[Coin-standards] INFORMS / directions

[Robert Fourer]

In reply to the message from >>> Laurent Perron <lperron at ilog.fr>
received 10/17/02 11:15 AM:

> Let me present myself, I am Laurent Perron and I am project leader of
> ILOG Solver and ILOG Concert. ILOG Concert is a common modeling API
> shared by ILOG Cplex and ILOG Solver (our CP engine).
> 
> The main idea behind an XML format suitable for CLP or CP is about
> expressiveness. For instance, we have some "logical" constraint that we
> use to model our problem. Such logical constraints are "all different",
> "sequence". They may be very difficult to linearize, or impossible to
> linearize. They may also have different linearizations that adapted to
> different situations.
> 
> Therefore, the implication is that for each of these logical
> constraints, there will be a special xml tag for them.
> 
> Please note that our CP engine deals with other kind of basic objects
> (integer set variable, integer bag (or multi-set) variables). There are
> also extensions to deal with scheduling, dispatching and routing,
> configuration... Each of these extensions introduce new basic objects
> and thus new XML tags.
> 
> We have also developed a internal XML layer to save out models. This
> layer is mapped quite directly from our Concert layer. I think it would
> be a good idea to show some samples of these generated XML to compare
> with this standardization effort (that we want to support in addition to
> this simple XML format).

In reply to the message from >>> Richard Damian
<richard_damian at peoplesoft.com> received 10/17/02 1:07 PM:

> I got a copy of your note from a colleague.  We work in an Optimization
> group that is facing the problem of storing mathematical expressions in a
> database.  On the front end, I have a parser that converts our
> representation to XML and then to MathML for display in a browser.  But on
> the back end, we need to be able to store and retrieve math expressions
> from a database to run a linear model.
> 
> I have been thinking for some time about using XML capabilities of most
> modern databases to store our expressions as XML.  In that way we would not
> need to worry so much about a potentially complex database schema needed
> for storage of the tree.  We could depend on the QPath capabilities of SQL
> to traverse our data and regenerate our expression.
> 
> Do you have an experience in this area?

Could you supply an example of an XML representation of the kind you are
generating?  An example could help others to appreciate what you're trying
to represent, in particular whether you're working with models or instances
(or something in-between).  Even a very small example can be helpful, but an
example of at least a few hundred variables and constraints can be valuable
in showing how big a file is generated and what takes up most of the space.

-- Bob Fourer
   4er at iems.nwu.edu