[Cgl] cgl, cpx, mir, error

[Matthew Galati]

I take that back - getColSolution is ok, if I calculate rowAct myself in CglMir.

And, I am not sure how to tell that I have OsiCpx when I am inside Cgl. So, I just calculate rowAct in all cases. It's ugly, and unnecessarily slow, but I think it works. I still think the real fix is to re-design Cgl to not need Osi.

Here is a simple patch if you are interested:
  https://coral.ie.lehigh.edu/projects/DECOMP/browser/trunk/TempFix/CglMir 
  https://coral.ie.lehigh.edu/projects/DECOMP/changeset/237



-----Original Message-----
From: Matthew Galati 
Sent: Wednesday, April 11, 2007 6:49 PM
To: Matthew Galati; Lou Hafer; cgl at list.coin-or.org
Subject: RE: [Cgl] cgl, cpx, mir, error

Hmmm... rowAct is not the only problem. The error also occurs because of getColSolution. Yet, I just set it with setColSolution. 



-----Original Message-----
From: cgl-bounces at list.coin-or.org [mailto:cgl-bounces at list.coin-or.org] On Behalf Of Matthew Galati
Sent: Wednesday, April 11, 2007 6:29 PM
To: Lou Hafer; cgl at list.coin-or.org
Subject: RE: [Cgl] cgl, cpx, mir, error

Probably the easiest workaround is to fix CglMir to calculate Ax* rather than query it when using OsiCpx. 


-----Original Message-----
From: cgl-bounces at list.coin-or.org [mailto:cgl-bounces at list.coin-or.org] On Behalf Of Lou Hafer
Sent: Wednesday, April 11, 2007 6:15 PM
To: cgl at list.coin-or.org
Subject: Re: [Cgl] cgl, cpx, mir, error

Ted, Matt,

	I know I must sound like a perpetual fussbudget, but ...  it's not that simple.  For one thing, how will you know there's an error?  Look for a bogus return value?  What's bogus?  Look for a throw?  Hah!  That's more inconsistent than most across implementations.  Who's going to check?  Only the client can really say whether they're looking at a bogus return value --- that'll depend on the state of the OSI (unloaded, imposed solution, solution from solver).

	It's a reasonable idea to provide a default (if inefficient) implementation in the base class.  But arguably better to fix OsiCpx so that it knows when to use a solution supplied by the user and when to use a solution from the solver, and when it has no solution available.

							Lou

_______________________________________________
Cgl mailing list
Cgl at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/cgl

_______________________________________________
Cgl mailing list
Cgl at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/cgl