Re: [Help-glpk] un reduced filed

[Kevin Hunter]

At 3:34pm -0400 Sat, 06 Nov 2010, Andrew Makhorin wrote:
> > Am I incorrect in thinking that no matter where it is, a term
> > multiplied by 0 is effectively useless information? It is perhaps
> > of note that I only see any +0 terms in the objective function.
>
> Zero objective and constraint entries (i.e. entries like 0 * x) are
> allowed, however, you do not need to specify them, because glpk
> assumes sparse data format: if some constraint matrix or objective
> coefficient is omitted, it is considered as zero by default.

Right.  I'm aware that I don't need to specify them, but they're in the 
LP output of another like program, with which I've recently become 
acquainted.  I'm asking why does GLPK leaves them in?  Aren't they 
superfluous.

> Btw, you can reduce the lp file size much more by appending the
> suffix '.gz', for instance, --wlp smaller_file.lp.gz.

Oh!  I must have missed that in the documentation.  I knew that Gurobi 
transparently supported .gz, and I guessed that GLPK did via the output 
from ./configure (--with-zlib), but I hadn't yet figured out where.

However, I actually want to reduce the LP as much as possible so that I 
can have an easier time visually inspecting it.  The LP is the output of 
another project that I'm working on, of which I'm sometimes suspicious.

In a similar vein, is it possible to presolve *without* solving?  In an 
effort to reduce the file size, I tried this:

parm.presolve = GLP_ON;
parm.tm_lim = 500;  // Something really small because what I want
                    // is the presolved instance.
glp_simplex( lp, &parm );
glp_write_lp( lp, some_file );

As I'm sure you're aware, this didn't do any good.

Finally, I was hoping to use Heinrich's code to reduce the output some 
more, because my LP file contains a number of constraints that could be 
merged:

C1: x >= 5
C2: x <= 5

But after compiling, I wasn't sure how to use it.  Do you have any 
suggestions of how else I could reduce the redundancies in my LP file 
(of which there are *many*)?

Thanks,

Kevin

P.S. If you're curious, here's an xzipped version of my 296 MiB LP file:

http://cygnus.ce.ncsu.edu/~kevin/tmp/co2ng_4branch.lp.xz [5.9MiB]