Thursday, February 12, 2009


I doubt that the message from GAMS/COINCBC:

Problem statistics: 352037 columns and 122283 rows.
                    231136 variables have integrality restrictions.

actually is written by code from John Forrest. I guess he would write rows followed by columns (that is how I would have presented this message). Indeed further down we see messages like:

processed model has 50182 rows, 152495 columns (152495 integer) and 429919 elements

It is interesting to note how large a role such conventions play. No one would write a[j,i] unless for a specific reason. When I was teaching numerical programming, I always asked students to redo their code if it contained:

var x : integer;

The strict use of such conventions make things more predictable and thus easier to read quickly.