README

PHP wrapper around LPSolve 5.5, providing a safe and convenient interface for building and solving linear optimization problems.

Installation

Install via Composer

composer require kerigard/lpsolve

Load the Composer autoloader

require 'vendor/autoload.php'

Important

This library requires the lp_solve PHP extension.

Note

The official lp_solve extension only supports PHP 5, so it needs to be compiled manually for modern PHP versions.

Building the lp_solve extension

(for PHP 5.6 - 8.x)

The repository Kerigard/lp-solve-php-docker contains maintained source directories for building the lp_solve extension for different PHP versions:

1. Install lp-solve system package

apt update && apt install -y lp-solve

2. Download source code

curl -fsSL https://github.com/Kerigard/lp-solve-php-docker/archive/8.x.tar.gz \ | tar -xz -C /tmp --strip-components=1

Replace 8.x with the desired branch.

3. Build and install

(cd /tmp/lp-solve/extra/PHP && phpize && ./configure && make && make install) rm -r /tmp/lp-solve

4. Enable the extension

Add to your php.ini:

extension=phplpsolve55.so 

Usage

Maximization with constraints

use Kerigard\LPSolve\Constraint; use Kerigard\LPSolve\Problem; use Kerigard\LPSolve\Solver; $problem = new Problem( objective: [143, 60, 195], constraints: [ new Constraint([120, 210, 150.75], LE, 15000), new Constraint([110, 30, 125], LE, 4000), new Constraint([1, 1, 1], LE, 75), ] ); $solver = new Solver(Solver::MAX); $solution = $solver->solve($problem); var_dump($solution->getStatus()); // OPTIMAL solution var_dump($solution->getCount()); // 1 var_dump($solution->getIterations()); // 2 if ($solution->getCode() === OPTIMAL) { var_dump($solution->getObjective()); // 6986.842105... var_dump($solution->getVariables()); // [0, 56.578947..., 18.421052...] }

Minimization with bounds

$problem = new Problem( objective: [-1, -2, 0.1, 3], constraints: [ new Constraint([1, 1, 0, 0], LE, 5), new Constraint([2, -1, 0, 0], GE, 0), new Constraint([-1, 3, 0, 0], GE, 0), new Constraint([0, 0, 1, 1], GE, 0.5), ], lowerBounds: [0, 0, 1.1, 0], upperBounds: [] ); $solver = new Solver(Solver::MIN); $solution = $solver->setScaling(SCALE_MEAN | SCALE_INTEGERS)->solve($problem); var_dump($solution->getObjective()); // -8.223333...

Integer and Binary Variables

$problem = new Problem( objective: [-1, -2, 0.1, 3], constraints: [ new Constraint([1, 1, 0, 0], LE, 5), new Constraint([2, -1, 0, 0], GE, 0), new Constraint([-1, 3, 0, 0], GE, 0), new Constraint([0, 0, 1, 1], GE, 0.5), ], integerVariables: [0, 0, 1, 0], // integer variables (only variable #3 is integer) binaryVariables: [1, 0, 0, 1] // binary variables (variables #1 and #4 are binary) ); $solver = new Solver(Solver::MIN); $solution = $solver->setVerbose(DETAILED)->solve($problem); var_dump($solution->getVariables()); // [1, 2, 1, 0]

If you pass true instead of an array, all variables will become integer or binary.

Error handling

Invalid problem definition

$problem = new Problem( [1], [new Constraint([0, 78.26, 0, 2.9], GE, 92.3)] ); try { $solver = new Solver(); $solver->setTimeout(0)->solve($problem); } catch (\LPSolveException $e) { var_dump($e->getMessage()); // Invalid vector }

Throwing exceptions for non-optimal solutions

$problem = new Problem( [10, 10], [ Constraint::fromString('1x + 1y = 20'), Constraint::fromString('0x + 1y <= 5'), Constraint::fromString('1x + 0y <= 5'), ] ); try { $solver = new Solver(Solver::MIN); $solver->throw()->solve($problem); } catch (\LPSolveException $e) { var_dump($e->getMessage()); // Model is primal INFEASIBLE var_dump($e->getCode() === INFEASIBLE); // true }

Additional behavioral notes:

• The original lp_solve extension may produce fatal errors for malformed equations.
• When built via Kerigard/lp-solve-php-docker, internal errors become LPSolveException.
• Calling ->throw() forces an exception for all non-optimal results.

Callbacks

$problem = new Problem( [1, 3, 6.24, 0.1], [ new Constraint([0, 78.26, 0, 2.9], GE, 92.3), new Constraint([0.24, 0, 11.31, 0], LE, 14.8), new Constraint([12.68, 0, 0.08, 0.9], GE, 4), ] ); $solver = (new Solver()) ->beforeSolve(function ($lpsolve, $problem) { lpsolve('set_improve', $lpsolve, IMPROVE_SOLUTION); }) ->afterSolve(function ($lpsolve, $problem, $solution) { lpsolve('write_lp', $lpsolve, 'model.lp'); }); $solution = $solver->solve($problem);

License

MIT. Please see the LICENSE FILE for more information.