Basis read/write

[prem swaroop]

Team

There #39;s a couple of features documented in 4.34 in the stand-alone solver section (Appendix D):

(Options specific to simplex method)

--bas filename read initial basis from filename in MPS format

--wbas filename write nal basis to filename in MPS format

Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39.

Now, documentation for 4.39 does NOT mention these features - so that #39;s at least consistent.

I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here.

I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.

Regards
Prem
 
Team

There's a couple of features documented in 4.34 in the stand-alone solver section (Appendix D):

(Options specific to simplex method)

--bas filename read initial basis from filename in MPS format

--wbas filename write nal basis to filename in MPS format

Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39.

Now, documentation for 4.39 does NOT mention these features - so that's at least consistent.

I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here.

I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.

Regards
Prem

_______________________________________________
Help-glpk mailing list
Help-glpk@...
http://lists.gnu.org/mailman/listinfo/help-glpk

[prem swaroop]

Thanks Dave.

In the hope of achieving the same, I found and tried -w and -r options documented thus in version 4.39:

   -r filename, --read filename
                     read solution from filename rather to find it with
                     the solver
   -w filename, --write filename
                     write solution to filename in plain text format

However, after -r, the solver does not solve the problem at all (despite my specifying --min just to try).  This is consistent to documentation for sure.

I am unsure how far this can go - but there apparently are APIs that do these jobs, we only need their equivalent command-line versions:

3.3.2 glp read sol—read basic solution from text file
Description
The routine glp_read_sol reads basic solution from a text file whose name
is specified by the parameter fname into the problem object.
For the file format see description of the routine glp_write_sol.

3.3.3 glp write sol—write basic solution to text file
Description
The routine glp_write_sol writes the current basic solution to a text file
whose name is specified by the parameter fname. This file can be read back
with the routine glp_read_sol.

Thanks for a great work!

Regards
Prem

On Sun, Oct 25, 2009 at 12:31 PM, David Posner <dposner@...> wrote:
Hi Prem, I don #39;t have an answer but I asked about a few months ago in connection with a port I was trying to do from a commercial LP solver which supports saving/restoring bases from files and I never got an answer.  Anyway I second your request!

Dave Posner



----- Original Message ----
From: prem swaroop <premswaroop@...>
To: help-glpk@...
Sent: Sunday, October 25, 2009 5:08:46 AM
Subject: [Help-glpk] Basis read/write

Team

There #39;s a couple of features documented in 4.34 in the stand-alone solver section (Appendix D):

(Options specific to simplex method)

--bas filename read initial basis from filename in MPS format

--wbas filename write nal basis to filename in MPS format

Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39.

Now, documentation for 4.39 does NOT mention these features - so that #39;s at least consistent.

I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here.

I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.

Regards
Prem






 
Thanks Dave.

In the hope of achieving the same, I found and tried -w and -r options documented thus in version 4.39:

   -r filename, --read filename
                     read solution from filename rather to find it with
                     the solver
   -w filename, --write filename
                     write solution to filename in plain text format

However, after -r, the solver does not solve the problem at all (despite my specifying --min just to try).  This is consistent to documentation for sure.

I am unsure how far this can go - but there apparently are APIs that do these jobs, we only need their equivalent command-line versions:

3.3.2 glp read sol—read basic solution from text file
Description
The routine glp_read_sol reads basic solution from a text file whose name
is specified by the parameter fname into the problem object.
For the file format see description of the routine glp_write_sol.

3.3.3 glp write sol—write basic solution to text file
Description
The routine glp_write_sol writes the current basic solution to a text file
whose name is specified by the parameter fname. This file can be read back
with the routine glp_read_sol.

Thanks for a great work!

Regards
Prem

On Sun, Oct 25, 2009 at 12:31 PM, David Posner <dposner@...> wrote:

Hi Prem, I don't have an answer but I asked about a few months ago in connection with a port I was trying to do from a commercial LP solver which supports saving/restoring bases from files and I never got an answer.  Anyway I second your request!

Dave Posner



----- Original Message ----
From: prem swaroop <premswaroop@...>
To: help-glpk@...
Sent: Sunday, October 25, 2009 5:08:46 AM
Subject: [Help-glpk] Basis read/write

Team

There #39;s a couple of features documented in 4.34 in the stand-alone solver section (Appendix D):

(Options specific to simplex method)

--bas filename read initial basis from filename in MPS format

--wbas filename write nal basis to filename in MPS format

Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39.

Now, documentation for 4.39 does NOT mention these features - so that #39;s at least consistent.

I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here.

I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.

Regards
Prem

_______________________________________________
Help-glpk mailing list
Help-glpk@...
http://lists.gnu.org/mailman/listinfo/help-glpk

[xypron]

Hello Prem

reading the basis in MPS format is implemented in the stand alone solver
of GLPK 4.26 in file glplpx20.c. You can download it at
ftp://ftp.gnu.org/gnu/glpk/glpk-4.26.tar.gz

Best regards

Xypron

prem swaroop wrote:

Team

There #39;s a couple of features documented in 4.34 in the stand-alone solver section (Appendix D):

(Options specific to simplex method)

--bas filename read initial basis from filename in MPS format

--wbas filename write nal basis to filename in MPS format

Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39.

Now, documentation for 4.39 does NOT mention these features - so that #39;s at least consistent.

I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here.

I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.

Regards
Prem
 
Team There's a couple of features documented in 4.34 in the stand-alone solver section (Appendix D): (Options specific to simplex method) --bas filename read initial basis from filename in MPS format
--wbas filename write nal basis to filename in MPS format Both are missing from the glpsol that I built on cygwin/Windows Vista, for versions 4.34 as well as 4.39. Now, documentation for 4.39 does NOT mention these features - so that's at least consistent.
I suspect this feature existed earlier on.  I think this could be a useful feature to keep, and I can quote my case here. I am essentially solving a large-ish LP whose parameters are generated by incrementing one parameter value.  And so, I dont expect too many changes in the solution for a large part of my increments, and hence using a final basis for one solution as an initial basis for the next could save a lot in time.
Regards Prem

_______________________________________________
Help-glpk mailing list
[e-mail]
http://lists.gnu.org/mailman/listinfo/help-glpk

[Andrew Makhorin]

> There #39;s a couple of features documented in 4.34 in the
> stand-alone solver section (Appendix D):

> (Options specific to simplex method)

> --bas filename read initial basis from filename in MPS format

> --wbas filename write nal basis to filename in MPS format

> Both are missing from the glpsol that I built on cygwin/Windows
> Vista, for versions 4.34 as well as 4.39.

> Now, documentation for 4.39 does NOT mention these features - so that
> #39;s at least consistent.

> I suspect this feature existed earlier on.  I think this could be a
> useful feature to keep, and I can quote my case here.

> I am essentially solving a large-ish LP whose parameters are
> generated by incrementing one parameter value.  And so, I dont expect
> too many changes in the solution for a large part of my increments,
> and hence using a final basis for one solution as an initial basis for
> the next could save a lot in time.

Options --bas and --wbas were temporarily removed from glpsol, since
they supported only fixed mps format. This made them useless in most
cases due to limitation of row/column name length to 8 chars (in
mathprog models the names produced by the translator are, as a rule,
longer).



_______________________________________________
Help-glpk mailing list
Help-glpk@...
http://lists.gnu.org/mailman/listinfo/help-glpk