Re: [Help-glpk] Writing Output File to MS Access

[Alex Morelli]

Hello Andrew,

Thanks for your reply.  I have tried the proposed approach and I get the
following:

-----------------------------------------------
Parameter(s) specified in the command line:
 --glp foo.glp -w foo.sol
Reading problem data from 'foo.glp'...
Foo.glp:7: error: too many data fields specified
GLPK LP/MIP processing error
-----------------------------------------------

When I run this model through gusek (as the original .mod file with the
reduced data set) the model solves, so I'm not sure how to prepare the
mps model to adopt your proposal.  I would appreciate any guidance on
this error.

I have interpreted your proposal as an iterative process - something
like this:

1.      prepare the problem with the reduced data set then follow the
proposal
2.      expand the data set and repeat
3.      continue performing 3 until solved (with all data) or out of
memory.

Is that correct?

Regards, Alex 
_________ 
Alex Morelli 
Bldg 6, 650 Church Street, Richmond VIC 3121
Mobile +61 400 365 754 / address@hidden


-----Original Message-----
From: Andrew Makhorin [mailto:address@hidden 
Sent: Friday, 2 September 2011 6:21 PM
To: Alex Morelli
Cc: address@hidden
Subject: Re: [Help-glpk] Writing Output File to MS Access

> Now, however, writing the output causes a lack of memory error and the
> model fails even with the reduced data set.  I have been watching the
> memory monitor through Windows Task Manager and note that the reduced
> data set finds an integer (they are all binary) optimal solution after
> about 2500 secs at which time the reported available memory is c.c.
> 1.9GB.
> 
> When I include the Access table output the memory reduces rapidly and
> the program fails.  The following is one of the Access output table
> syntax (which I assume is correct, because the smaller output tables
> update correctly).
> 

To reduce the memory requirement you may try to split the job into
several stages like follows:

1) translate the model and write it to foo.glp:

   glpsol -m foo.mod -d foo.dat --check --wglp foo.glp

2) solve the translated model and write its solution to foo.sol

   glpsol --glp foo.glp -w foo.sol

3) process the model using the previously found solution

   glpsol -m foo.mod -d foo.dat -r foo.sol