(model.inner, x.value - 1, KN_VARTYPE_INTEGER)
KN_set_var_type(model.inner, _c_column(x), KN_VARTYPE_INTEGER)
return MOI.ConstraintIndex{MOI.VariableIndex,MOI.Integer}(x.value)
end

Expand Down Expand Up @@ -880,10 +882,10 @@ function MOI.add_constraint(
_throw_if_solved(model, func, set)
# Add constraints inside KNITRO.
index_con = KN_add_con(model.inner)
indv = [v.value - 1 for v in func.variables]
indv = _c_column.(func.variables)
KN_set_con_upbnd(model.inner, index_con, 0.0)
KN_add_con_linear_struct(model.inner, index_con, indv[1], -1.0)
indexVars = convert.(Cint, indv[2:end])
indexVars = indv[2:end]
nnz = length(indexVars)
indexCoords = Cint[i for i in 0:(nnz-1)]
coefs = ones(Float64, nnz)
Expand All @@ -898,9 +900,8 @@ function MOI.add_constraint(
coefs,
constants,
)
# Add constraints to index.
ci = MOI.ConstraintIndex{typeof(func),typeof(set)}(index_con)
model.constraint_mapping[ci] = convert.(Cint, indv)
model.constraint_mapping[ci] = indv
return ci
end

Expand Down Expand Up @@ -947,7 +948,7 @@ function MOI.add_constraint(
set::MOI.Complements,
)
_throw_if_solved(model, func, set)
indv = Cint[v.value - 1 for v in func.variables]
indv = _c_column.(func.variables)
# Number of complementarity in Knitro is half the dimension of the MOI set
n_comp = div(set.dimension, 2)
# Currently, only complementarity constraints between two variables
Expand Down Expand Up @@ -1025,22 +1026,24 @@ function _add_objective(model::Optimizer, f::MOI.ScalarQuadraticFunction)
I, J, V = _canonical_quadratic_reduction(f)
KN_add_obj_quadratic_struct(model.inner, I, J, V)
columns, coefficients = _canonical_linear_reduction(f)
KN_add_obj_linear_struct(model.inner, columns, coefficients)
nnz = length(columns)
KN_add_obj_linear_struct(model.inner, nnz, columns, coefficients)
KN_add_obj_constant(model.inner, f.constant)
model.objective = nothing
return
end

function _add_objective(model::Optimizer, f::MOI.ScalarAffineFunction)
columns, coefficients = _canonical_linear_reduction(f)
KN_add_obj_linear_struct(model.inner, columns, coefficients)
nnz = length(columns)
KN_add_obj_linear_struct(model.inner, nnz, columns, coefficients)
KN_add_obj_constant(model.inner, f.constant)
model.objective = nothing
return
end

function _add_objective(model::Optimizer, f::MOI.VariableIndex)
KN_add_obj_linear_struct(model.inner, f.value - 1, 1.0)
KN_add_obj_linear_struct(model.inner, 1, [_c_column(f)], [1.0])
model.objective = nothing
return
end
Expand Down

0 comments on commit 35d88fc

Please sign in to comment.