Skip to content

Commit

Permalink
Various improvements to MOI wrapper (#291)
Browse files Browse the repository at this point in the history
  • Loading branch information
@odow
odow authored Nov 3, 2021
1 parent b0aa7de commit b0a499d
Showing 1 changed file with 16 additions and 107 deletions.
123 changes: 16 additions & 107 deletions src/MOI_wrapper.jl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,30 +2,6 @@ import MathOptInterface

const MOI = MathOptInterface

mutable struct _VariableInfo
lower_bound::Float64
has_lower_bound::Bool
lower_bound_dual_start::Union{Nothing,Float64}
upper_bound::Float64
has_upper_bound::Bool
upper_bound_dual_start::Union{Nothing,Float64}
is_fixed::Bool
start::Union{Nothing,Float64}
end

function _VariableInfo()
return _VariableInfo(
-Inf,
false,
nothing,
Inf,
false,
nothing,
false,
nothing,
)
end

mutable struct _ConstraintInfo{F,S}
func::F
set::S
Expand All @@ -35,18 +11,18 @@ end
_ConstraintInfo(func, set) = _ConstraintInfo(func, set, nothing)

mutable struct Optimizer <: MOI.AbstractOptimizer
inner::Union{IpoptProblem,Nothing}
inner::Union{Nothing,IpoptProblem}
variables::MOI.Utilities.VariablesContainer{Float64}
variable_primal_start::Vector{Union{Nothing,Float64}}
variable_lower_start::Vector{Union{Nothing,Float64}}
variable_upper_start::Vector{Union{Nothing,Float64}}
nlp_data::MOI.NLPBlockData
sense::MOI.OptimizationSense
objective::Union{
Nothing,
MOI.VariableIndex,
MOI.ScalarAffineFunction{Float64},
MOI.ScalarQuadraticFunction{Float64},
Nothing,
}
linear_le_constraints::Vector{
_ConstraintInfo{
Expand Down Expand Up @@ -89,41 +65,16 @@ mutable struct Optimizer <: MOI.AbstractOptimizer
end

### _EmptyNLPEvaluator
### This AbstractNLPEvaluator is used when the user hasn't defined one.

struct _EmptyNLPEvaluator <: MOI.AbstractNLPEvaluator end

MOI.features_available(::_EmptyNLPEvaluator) = [:Grad, :Jac, :Hess]

MOI.initialize(::_EmptyNLPEvaluator, features) = nothing

MOI.eval_objective(::_EmptyNLPEvaluator, x) = NaN

function MOI.eval_constraint(::_EmptyNLPEvaluator, g, x)
@assert length(g) == 0
return
end

function MOI.eval_objective_gradient(::_EmptyNLPEvaluator, g, x)
fill!(g, 0.0)
return
end

MOI.initialize(::_EmptyNLPEvaluator, ::Any) = nothing
MOI.eval_constraint(::_EmptyNLPEvaluator, g, x) = nothing
MOI.jacobian_structure(::_EmptyNLPEvaluator) = Tuple{Int64,Int64}[]

MOI.hessian_lagrangian_structure(::_EmptyNLPEvaluator) = Tuple{Int64,Int64}[]

function MOI.eval_constraint_jacobian(::_EmptyNLPEvaluator, J, x)
@assert length(J) == 0
return
end

function MOI.eval_hessian_lagrangian(::_EmptyNLPEvaluator, H, x, σ, μ)
@assert length(H) == 0
return
end

_empty_nlp_data() = MOI.NLPBlockData([], _EmptyNLPEvaluator(), false)
MOI.eval_constraint_jacobian(::_EmptyNLPEvaluator, J, x) = nothing
MOI.eval_hessian_lagrangian(::_EmptyNLPEvaluator, H, x, σ, μ) = nothing

function Optimizer(; kwargs...)
if length(kwargs) > 0
Expand All @@ -140,7 +91,7 @@ function Optimizer(; kwargs...)
Float64[],
Float64[],
Float64[],
_empty_nlp_data(),
MOI.NLPBlockData([], _EmptyNLPEvaluator(), false),
MOI.FEASIBILITY_SENSE,
nothing,
[],
Expand All @@ -166,7 +117,7 @@ function MOI.empty!(model::Optimizer)
empty!(model.variable_primal_start)
empty!(model.variable_lower_start)
empty!(model.variable_upper_start)
model.nlp_data = _empty_nlp_data()
model.nlp_data = MOI.NLPBlockData([], _EmptyNLPEvaluator(), false)
model.sense = MOI.FEASIBILITY_SENSE
model.objective = nothing
empty!(model.linear_le_constraints)
Expand Down Expand Up @@ -555,10 +506,8 @@ function MOI.set(
},
S,
}
MOI.throw_if_not_valid(model, ci)
constraints = _constraints(model, F, S)
if !(1 <= ci.value <= length(constraints))
throw(MOI.InvalidIndex(ci))
end
constraints[ci.value].dual_start = value
return
end
Expand All @@ -574,10 +523,8 @@ function MOI.get(
},
S,
}
MOI.throw_if_not_valid(model, ci)
constraints = _constraints(model, F, S)
if !(1 <= ci.value <= length(constraints))
throw(MOI.InvalidIndex(ci))
end
return constraints[ci.value].dual_start
end

Expand Down Expand Up @@ -746,8 +693,8 @@ function MOI.supports(
::MOI.ObjectiveFunction{
<:Union{
MOI.VariableIndex,
MOI.ScalarAffineFunction,
MOI.ScalarQuadraticFunction,
MOI.ScalarAffineFunction{Float64},
MOI.ScalarQuadraticFunction{Float64},
},
},
)
Expand All @@ -761,8 +708,8 @@ function MOI.set(
) where {
F<:Union{
MOI.VariableIndex,
MOI.ScalarAffineFunction,
MOI.ScalarQuadraticFunction,
MOI.ScalarAffineFunction{Float64},
MOI.ScalarQuadraticFunction{Float64},
},
}
_check_inbounds(model, func)
Expand Down Expand Up @@ -850,41 +797,9 @@ function _nlp_constraint_offset(model::Optimizer)
return x + length(model.quadratic_eq_constraints)
end

_eval_function(::Nothing, x) = 0.0

function _eval_function(var::MOI.VariableIndex, x)
return x[var.value]
end
_eval_function(::Nothing, ::Any) = 0.0

function _eval_function(aff::MOI.ScalarAffineFunction, x)
function_value = aff.constant
for term in aff.terms
# Note the implicit assumtion that VariableIndex values match up with
# x indices. This is valid because in this wrapper ListOfVariableIndices
# is always [1, ..., NumberOfVariables].
function_value += term.coefficient * x[term.variable.value]
end
return function_value
end

function _eval_function(quad::MOI.ScalarQuadraticFunction, x)
function_value = quad.constant
for term in quad.affine_terms
function_value += term.coefficient * x[term.variable.value]
end
for term in quad.quadratic_terms
row_idx = term.variable_1
col_idx = term.variable_2
coefficient = term.coefficient
if row_idx == col_idx
function_value +=
0.5 * coefficient * x[row_idx.value] * x[col_idx.value]
else
function_value += coefficient * x[row_idx.value] * x[col_idx.value]
end
end
return function_value
end
_eval_function(f, x) = MOI.Utilities.eval_variables(xi -> x[xi.value], f)

### Eval_F_CB

Expand Down Expand Up @@ -1414,9 +1329,6 @@ function MOI.get(model::Optimizer, attr::MOI.DualStatus)
# Solutions are only guaranteed to satisfy the "acceptable" convergence
# tolerances.
return MOI.NEARLY_FEASIBLE_POINT
elseif status == :Infeasible_Problem_Detected
# TODO: What is the interpretation of the dual in this case?
return MOI.UNKNOWN_RESULT_STATUS
else
return MOI.UNKNOWN_RESULT_STATUS
end
Expand All @@ -1435,9 +1347,6 @@ end

### MOI.VariablePrimal

# TODO: This is a bit off, because the variable primal should be available
# only after a solve. If model.inner is initialized but we haven't solved, then
# the primal values we return do not have the intended meaning.
function MOI.get(
model::Optimizer,
attr::MOI.VariablePrimal,
Expand Down

2 comments on commit b0a499d

@odow
Copy link
Member Author

@odow odow commented on b0a499d Dec 3, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JuliaRegistrator register

@JuliaRegistrator
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Registration pull request created: JuliaRegistries/General/49820

After the above pull request is merged, it is recommended that a tag is created on this repository for the registered package version.

This will be done automatically if the Julia TagBot GitHub Action is installed, or can be done manually through the github interface, or via:

git tag -a v0.9.0 -m "<description of version>" b0a499d00508337e5da87f036205c838b4e40271
git push origin v0.9.0

Please sign in to comment.