NLPAPI is a subroutine library with routines for building nonlinear programming problems. The general form is an objective with a set of simple bounds, equality and inequality constraints. It is built around the "Group Partially Separable" structure that LANCELOT defines, but constraints and objective may also be defined as functions of the problem variables.
The user creates a "NLProblem", which is a pointer to a data structure. Then the objective and constraints are added with various calls. Once the problem is defined our interface to LANCELOT may be used to create a NLLancelot data structure, and pass it and the problem to a routine which either minimizes or maximizes the objective subject to the constraints.
The source code for NLPAPI is stand-alone. If the user wishes to use our LANCELOT interface he must obtain his own copy of LANCELOT.
The very short version of the download and installation instructions for UNIX-like system (including Linux and Cygwin) is this:
-
Get the code via subversion using the following command
svn co https://projects.coin-or.org/svn/NLPAPI/trunk NLPAPI
-
Go into the downloaded directory and edit the file share/config.site to give local lib and include dirs for LANCELOT. (Last two lines in the file).
-
Run the configuration script
./configure
Make sure the last line of output says that the configuration was successful. The
configure
script has many customization features. To see them, type./configure --help
. In particular, to install a COIN-OR package at a location accessible to all users on your machine, you can run./configure --prefix=/usr/local
. -
Compile the code:
make
-
Install the generated libraries, executables, and header files:
make install
This will create subdirectories
bin
,lib
, andinclude
in the download directory with the product of the compilation.
NLPAPI uses autoconf. It has been tested on AIX 4.3, AIX5.1, Cygwin, and !RedHat 8.0, but should install cleanly on any UNIX machine.
Source for several examples is included, as well as a SIF decoder which will translate many SIF files (no guarantees!) into C and Fortran source for defining the NLPAPI problem.
Documentation is included in a tex file, or the user may view a pdf version of the User's Guide, or the Programming Reference.