Re: [Help-glpk] Strict inequalities

[Andrew Makhorin]

>I would like to enforce strict inequality constraints.
>A natural solution seems to be: work with the supported non-strict upper
and
>lower bounds and add some very small quantity epsilon (some machine delta
>like number) to one side of the inequality.
>
>Is this the way it is supposed to be done, and if so, are there any GLPK
>builtin constants (perhaps any of the real-valued parameters) that are
>suitable to ac as such an "epsilon" (i.e. small enough, but not so small
that
>it would lose its use due to tolerance levels used by glpk) ?

You can increase a lower bound in the following way:

eps = 10.0 * lpx_get_real_parm(lp, LPX_K_TOLBND);

new_lb = old_lb + eps * (1.0 + fabs(old_lb));