Re: [Help-glpk] Linear Programming in Java using GLPK and MathProg

[i.van.baren]

Hi Andrew,

The point here is that GLPK has no api routines available to map MathProg 
variable and index name to the associated column number in the resulting MIP or 
LP problem. Only this mapping would help Benson to read the values of the 
objective function and the decision variables without parsing the output file. 

Following your declarative approach of MathProg, I guess that Benson has to 
find another way to extract the solution components from a MathProg model in 
GLPK. In my previous response to his question I formulated a possible solution 
with xml files. 

In this solution you create a xml file as output of the MathProg program with a 
self made tag and tree structure. For reading xml files Java has many routines 
available. So, accessing the xml file with the appropriate variable tag will 
give the required value. 

I hope this clarifies the discussion. I can remember a previous discussion 
thread about this subject.

Best regards,

Ivo van Baren
Netherlands

---- Andrew Makhorin <address@hidden> schrijft: 
> > I am writing a university timetabling language and solver in Java and
> >  I want to use the linear programming approach. I want to use a free 
> >  solver and GLPK seems to be a good candidate. I intend the program to
> >  translate the timetabling problem into MathProg format and let the 
> >  solver solves it. It seems that readModel in GlpkSolver allows me to 
> >  do this, however it seems to output the result of the calculation in a
> >  file, is it possible to access parts of the solution without going to 
> >  the file? Can I access the values, e.g. the value of the objective 
> >  function in some ways? 
> 
> To obtain solution components you can use glpk api routines
> glp_mip_status, glp_mip_obj_val, glp_mip_row_val, and glp_mip_col_val.
> For more details see the reference manual, section "Mixed integer
> programming routines".
> 
> 
> 
> 
> _______________________________________________
> Help-glpk mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/help-glpk