Dear all,
I would like to use glpk in order to obtain exact rational solutions
of linear programming problems (all variables rationals, no integer).
Unfortunately, since GLPK is not thread safe, I have to use the
glpsol as an external program called from a C++ program.
When I use the option --exact and --xcheck I get unfortunately
a vertex with fractional coordinates expressed as 1.666667 instead
of the expected 5/3. I know I can compute the exact solution
since I have a basis of incident inequalities. But it is a little bit sad
to have to recompute something which is likely available from the
model itself.
Also, the documentation glpk.pdf gives the following for using
the model. It says that it is better to first use glp_simplex and then
use glp_exact in order to achieve good result. Reading the source
code, it seems that this is achieved by using --xcheck (but not --exact)
for solving the problem.
Mathieu