[Coin-discuss] osicplex strange behavior, writeMps bugs.

[John J Forrest]

Matt,

That sounds like a good idea.  In these matters we have in the past looked
at the MPSX 370 definitions to see what a "standard" is.  As far as I can
see from the manual BCDOUT just tries to be a mirror image of CONVERT and
so I think the default should be that writeMps produces a mirror image of
readMps.

I have modified all the interfaces.  It should compile with all solvers but
I have not tested them all - it will be trivial to correct any typos.

To avoid ambiguity when the native Coin writeMps is called by non virtual
interface I have renamed that writeMpsNative.  The only code, other than
Clp, that uses it directly is Dylp.  Lou - I added parameter but did not
change code.  If you want I can do so.

At present Clp is only one that takes advantage of objSense.  OSL does not.

John Forrest


                                                                                                                          
                      Matthew Saltzman                                                                                    
                      <mjs at ces.clemson.edu>                        To:       coin-discuss at www-124.southbury.usf.ibm.com   
                      Sent by:                                     cc:                                                    
                      coin-discuss-admin at www-124.southbury.        Subject:  Re: [Coin-discuss] osicplex strange          
                      usf.ibm.com                                   behavior, writeMps bugs.                              
                                                                                                                          
                                                                                                                          
                      01/23/2003 11:48 AM                                                                                 
                      Please respond to coin-discuss                                                                      
                                                                                                                          




On Thu, 23 Jan 2003, John J Forrest wrote:

> It would be trivial to add another parameter to writeMps i.e. bool
> forceMinimization=false).
>
> Clp uses the OsiSolverInterface version anyway.

How about an additional parameter of objSense with value 1 to write the
minimization objective, -1 to write the maximization objective (like
getObjSense in OSI), or 0 (default) to write coefficients unmodified?

Adding the parameter at the end of the list with default 0 shouldn't break
any existing code.

--
                         Matthew Saltzman

Clemson University Math Sciences
mjs at clemson.edu
http://www.math.clemson.edu/~mjs
_______________________________________________
Coin-discuss mailing list
Coin-discuss at www-124.ibm.com
http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-discuss