Organization of the OCaml interface

Not all functions of type C library are interfaced; for instance, you can't manipulate saturation matrices in OCAML (buf of course, if someone asks me this feature, I can add it). The three datatypes that are interfaced are vectors, matrices, and polyhedra, and there is a corresponding module for each of these datatypes. All objects are properly finalized, thanks to the use of "custom" blocks.

The interface was developped with the CAMLIDL tool. Each `.idl' file produces a `_caml.c', a `.mli' and a `.ml' file. Each file is represents a module.

The interface is decomposed as follows:

• Files `polka_lexer.mli', `polka_lexer.mll' and `polka_parser.mly' implements pretty-input of vectors and matrices;
• Files `polka_caml.h', `polka_caml.c', `polka.mli' and `polka.ml' take care of global stuff and defines the OCAML module Polka.
• `vector.idl' defines the module Vector.
• `matrix.idl' defines the module Matrix.
• `poly.idl' defines the module Poly.

When invalid arguments are given to some function (incompatible dimensions, out-of-bound access, ...), an Invalid_argument exception is usually thrown.

We provide a unified type for coefficients, which is the OCAML native integers. Converting to OCAML native integers coefficients as used in the C library, which can be long int, long long int or mpz_t, may produce overflow. In this case, an exception Polka.Overflow(str) is raised, carrying the string representation of the coefficient in base 10. This concerns only the functions Vector.get and Matrix.get.