[Scip] How to add linear constraints during solve s.t. these survive complete restart?

[Matthias Walter]

Hi Marc,

Thank you for the explanation! Your suggestions are good, but I'm in the
lucky position that I am in fact solving LPs with different objectives
and hence can use (although maybe not recommended?) the
SCIPlpi-interface instead. And a warm-start for the LP-solver with the
same unchanged constraint-set shouldn't be too hard. I only have to
replace my constraint handler by a simple loop which searches for
inequalities to be separated.

Best & many thanks,

Matthias

On 07/04/2014 05:14 PM, Marc Pfetsch wrote:
> 
> 
> Hi Matthias,
> 
> as far as I see, this is currently only possible if you add the
> constraint if SCIP is in the problem stage, i.e., if you already freed
> the solving data. Thus, you would need to store the data of the cuts and
> then create the constraints after freeing the transformed problem. Of
> course, you would then need to use the original variables (and not the
> transformed ones).
> 
> If you want to proceed this way, you need to establish a way of storing
> the information and making sure that the data is not freed if you free
> the transformed problem (this should not be too hard). But if you have
> written this part of code, it seems that it would be no problem to
> separate over this data during an ordinary run. However, adding the
> constraints before the solution would include it in the presolving process.
> 
> Best
> 
> Marc
> 
> On 07/04/2014 04:05 PM, Matthias Walter wrote:
>> Dear list,
>>
>> I solve a certain problem very often and (using a constraint handler) I
>> generate linear constraints during the solving process. What I want is
>> that after I free the transformed problem/solve data, change some
>> variable objectives and call solve again, these linear constraints are
>> available so I don't need to separate them again. I tried to create real
>> constraints within the enforcement/separation callbacks (and not just
>> add rows) but that didn't help them to survive a new start.
>>
>> Is this possible with SCIP, e.g., by adding the constraints w.r.t. the
>> original variables and not their transformed counterparts?
>>
>> Best regards,
>>
>> Matthias
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> http://listserv.zib.de/mailman/listinfo/scip
>>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip