- SEE ALSO
Math::LP::Solve - perl wrapper for the lp_solve linear program solver
use Math::LP::Solve qw(:ALL); # imports all functions and variables # construct an LP with 0 initial constraints and 2 variables $lp = make_lp(0,2); # add the constraint x1 + 2 x2 <= 3 $coeffs = ptrcreate('double',0.0,2); # mallocs a C array ptrset($coeffs,1.0,0); ptrset($coeffs,2.0,1); add_constraint($lp,$coeffs,$LE,3); ptrfree($coeffs); # frees the C array # set the objective function to x1+x2 and solve for a maximum $obj = ptrcreate('double',1.0,2); set_obj_fn($lp,$obj); ptrfree($obj); set_maxim($lp); # solve the LP solve($lp) == $OPTIMAL or die "No solution found"; $solution = lprec_best_solution_get($lp); # extract the results from the solution array $obj_fn_val = ptrvalue($solution,0); $constr_val = ptrvalue($solution,1); $x1 = ptrvalue($solution,2); $x2 = ptrvalue($solution,3);
Math::LP::Solve is a wrapper around the freeware lp_solve library, which solves linear and mixed linear/integer programs. Most functions and data structures in the file lpkit.h of the lp_solve distribution are made available in the Math::LP::Solve namespace.
This document does not go into the details of how to setup and solve a linear program using the lp_solve library. For details on this you are referred to the documentation included in the source code for lp_solve.
That being said, a few details of the Perl wrappers around the underlying lp_solve library need explaining in order to be able to use them. (For those interested, the wrapping was done using SWIG, more info at http://www.swig.org/) All symbols (functions and variables) are divided into 4 categories. All these symbols are in the
Math::LP::Solve namespace and are not exported by default. They are however tagged so that you can easily import them into your own code. The following
%EXPORT_TAGS are available:
pointer library functions, needed to handle C-style arrays;
pairs of get/set functions to access data fields of structs;
wrappers for lp_solve library functions;
perl scalar variables mapping
#define'd constants in lpkit.h.
A 5th category named ALL is available in
%EXPORT_TAGS, which includes all symbols of the 4 mentioned categories.
The pointer library functions are needed to pass arrays of coefficients etc. to and from the lp_solve functions and data structures. In the underlying C library, this is done using
double* pointers, which are not available in Perl. The pointer library functions provide a Perl interface to get around this problem.
There are several pointer library functions, and they are fully explained in the SWIG documentation. However, the following is all you need to know to use them with lp_solve:
Creates and returns a pointer to type
$type, which is an array with
$sizefields initialized to
$initval. E.g. an array of 2 doubles initialized to zero is created with the command
$arr_double = Math::LP::Solve::ptrcreate('double',0.0,2);
sets the value of the
$index'th field of the array pointed to in
$ptrto the value
$val. E.g. the 2nd entry of an array of doubles is set to 3.14 using
Note that the 1st entry is denoted by index 0, as in C.
$index'th entry of the array pointed to in
$ptr. E.g. the 1st value of an array of doubles is requested using
$d0 = Math::LP::Solve::ptrvalue($arr_double,0);
frees the memory allocated for
$ptr. Always do this when you are finished with an array you allocated yourself using ptrcreate(), or you will end up with memory leaks. Also, take care not to invoke ptrfree() twice on the same pointer if it is not re-created.
The functions have the same name as in
lpkit.h. Note however that
double* parameters need to be handled with the aforementioned pointer library functions. The pointer library functions are not needed for the
lprec* parameters, as their creation, manipulation and freeing is completely covered by the
lpkit.h functions. E.g. an LP is created with
$lp = Math::LP::Solve::make_lp(0,0);
subsequently manipulated with
and finally freed using
Some functions have been added to the ones available in
lpkit.h to ease file manipulation and handling names of rows and columns:
returns the name of the LP;
sets the name of the LP to
- lprec_row_name_get($lp,$i) and lprec_col_name_get($lp,$i)
returns the name of the row resp. column with index
- lprec_row_name_set($lp,$i,$name) and lprec_col_name_set($lp,$i,$name)
sets the name of the row resp. column with index
opens the file
$mode, which is specified as a string. Calls the C function fopen() internally;
closes a filehandle obtained with open_file().
Following constants are available in the Math::LP::Solve namespace:
- General constants
- Constraint types
- Boolean values
- Status values obtained from solve()
- Extra status values obtained from lag_solve()
Each data field in
struct lprec can be queried from a Perl variable holding an LP using
Math::LP::Solve::lprec_FIELD_get($lp) and set using
Note that the row and column names are accessed using the functions lprec_row_name_get(), lprec_col_name_get(), lprec_row_name_set() and lprec_col_name_set() described above.
The underlying lp_solve library has been written by Michel Berkelaar and adapted by Jeroen Dirks. Its source code is available at ftp://ftp.ics.ele.tue.nl/pub/lp_solve/
More information on the exporting of symbols is found in Exporter.
The wrapping of the C library was done with the aid of SWIG. The SWIG homepage is located at http://www.swig.org/
An object oriented interface to the lp_solve library has been written on top of Math::LP::Solve. For more info look at Math::LP.
Wim Verhaegen <firstname.lastname@example.org>
Copyright (c) 2000-2001 Wim Verhaegen. All rights reserved. This program is free software; you can redistribute and/or modify it under the same terms as Perl itself.
Consult the lp_solve documentation for copyright information on the lp_solve library.