SolverAPI.jl

This Julia package takes serialized Rel SolverAPI library models, translates these to enable access to external solvers, and returns serialized solution information.

Currently, JSON is used for serialization, and only optimization solvers and constraint programming solvers are supported.

API

The main API is solve which takes a JSON request or Julia dictionary and a solver and returns a response.

using SolverAPI

import HiGHS

tiny_min = Dict(
    "version" => "0.1",
    "sense" => "min",
    "variables" => ["x"],
    "constraints" => [["==", "x", 1], ["Int", "x"]],
    "objectives" => ["x"],
)

solve(tiny_min, HiGHS.Optimizer())

Format

Request

Request format example:

{
  "version": "0.1",
  "sense": "min",
  "variables": [
    "x"
  ],
  "constraints": [
    [
      "==",
      "x",
      1
    ],
    [
      "Int",
      "x"
    ]
  ],
  "objectives": [
    "x"
  ],
  "options": {
    "silent": 0,
    "print_format": "LP"
  }
}

Required fields:

• version: [String] The version of the API that is used.
• options: [Array] Options, such as the time limit, if the model should be printed, or general solver attributes. For a complete list, please refer to the documentation of the solver.
• sense: [String] One of feas, min, or max
• variables: [Array] A list of variables that are used in the model.
• constraints: [Array] A list of constraints. Each constraint contains an operation and a set of arguments, such as ["==", "x", 1].
• objectives: [Array] The objective.

Options

We always support the following options:

• silent: [Bool] Controls if the solver prints any logs.
• time_limit_sec: [Float64] Limits the total time expended. The optimization returns a TIME_LIMIT status. If not provided, a default of 300s is used.
• print_only: [Bool] If set to true the model will only be printed and not solved.
• print_format: [String] If and how the model should be printed. Currently supported formats: MOI, LaTeX, MOF, LP, MPS, NL.
• relative_gap_tolerance: [Float64] The relative gap tolerance. Must be within [0,1].
• absolute_gap_tolerance: [Float64] Absolute gap tolerance. Must be non-negative.

Response

Response format examples:

{
  "version": "0.1",
  "results": [
    {
      "objective_value": 0.0,
      "primal_status": "FEASIBLE_POINT",
      "names": [
        "\"x\""
      ],
      "values": [
        1
      ]
    }
  ],
  "termination_status": "OPTIMAL",
  "solve_time_sec": 0.000112959,
  "solver_version": "HiGHS_1.5.3",
  "relative_gap": 0.0
}

Example with a model error:

{
  "version": "0.1",
  "errors": [
    {
      "type": "InvalidModel",
      "message": "Objectives must be empty when `sense` is `feas`."
    }
  ],
  "termination_status": "OTHER_ERROR"
}

Example with printing and no solving:

{
  "model_string": "{\"name\":\"MathOptFormat Model\",\"version\":{\"major\":1,\"minor\":4},\"variables\":[{\"name\":\"x\"}],\"objective\":{\"sense\":\"min\",\"function\":{\"type\":\"Variable\",\"name\":\"x\"}},\"constraints\":[{\"name\":\"c1\",\"function\":{\"type\":\"ScalarAffineFunction\",\"terms\":[{\"coefficient\":1.0,\"variable\":\"x\"}],\"constant\":0.0},\"set\":{\"type\":\"EqualTo\",\"value\":1.0}},{\"function\":{\"type\":\"Variable\",\"name\":\"x\"},\"set\":{\"type\":\"Integer\"}}]}",
  "termination_status": "OPTIMIZE_NOT_CALLED",
  "version": "0.1"
}

Required fields:

• version: [String] The version of the API that is used.
• termination_status: [String] The MOI termination status.

Optional fields:

• results: [Array] The results array. Zero, one, or multiple results will be present. Each result will contain multiple fields describing the solution, such as objective_value, primal_status, names, and values.
• errors: [Array] None, one, or multiple errors that were present.
• model_string: [String] If requested via print_format the model as a string.
• solver_version: [String] The version of the solver used.
• solve_time_sec: [Float64] Duration of solving.
• relative_gap: [Float64] The relative gap.