Re: [Help-glpk] inf as RHS of constraint

[Andrew Makhorin]

> Is this "inf" as a key word or string -- or "inf" as a
> genuine IEEE 754 floating point infinity?
> 
>   http://en.wikipedia.org/wiki/IEEE_754

The "inf" keyword in cplex lp format is just an attribute indicating
that a variable has no lower/upper bound. Though this attribute can be
implemented as a ieee754 inf, it is not essential, because such value is
not used in computations.

As to using inf in rhs, the cplex doc says nothing about that. This
would have sense only to allow free rows (like in mps), however, free
rows being redundant do not need to be included in the model.

> 
> The reason I ask is that I recently fed GLPK a NaN and
> it simply printed a short message and died:
> 
>   http://lists.gnu.org/archive/html/bug-glpk/2011-01/msg00001.html
> 
> In this regard, might it also be useful if GLPK
> recognized and processed floating point 'inf's ?
> 
> Clearly, 'inf's carry more meaning than 'NaN's!
> 

This problem will be fixed. As I explained earler I would like to keep
the glpk code written in ansi c 89 (not in 99), where fpclassify is not
available.