ortools-clone/ortools/fsharp/OrTools.FSharp/OrTools.fs

namespace Google.OrTools

module FSharp =

  open System
  open Google.OrTools.Algorithms
  open Google.OrTools.LinearSolver

  type Goal =
    /// Maximize the Objective Function
    | Maximize
    /// Minimize the Objective Function
    | Minimize

  /// Linear Programming Algorithm
  type LinearProgramming =
    /// Coin-Or (Recommended default)
    | CLP
    /// GNU Linear Programming Kit
    | GLPK
    /// Google Linear Optimization
    | GLOP
    /// Gurobi Optimizer
    | GUROBI
    /// IBM CPLEX
    | CPLEX
    /// Solver Name
    override this.ToString() =
      match this with
      | CLP -> "CLP_LINEAR_PROGRAMMING"
      | GLPK -> "GLPK_LINEAR_PROGRAMMING"
      | GLOP -> "GLOP_LINEAR_PROGRAMMING"
      | GUROBI -> "GUROBI_LINEAR_PROGRAMMING"
      | CPLEX -> "CPLEX_LINEAR_PROGRAMMING"
    /// Solver Id
    member this.Id =
      match this with
      | CLP -> 0
      | GLPK -> 1
      | GLOP -> 2
      | GUROBI -> 6
      | CPLEX -> 10

  /// Integer Programming Algoritm
  type IntegerProgramming =
      /// Solving Constraint Integer Programs (Recommended default)
      | SCIP
      /// GNU Linear Programming Kit
      | GLPK
      /// Coin-Or Branch and Cut
      | CBC
      /// Gurobi Optimizer
      | GUROBI
      /// IBM CPLEX
      | CPLEX
      /// Binary Optimizer
      | BOP
      /// Solver Name
      override this.ToString() =
        match this with
        | SCIP -> "SCIP_MIXED_INTEGER_PROGRAMMING"
        | GLPK -> "GLPK_MIXED_INTEGER_PROGRAMMING"
        | CBC -> "CBC_MIXED_INTEGER_PROGRAMMING"
        | GUROBI -> "GUROBI_MIXED_INTEGER_PROGRAMMING"
        | CPLEX -> "CPLEX_MIXED_INTEGER_PROGRAMMING"
        | BOP -> "BOP_INTEGER_PROGRAMMING"
      /// Solver Id
      member this.Id =
        match this with
        | SCIP -> 3
        | GLPK -> 4
        | CBC -> 5
        | GUROBI -> 7
        | CPLEX -> 11
        | BOP -> 12

  type LinearSolverAlgorithm =
    | LP of LinearProgramming
    | IP of IntegerProgramming

  /// Max Flow Solver Result
  type MaximumFlowResult =
    | Optimal
    | IntegerOverflow
    | BadInput
    | BadResult
    member this.Id =
        match this with
        | Optimal -> 0
        | IntegerOverflow -> 1
        | BadInput -> 2
        | BadResult -> 3

  /// Minimum Cost Flow Result
  type MinimumCostFlowResult =
    | NotSolved
    | Optimal
    | Feasible
    | Infeasible
    | Unbalanced
    | BadResult
    | BadCostRange
    member this.Id =
      match this with
      | NotSolved -> 0
      | Optimal -> 1
      | Feasible -> 2
      | Infeasible -> 3
      | Unbalanced -> 4
      | BadResult -> 5
      | BadCostRange -> 6


  /// Knapsack Solver Algorithm
  type KnapsackSolverAlgorithm =
    | BruteForce
    | SixtyFourItems
    | DynamicProgramming
    | MultidimensionCBC
    | MultidimensionGLPK
    | MultidimensionBranchAndBound
    | MultidimensionSCIP
    /// Solver Id
    member this.Id =
      match this with
      | BruteForce -> 0
      | SixtyFourItems -> 1
      | DynamicProgramming -> 2
      | MultidimensionCBC -> 3
      | MultidimensionGLPK -> 4
      | MultidimensionBranchAndBound -> 5
      | MultidimensionSCIP -> 6

  let knapsackSolve (name: string) (solverAlgorithm:KnapsackSolverAlgorithm) (profits:int64 list) (weights:int64 list) (capacities:int64 list) =
    // extract the specific algorithm so its Id can be used to create solver
    let algorithm =
      match solverAlgorithm with
      | MultidimensionBranchAndBound ->
          MultidimensionBranchAndBound.Id
      | BruteForce ->
          BruteForce.Id
      | SixtyFourItems ->
          SixtyFourItems.Id
      | DynamicProgramming ->
          DynamicProgramming.Id
      | MultidimensionCBC ->
          MultidimensionCBC.Id
      | MultidimensionGLPK ->
          MultidimensionGLPK.Id
      | MultidimensionSCIP ->
          MultidimensionSCIP.Id

    let solver = new KnapsackSolver(algorithm, name)

    // transform lists to compatible structures for C++ Solver
    let profits = new KInt64Vector( List.toArray profits )

    let weights =
      let tempVector = new KInt64VectorVector(1)
      let tempWeights = new KInt64Vector(List.toArray weights)
      tempVector.Add(tempWeights)
      tempVector

    let capacities = new KInt64Vector (List.toArray capacities)

    solver.Init(profits, weights, capacities)
    solver

  type SolverOpts = {
    /// Name of the solver
    SolverName: string;
    /// Linear objective function vector
    ObjectiveFunction: float list;
    /// Matrix for linear inequality constraints
    ConstraintMatrix: float list list option;
    /// Upper Bound vector for linear inequality constraints
    ConstraintVectorUpperBound: float list option;
    /// Lower Bound vector for linear inequality constraints
    ConstraintVectorLowerBound: float list option;
    /// Vector of variable upper bounds
    VariableUpperBound: float list;
    /// Vector of variable lower bounds
    VariableLowerBound: float list;
    /// Matrix for linear equality constraints
    ConstraintMatrixEq: float list list option;
    /// Vector for linear equality constraints
    ConstraintVectorEq: float list option;
    /// Solver Algorithm to use
    SolverAlgorithm: LinearSolverAlgorithm;
    /// Solver Goal (Minimize/Maximize)
    SolverGoal: Goal
  } with
    member this.Name(nm:string)=
      {this with SolverName = nm}
    member this.Objective(objVector:float list) =
      {this with ObjectiveFunction = objVector}
    member this.Matrix(mat:float list list) =
      {this with ConstraintMatrix = Some(mat)}
    member this.VectorUpperBound(vec:float list) =
      {this with ConstraintVectorUpperBound = Some(vec)}
    member this.VectorLowerBound(vec:float list) =
      {this with ConstraintVectorLowerBound = Some(vec)}
    member this.MatrixEq(mat:float list list) =
      {this with ConstraintMatrixEq = Some(mat)}
    member this.VectorEq(vec:float list) =
      {this with ConstraintVectorEq = Some(vec)}
    member this.VarUpperBound(ub:float list) =
      {this with VariableUpperBound = ub}
    member this.VarLowerBound(lb:float list) =
      {this with VariableLowerBound = lb}
    member this.Algorithm(algo:LinearSolverAlgorithm) =
      {this with SolverAlgorithm = algo}
    member this.Goal(goal:Goal) =
      {this with SolverGoal = goal}
    static member Default =
      {
        SolverName = "Solver";
        ObjectiveFunction = [];
        ConstraintMatrix = None;
        ConstraintVectorUpperBound = None;
        ConstraintVectorLowerBound = None;
        VariableUpperBound = [];
        VariableLowerBound = [];
        ConstraintMatrixEq = None;
        ConstraintVectorEq = None;
        SolverAlgorithm = LP CLP
        SolverGoal = Maximize
      }

  let lpSolve (solverOptions:SolverOpts) =
    // extract the specific algorithm so its Id can be used to create solver
    let algorithm =
      match solverOptions.SolverAlgorithm with
      | LP lp -> lp.Id
      | IP ip -> ip.Id

    let solver = new Solver(solverOptions.SolverName, algorithm)

    // Detect errors on required parameters
    if (solverOptions.ConstraintMatrix.IsNone && solverOptions.ConstraintVectorUpperBound.IsNone && solverOptions.ConstraintVectorLowerBound.IsNone)
      && (solverOptions.ConstraintMatrixEq.IsNone && solverOptions.ConstraintVectorEq.IsNone) then
      failwith "Must provide at least one Matrix/Vector pair for inequality/equality contraints"

    if solverOptions.ObjectiveFunction.IsEmpty then
      failwith "Objective function cannot be empty"

    if solverOptions.VariableUpperBound.IsEmpty then
      failwith "Variable upper bound values cannot be empty"

    if solverOptions.VariableLowerBound.IsEmpty then
      failwith "Variable lower bound values cannot be empty"

    // check inequality matrix dimensions against bounds
    match (solverOptions.ConstraintMatrix, solverOptions.ConstraintVectorLowerBound, solverOptions.ConstraintVectorUpperBound) with
    | _, Some lb, Some ub when lb.Length <> ub.Length ->
      failwithf "Constraint vector dimensions should be equal.\nLower Bound Length: %i\nUpper Bound Length: %i" lb.Length ub.Length
    | _ -> ()

    // Check Objective function dimensions with variable bounds dimensions
    match (solverOptions.ObjectiveFunction.Length, solverOptions.VariableLowerBound.Length, solverOptions.VariableUpperBound.Length) with
    | _ , lb, ub when lb <> ub ->
      failwithf "Variable vector dimensions should be equal.\nLower Bound Length: %i\nUpper Bound Length: %i" lb ub
    | obj, _ , ub when obj <> ub ->
      failwithf "Variable vector dimensions should be equal.\nUpper Bound Length: %i\nObjective Function Length: %i" ub obj
    | obj, lb, _ when obj <> lb ->
      failwithf "Variable vector dimensions should be equal.\nLower Bound Length: %i\nObjective Function Length: %i" lb obj
    | _ -> ()

    // Variables
    let vars =
      match solverOptions.SolverAlgorithm with
        | LP lp ->
            [ for i in 0 .. (solverOptions.VariableLowerBound.Length-1) -> solver.MakeNumVar(solverOptions.VariableLowerBound.[i], solverOptions.VariableUpperBound.[i], (sprintf "var[%i]" i ) ) ]
        | IP ip ->
            [ for i in 0 .. (solverOptions.VariableLowerBound.Length-1) -> solver.MakeIntVar(solverOptions.VariableLowerBound.[i], solverOptions.VariableUpperBound.[i], (sprintf "var[%i]" i ) ) ]

    // Constraints
    let cols = [ for i in 0 .. (solverOptions.VariableLowerBound.Length-1) -> i ]   // generate column index selectors

    // Inequality Constraints
    match (solverOptions.ConstraintMatrix, solverOptions.ConstraintVectorLowerBound, solverOptions.ConstraintVectorUpperBound) with
    | (None, Some lb, Some ub) ->
        failwithf "Matrix for Equality Constraints undefined."
    | (Some mat, None, Some ub) ->
        for row = 0 to (ub.Length-1) do
            // generate constraint operands based on indices
            let constraintOperands = List.map (fun c ->  vars.[c] * mat.[c].[row]) cols
            let linearExp = List.reduce (+) constraintOperands

            // create the constraint
            let rangeConstraint = RangeConstraint(linearExp, Double.NegativeInfinity, ub.[row])
            solver.Add(rangeConstraint) |> ignore
    | (Some mat, Some lb, None) ->
        for row = 0 to (lb.Length-1) do
            // generate constraint operands based on indices
            let constraintOperands = List.map (fun c ->  vars.[c] * mat.[c].[row]) cols
            let linearExp = List.reduce (+) constraintOperands

            // create the constraint
            let rangeConstraint = RangeConstraint(linearExp, lb.[row], Double.PositiveInfinity)
            solver.Add(rangeConstraint) |> ignore
    | (Some mat, Some lb, Some ub) ->
        for row = 0 to (ub.Length-1) do
            // generate constraint operands based on indices
            let constraintOperands = List.map (fun c ->  vars.[c] * mat.[c].[row]) cols
            let linearExp = List.reduce (+) constraintOperands

            // create the constraint
            let rangeConstraint = RangeConstraint(linearExp, lb.[row], ub.[row])
            solver.Add(rangeConstraint) |> ignore
    | _ -> ()


    // Equality Constraints
    match (solverOptions.ConstraintMatrixEq, solverOptions.ConstraintVectorEq) with
    | (None, Some b) ->
        failwithf "Matrix for Equality Constraints undefined."
    | (Some mat, Some vec) ->
        for row = 0 to (vec.Length-1) do
            // generate constraint operands based on indices
            let constraintOperands = List.map (fun c ->  vars.[c] * mat.[c].[row]) cols
            let linearExp = List.reduce (+) constraintOperands

            // create the constraint
            let equalityConstraint = RangeConstraint(linearExp, vec.[row], vec.[row])
            solver.Add(equalityConstraint) |> ignore
    | _ -> ()

    // Objective
    let objectiveOperands = List.map (fun c ->  solverOptions.ObjectiveFunction.[c] * vars.[c]) cols
    let objectiveExp = List.reduce (+) objectiveOperands

    match solverOptions.SolverGoal with
    | Minimize ->
        solver.Minimize(objectiveExp)
    | Maximize ->
        solver.Maximize(objectiveExp)

    solver

  /// Solves the optimization problem and prints the results to console
  let SolverSummary (solver:Solver) =
    let resultStatus = solver.Solve();

    match resultStatus with
    | status when status <> Solver.OPTIMAL ->
        printfn "The problem does not have an optimal solution!"
        exit 0
    | _ ->
        printfn "\nProblem solved in %d milliseconds" (solver.WallTime())
        printfn "Iterations: %i\n" (solver.Iterations())

    printfn "Objective: %f" (solver.Objective().Value())
    for i in 0 .. (solver.NumVariables()-1) do
      printfn "%-10s: %f " (sprintf "var[%i]" i) ((solver.LookupVariableOrNull(sprintf "var[%i]" i)).SolutionValue())

    solver