[Coin-standards] Technical Item! CLP

[Irv Lustig]

At 12:17 PM 4/12/02 -0500, Leonardo B. Lopes wrote:
>Could you invite some ILOG CP person into the list?

I spoke with Laurent Perron of the ILOG R&D person.  Laurent is a
key contributor to ILOG Concert Technology, which is our effort to
come up with representations (in C++ and Java) for CP and MP.
Laurent is busy on another project right now, but intends to
contribute an XML proposal for CP to the list in the near future.

One of the points that Laurent made is that an LP matrix
of constraints needs to be considered a potential "constraint"
in the CP representation.  This means that one needs to treat
variables as objects, so that when a set of linear constraints
are given, you know which variables are involved.

For example, one could imagine a problem of the following form
(using a bastardized OPL):

var int x[1..10] in 0..100;
var int y[1..10] in 0..100;

/* Constraints */

forall (i in 1..5) {
    sum (j in 1..10) ax[i,j]*x[j] +
    sum (j in 1..10) ay[i,j]*y[j] = b[i];
};
alldifferent(x);

The first set of constraints are linear constraints and correspond
to a matrix.  The next constraint is alldifferent(), but only on
a subset of the variables.

The point here is that a set of linear constraints is at the
same "object level" as any of the CP constraints.

         -Irv