Re: [Help-glpk] glpk 4.42 release information

[Andrew Makhorin]

> Thanks, quickly looked at the reference manual and it looks really
> good. The sensitivity report preamble message has been turned off,
> which is nice. But I'm not sure what would happen if you give
> glp_analyze_coef a none-basic variable?

This is a trivial case which needs no analysis. For example, if
non-basic variable x have its lower bound active and its reduced
cost, say, 1.23, you cannot decrease the objective coefficient more
than on 1.23; otherwise the reduced cost becomes negative and the
basis becomes dual infeasible (non-optimal).

>  Similarly, what would happen
> if you give glp_analyze_bound a basic variable?

This is also a trivial case, even easier than the previous one.
For example, if primal value of basic variable x is 3.21, its lower
bound cannot be greater and its upper bound cannot be less than 3.21;
otherwise the basis becomes primal infeasible.

>  The reference manual
> does not mention this, it seems to me. BTW, I'm not sure how to check
> if a variable is basic or not, simply checking if the value is at one
> of its bounds is easy but not a reliable way to do this.

glp_get_row_stat/glp_get_col_stat should be used.

> I am very excited about this new release, maybe I will directly
> incorporate GLPK into PyMathProg, so that you don't need to install
> GLPK separately, nor worry about which version compatibility issues.