[Osi] milp support in OSI

Matthew Saltzman mjs at clemson.edu
Thu Apr 23 11:06:27 EDT 2009


On Thu, 2009-04-23 at 10:55 -0400, Laszlo Ladanyi wrote:
> I think Osi should not even support solving MILPs... Even that 
> branchAndBound() call should be taken out... It's an entirely 
> different story that there could be another class that supports MILPs 
> and is built on top of Osi. There is a plan (for x years now, x>3) to 
> rewrite Osi with allowance for simplex based methods and MILP related 
> methods all nicely separated into distinct classes, data handling and 
> algorithms are separated, too, etc... However, I have no idea when 
> anyone will have the time to actually implement that :-(

OK In a broad sense, though, the idea of having a generic API for MILPs
is a worthy one.  Conceptually, the C++ class we currently call
OsiSoverInterface and its descendents are distinct from the idea of a
generic solver API, which may consist of different classes realizing
different pieces of it.

> 
> --Laci
> 
> On Thu, 23 Apr 2009, Matthew Galati wrote:
> 
> > Currently OSI does support the branchAndBound function to solve a MILP.
> > However, I don't see any access methods for MILP statistics afterwards (or
> > any other MILP-type manipulation methods). Is there any intention to add
> > this? For example, how do I get the number of nodes used, the gap, how do I
> > set options specific to MILP for say, Cplex or Xpress calls?
> >
> > I can always cast an Osi to an OsiCpx and then get CpxLp and CpxEnv ptrs and
> > then use Cpx directly -- but obviously all of this defeats the purpose of
> > having OSI.
> >
> > Thanks,
> > Matt
> >
> 
> _______________________________________________
> Osi mailing list
> Osi at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/osi
-- 
                Matthew Saltzman

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




More information about the Osi mailing list