Originally posted by: JPatrick

I want to create a decision variable that is the upper half of a nxn matrix.  I know I can define dvar x[Size][Size] and then simply force the lower half of the matrix to be zero but that seems wasteful.  Is there no way to conditionally define the decision variable?  I tried the following:

 dvar int x[i in Size][j in Size: i < j] in 0..1;
but OPL did not like it.

Jonathan

Hi,

you could use a tuple set in order to deal with half the matrix.

And then if you need an index with 2 dimensions, you may use a dexpr. (y in my tiny example)

int Size=4;

tuple t
{
int i;
int j;
}

{t} index={<i,j> | ordered i,j in 1..Size};

dvar int x[index] in 0..1;
dexpr int y[i in 1..Size,j in 1..Size]=(i==j)?0:((i<j)?x[<i,j>]:x[<j,i>]);

subject to
{
x[<1,2>]==1;
y[3][2]==1;

}

regards

PS: Many how to with OPL at https://www.linkedin.com/pulse/how-opl-alex-fleischer/