[Coin-standards] Technical Item! CLP

[Irv Lustig]

Leo:

At 06:41 PM 4/23/02 -0500, Leonardo B. Lopes wrote:
>Wow! This sounds like real progress. Cols and Rows are objects in SNOML,
>and something like your example could be done today using it. It might
>actually be quite simple to merge the CLP expressions and the current
>version of SNOML, assuming the CPML (?) has a closing tag for expressions,
>like <math> in MathML. It also shouldn't be hard to keep the
>column-as-object requirement as we move forward in other directions.

I think you may want to change your naming scheme.  Rather than calling the 
objects "cols" and "rows", they should be "variables" and 
"constraints".  That is a language common to math programming and 
constraint programming.

Also, you want to have the ability to treat a group of constraints (and 
associated variables) as a single "global constraint".  In other words, a 
group of linear constraints becomes a single "constraint" that is at the 
same level in the object hierarchy as something like "alldifferent".

I haven't looked at your representation, but from the discussion, I infer 
that it is too matrix-oriented (i.e., rows, columns), rather than 
"problem"-oriented (i.e., constraints, variables).  CP needs the latter 
representation.

         -Irv