This file describes how to install the constraint-solver library and
run the c++ and python examples. This has been tested under ubuntu
10.10, Mac OS X Snow Leopard with xcode 3.2.3, and Microsoft Windows
with Visual Studio 2010 express C++.

Upon untarring the given operation_research tar file, you will get the
following structure:

operations_research-1.0/
  LICENSE-2.0.txt    <- Apache License
  Makefile           <- Makefile for unix platform (ubuntu linux, mac os X)
  Makefile.msv       <- Makefile for Microsoft Visual Studio
  README             <- This file
  base/              <- Directory containing basic utilities.
  constraint_solver/ <- The main directory for the constraint solver library
  examples/          <- C++ examples
  graph/             <- Graph algorithms used in the constraint solver
  objs/              <- Where C++ objs files will be stored
  python/            <- Python examples
  util/              <- More utilities needed by the constraint solver

First, you will need to install the google-gflags library, compile it
and install it.  It can be obtained at the following address:

  http://code.google.com/p/google-gflags/

By default, the makefile assumes the installation path is
../gflags-1.3 from the Makefile, thus at the same level as the
operations_research directory.

Then you need to install zlib. The makefile default path is
../zlib-1.2.5 from the location of the makefile.

Please edit the top of the makefile to accomodate your installation.

On Microsoft Windows, please do the following:
  - decompress the gflags directory, do not compile it
  - modify the Makefile.msv to point to the source of the gflags source
    directory.
  - install the zlib source archive, step into it
    and run "nmake -f win32/Makefile.msc" directly at this point.
    (please note that there is a bug in zlib-1.2.5 makefile as the OBJS
     list of files misses the inffast.obj entry).
  - modify the Makefile.msv to point to the root of the zlib source tree.

You can then compile the library, examples (linux, mac, and windows)
and python wrapping (linux, mac):

  make (linux, mac)
  nmake -f Makefile.msv (windows)

To compile in debug mode, please use

  make DEBUG=-g (linux, mac)
  nmake -f Makefile.msv DEBUG="/Od /Zi"

You can clean everything using

  make clean (linux, mac)
  nmake -f Makefile.msv clean (windows)

When everything is compiled, you will find under operations_research:
  3 libraries (libcp.a, libutil.a and libbase.a) for unix, or
  4 libraries (cp.lib, util.lib, base.lib and gflags.dll) for Windows
  One binary per C++ example (e.g. nqueens or nqueens.exe on Windows)
  A C++ wrapping library (_pywrapcp.so)

You can execute C++ examples just by running then:

  ./magic_square (linux, mac)
  magic_square.exe (windows)

For the python examples, as we have not installed the
constraint_solver module, we need to use the following command:

PYTHONPATH= <python_binary> python/<sample.py>

As in

  PYTHONPATH= python2.6 python/golomb8.py

This has been tested with python 2.5 and 2.6.
The wrapping using SWIG has been tested with SWIG 1.3.40 and 1.3.31.