Re: [Help-glpk] Certificates of infeasibility or unboundness

[Andrew Makhorin]

> Thank you for your answer ! You make it sound like this is no more
> than a 10 minutes work :-)

In a particular case, may be.

> 
> I do not get your last sentence though "Of course, there must be api
> routines that peforms such calculations.". Do you mean "there should"

Oh, sorry for my poor English. I meant that it would be nice to have
such api routine, but currently glpk doesn't have it.

> ? Along you answer, you already pointed me to the names of the
> functions I would need like glp_transform_row...
> 
> Now, it is true that my intent was to write such a code in Sage, but
> it is hardly where it belongs. That's shame there is no dedicated
> function for this in GLPK : would you think it worth spending a few
> nights trying to write a patch for GLPK ? I do not know its code, so I
> don't expect to be able to produce it within the month, but it would
> definitely be smarter than writing it inside of Sage's source code
> where nobody can reuse it ....
> 

I don't think you need to patch the glpk code. You could write a simple
routine, which starting from the final basis reported by glp_simplex
minimizes the sum of primal infeasibilities (for example, using a
standard technique by including slack/surplus/artificial variables in
every constraint and then minimizing the sum of such variables).