[Osi] milp support in OSI

Matthew Saltzman mjs at clemson.edu
Thu Apr 23 12:02:07 EDT 2009


On Thu, 2009-04-23 at 08:40 -0700, Lou Hafer wrote:
> Folks,
> 
> 	I'm with Laci here.  We should have removed Osi::branchAndBound years
> ago.  Osi is has become far too bloated.  There's just no excuse for a class
> with 200 public methods.  Yet more functions will not make OsiSolverInterface a
> better, more useable design.  What's needed is some thought on how to properly
> divide it into a coherent base class and coherent subclasses.

Yes, absolutely.

No matter how you slice it up into classes, though, the API is still
going to have a boatload of public methods.  They'll just be better
organized (one hopes).

> 
> 	A good first step would be that we all stop adding new interface
> routines on the spur of the moment, and start working on the basis of consensus
> addition / modification / deletion.  

Hear, hear.  Willy-nilly addition of public routines that are supported
by only some OsiXxxSolverInterface classes is largely how we got to the
current (rather messy) state of affairs.

> The debate that comes out of that process
> would likely get us thinking on how to properly organise the existing OSI
> methods.

Do you envision a structured debate or do you want to handle things case
by case for now?

> 
> 							Lou

-- 
                Matthew Saltzman

Clemson University Math Sciences
mjs AT clemson DOT edu
http://www.math.clemson.edu/~mjs




More information about the Osi mailing list