[Cbc] Osi Interface for Cbc (and deprecation thereof)?

[Ted Ralphs]

To echo what Stefan said, there is not really general agreement about the
usefulness/appropriateness of OsiCbc  and/or the usefulness of Osi for
controlling MILP solvers in general. I personally find Osi is perfectly
suitable and useful for calling MIP solvers and allows the ability to
easily switch out different solvers, which is exactly what Osi is supposed
to do.  Yes, it is true that the design of Osi is geared towards LP solvers
and the user has to understand that some of the functionality will not work
when solving a MILP. Nevertheless, Osi is perfectly serviceable as an
interface to MILP solvers until there is a better alternative (which will
hopefully happen at some point).

With that said, OsiCbc is only "deprecated" inasmuch as it is not
well-maintained. However, I am personally using it for cases when I really
do need to access solvers through Osi. It works (subject to the
aforementioned performance penalties). I would be very happy to see it
worked on and its performance improved. I would also be happy to help
maintain it going forward.

Cheers,

Ted


On Tue, Feb 18, 2014 at 11:16 PM, Stefan Vigerske
<stefan at math.hu-berlin.de>wrote:

> Hi,
>
>
> On 02/18/2014 05:56 PM, Matthew Gidden wrote:
>
>> Hi all,
>>
>> I was curious about the status of an OsiSolverInterface for Cbc models. I
>> note that the current OsiCbcSolverInterface is deprecated, and it's
>> suggested that I use the CbcSolver interface, hence my conundrum: I'd like
>> to use the virtual Osi interface, but I'm wary of using something that's
>> deprecated. I was hoping the list could help me understand:
>> a) why the Osi Cbc interface was deprecated in the first place
>>
>
> Noone has been developing or updating it anymore. And the performance you
> get with it is inferior to what you get with the usual CbcSolver interface.
> Also MIP support in Osi is very limited, so interest for reviving OsiCbc
> by the Cbc developer has been very low, I believe.
>
>
>  b) a suggested path forward (I assume writing my own custom interface for
>> the CbcSolver)
>>
>
> driver3.cpp and driver4.cpp in Cbc/examples may be a starting point.
>
>
>  If the answer is writing a new Osi-to-(new)CbcSolver interface, I'd be
>> happy to contribute that back into coin.
>>
>
> I know some people who would like this to see.
>
> Stefan
>
>
>> Warm regards,
>>
>>
>>
>> _______________________________________________
>> Cbc mailing list
>> Cbc at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/cbc
>>
>>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>



-- 
Dr. Ted Ralphs
Associate Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20140219/15324099/attachment.html>