[Help-glpk] Re: Definition of boundaries

[Andrew Makhorin]

>Is it better to define boundaries with the lpx_set_col_bnds() command or
>to define an extra row with x > 3 or is there no difference between
>these ways to set a boundary. I'm asking because I have to parse an
>external problem description, where the boundaries are defined as extra
>row and it would be easier not to change that. 

If it possible to provide variables with their explicit bounds, it is
desirable for many reasons and significantly better than specifying
their bounds as separate constraints. However, this is not a necessary
requirement (for example, the lp presolver, if used, is able to recognize
implicit bounds).
 
>I'm having the same problem with definitions like:
>Q_dot = 3
>Q_dot = 7 delta_T
>Which is the same as delta_T = 3/7 , but this calculations are tricky to
>do automatically. Do you think it would do a big difference to the
>calculation speed to replace these equations?

The rule is the same as for bounds of variables. If all this requires
much extra work or makes the model too intricate, you can solve the model
"as is". But if it is possible to simplify the model on the generation
stage with moderate efforts, it would be reasonable to do that.

Although redundant constraints and implicit bounds may increase the
solution time, but in many practical cases it is 5-10%, not more.