[Coin-lpsolver] OsiSimplexInterface behavior for Clp

[Mikhail Nediak]

On Thu, 10 Apr 2003, Lou Hafer wrote:

> Mikhail,
>
> 	My initial reaction to this was ``you want to do what!?''. But after
> looking at the example code fragment, I'm sort of intrigued. I've just started
> back into coding, and have only glanced over the OsiSimplexInterface class.
> This will be an interesting case.
>
> 	But the later statement --- ``should the solver care at all ... when
> OsiSimplexInterface mode is enabled?''  --- does give me pause. There are lots
> of basic sanity and numerical accuracy checks that would fail, if the client
> expected dylp to perform a sequence of pivots that resulted in loss of
> primal (dual) feasibility while in primal II (dual II) phase.

I had the impression that OsiSimplexInterface should let user execute an
arbitary sequence of pivots (among other things). Of course, this is
not possible if a solver really is in primal/dual phase. The question
should probably be: does it really have to be at all times?

Best,
Mikhail

>
> 							Lou
>
> _______________________________________________
> Coin-lpsolver mailing list
> Coin-lpsolver at www-124.ibm.com
> http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-lpsolver
>