README

This document is no longer updated. Pkease consult the Normaliz manual for a full description of libnormaliz.



*******************************************************************************
**********             The easiest way to use libnormaliz            **********
*******************************************************************************

This version of libnormaliz is considered to be stable. We would be happy to
hear from you if you are using libnormaliz (mailto: [email protected]).

More information on normaliz:
http://www.mathematik.uni-osnabrueck.de/normaliz/

Preperation:

Check the compile instructions in ../INSTALL

Quick workflow overview:

1) include libnormaliz/cone.h in your source code
2) create a libnormaliz::Cone object
3) call one of the Cone.compute() methods
4) check if the desired data is computed: isComputed( X )
5) get the data: getX()
6) repeat 4+5 or 3+4+5

Then link against libnormaliz. It might be necessary to activate compilation with OpenMP (in g++ use the -fopenmp flag).


In version 3.0 the default template parameter should be mpz_class.
Internally libnormaliz will still use a smaller integer type in the
computations if possible and switch to mpz_class where necessary.
If you use another type, e.g. long, libnormaliz will use this type for
computations and will throw an exception if it fails.

Example:

typedef mpz_class Integer;

// get some input data
std::vector< std::vector <Integer> > Data = ...
libnormaliz::Type::InputType type = integral_closure; // for all input types see below
libnormaliz::Cone<Integer> MyCone = libnormaliz::Cone<Integer>(type, Data);

libnormaliz::ConeProperties Wanted;
Wanted.set(libnormaliz::ConeProperty::Triangulation, libnormaliz::ConeProperty::HilbertBasis);
MyCone.compute(Wanted);

// or just
MyCone.compute(libnormaliz::ConeProperty::Triangulation, libnormaliz::ConeProperty::HilbertBasis);

if (MyCone.isComputed(libnormaliz::ConeProperty::HilbertBasis)) {
	std::vector< std::vector< Integer> > HB& = MyCone.getHilbertBasis();
	//use it
}
if (MyCone.isComputed(libnormaliz::ConeProperty::Multiplicity)) {
   cout << "MyCone has multiplicity " << MyCone.getMultiplicity();
}



Possible InputTypes (from libnormaliz.h):
    integral_closure,
    polyhedron,
    normalization,
    polytope,
    rees_algebra,
    inequalities,
    strict_inequalities,
    signs,
    strict_signs,
    equations,
    congruences,
    inhom_inequalities,
    dehomogenization,
    inhom_equations,
    inhom_congruences,
    lattice_ideal,
    grading,
    excluded_faces,
    lattice,
    saturation,
    cone,
    offset,
    vertices,
    support_hyperplanes,
    cone_and_lattice

You can give up to 3 input type/matrix combinations directly in the constructor.
To combine more input types, build a
std::map <libnormaliz::InputType, vector< vector<Integer> > >
and construct the libnormaliz::Cone from it.

See cone_property.h for a complete list of possible properties.

Please see the Normaliz Documentation for more information on the input types and computation modes
http://www.mathematik.uni-osnabrueck.de/normaliz/



Alternative Configuration

The integer type in the libnormaliz library is templated. So in theory it is possible to use other integer types. Then you have to include libnormaliz-all.cpp. In this case you do not need to link libnormaliz.a. We suggest (and only tested) 'long long int' and the gmp type 'mpz_class'.


Other integer types

If you want to use other types, you probably have to implement some conversion functions which you can find in integer.h. Namely

bool libnormaliz::try_convert(TypeA, TypeB); // converts TypeB to TypeA, returns false if not possible

where one type is your type and the other is long or long long (and maybe mpz_class?).
Additionally if your type uses arbritary precision (i.e. it is some wrapper for mpz) also implement

template<> inline bool libnormaliz::using_GMP<YourType>() { return true; }