[Help-glpk] Re: current basis after solving the problem

[Andrew Makhorin]

>       I tried to apply 'lpx_eval_tab_row' after applying 'lpx_simplex',
>but it is saying that 'current basis is undefined'. I think in method 
>'simplex2', a new workspace has been created for the presolved problem, 
>and at the end of solution, the formulated problem has not copied to the 
>original problem( or the basis has not unloaded from the formulated 
>problem to the original). Am I right? How can I compute a row of simplex 
>table after finding the optmal LP solution in this case? 

Yes, you are right. Recovering basic solution the lp presolver stores
to the original problem object (LPX) all basis information, but
factorization of the basis matrix is not stored. To resolve the problem
you should either disable the lp presolver or call the routine
lpx_warm_up (after lpx_simplex), which factorizes the current basis
and recomputes basic solution components. Note, however, that due to
round-off errors it may happen that lpx_simplex reports that the basic
solution is optimal while lpx_warm_up reports that it is not. In this
case you can reoptimize your problem starting from the current basis
by calling lpx_simplex once again but with lp_presolver disabled (as a
rule it takes a few pivots).