gmane.comp.gnu.glpk

Re: "The conflict graph is either empty or too big"

glpk xypron — 2012-05-15T00:25:50

Hello,

looking at your model, I would suggest considering the following
solution approaches:

- a time windowing heuristic, which will try to find a solution for a first time windows, fixes the solution for a first part of the time windows and then moves the window for which a solution is sought

- a column generation approach
(see
Column Generation, Guy Desaulniers, Jacques Desrosiers und Marius M. Solomon, Springer, ISBN-13: 978-1441937995)

Best regards

Xypron

-------- Original-Nachricht --------

Re: "The conflict graph is either empty or too big"

spiritfire — 2012-05-14T15:14:47

I was told that it is quite normal for a MILP problem.

But the lines haven't change for quite a while :

+258023: mip =     not found yet >=  5.819440044e+004        (242; 34)
+260157: mip =     not found yet >=  5.819440044e+004        (242; 34)
+262192: mip =     not found yet >=  5.819440044e+004        (242; 34)
+264262: mip =     not found yet >=  5.819440044e+004        (242; 34)
+266054: mip =     not found yet >=  5.819440044e+004        (242; 34)
+267696: mip =     not found yet >=  5.819440044e+004        (242; 34)
+269282: mip =     not found yet >=  5.819440044e+004        (242; 34)
+269824: mip =     not found yet >=  5.819440044e+004        (243; 34)

For 15-30 minutes, only the inside of the parenthesis has changed...

Since I'm running on Gusek, I don't know which solver is used neither how to
change it :-/

If anybody can help ? Should I wait more or stop it and change some
parameters ?

Thanks a lot.

B.

spiritfire wrote:

Re: "The conflict graph is either empty or too big"

Haroldo Santos — 2012-05-14T14:43:37



Perfectly normal. Integer programming is hard and solvers can take enormous
amount of time to solve it.
You could try:
  - use another solver
  - improve your formulation
  - try different glpsol parameters (ex.: fpump )

"The conflict graph is either empty or too big"

spiritfire — 2012-05-14T14:40:00

Hi, 

I'm running a simulation and it's taking quite a while...

I'm running on Gusek, which is the only way I found to use GLPK (couldn't
figure out how to install on win)


The program seems to found a solution :

OPTIMAL SOLUTION FOUND
Integer optimization begins...
Gomory's cuts enabled
MIR cuts enabled
Cover cuts enabled
Clique cuts enabled
Creating the conflict graph...
The conflict graph is either empty or too big


But then it goes on searching for quite a while :
http://old.nabble.com/file/p33830878/simul.mod simul.mod 

Time used: 1294.0 secs.  Memory used: 55.1 Mb.
+176120: mip =     not found yet >=  5.813433080e+004        (195; 6)
+176321: mip =     not found yet >=  5.813433080e+004        (196; 6)

I have no clue if it's close to the solution, if it's taking too long or if
I should wait or ... ?

Does anybody know if it is normal ?

my code is attached.

Thank you.

Re: install GLPK

Andrew Makhorin — 2012-05-14T10:41:45

The tar file is not a program, it is an archive that contains a set of
files. You can extract files from a tar file with a suitable utility
like tar under Linux or WinZip under Windows. For more details see
http://en.wikipedia.org/wiki/Tar_(file_format) . 

Once you have unarchived glpk-4.47.tar into a directory, you may then
compile and install glpk by running the configure script and make
utility (under Linux) or by running one of batch scripts in subdirectory
w32 or w64 (under Windows). For more information on compiling glpk on
various platforms see http://en.wikibooks.org/wiki/GLPK .

Re: [Fwd: need syntax help for parameter statement]

Xypron — 2012-05-13T14:46:40

Hello,

your mail had to be forwarded manually because you did not subscribe to
the GLPK help list, see
https://lists.gnu.org/mailman/listinfo/help-glpk

You missed to write ":=" after the param statement.

If the value of TP does not depend on i, there is no need to use it in
your if statements.

Best regards

Xypron

set N := {1..4};
set M := {1..288};

param TP{i in N, k in M}, binary :=
  if (k<=96 or k>=216) then 1 else 0;

display TP
end;



On 13.05.2012 15:29, Andrew Makhorin wrote:

[Fwd: need syntax help for parameter statement]

Andrew Makhorin — 2012-05-13T13:29:27

-------- Forwarded Message --------
From: spiritfire <bmattlet< at >gmail.com>
To: Help-glpk< at >gnu.org
Subject: [Help-glpk] need syntax help for parameter statement
Date: Sun, 13 May 2012 03:10:20 -0700 (PDT)

Hi, 

I'm trying to declare a parameter of that kind :

param TP 'Time preference' {i in N, k in M}, binary;   # (with N = 1..4 and
M = 1..288)

The thing is that the values of my TP parameter should be easy to write
since it looks like this :

0,...,0,1,...,1,0,...,0
0,...,0,1,...,1,0,...,0
0,...,0,1,...,1,0,...,0
0,...,0,1,...,1,0,...,0

And I really don't want to write the 4x288 values !

Only the number of "1" and the location in the line is different for each
lines.

I tried this (I know it's quite ugly and wrong) : 
param TP 'Time preference' {i in N, k in M}, binary : 
if i=1 then (if (k<=96 or k>=216) then 1 else 0) 
else (if i=2 then (if (k<=96 or k>=216) then 1 else 0) 
else (if i=3 then (if (k<=96 or k>=216) then 1 else 0) 
else (if i=4 then (if (k<=96 or k>=216) then 1 else 0) ) ) ) 

Any he

Re: [Fwd: Calling GLPK with VBA Excel ?]

glpk xypron — 2012-05-11T22:14:47

Hello Samir,

your message had to be forwarded manually, because you are not subscribed to the GLPK help list, cf.
https://lists.gnu.org/mailman/listinfo/help-glpk

VBA offers command "SHELL" to call another program. You can use this to call glpsol, cf.
http://en.wikibooks.org/wiki/GLPK/Scripting_plus_MathProg#GLPSOL_and_Visual_Basic_for_Applications

Best regards

Xypron

[Fwd: Calling GLPK with VBA Excel ?]

Andrew Makhorin — 2012-05-11T21:08:22

-------- Forwarded Message --------
From: Samir S <s10mir< at >hotmail.com>
To: help-glpk< at >gnu.org
Subject: Calling GLPK with VBA Excel ?
Date: Fri, 11 May 2012 18:22:32 +0100

Hello all ! I succeeded to build with GLPK Lab a running model (.mod)
which collects data from an Excel file and record results on it.

But I wonder if there is any way to call and run this model straight
from Excel using VBA without using the interface of GLPK Lab. 

p.s. I'm not an IT specialist

Thx,

Re: [Fwd: GLPK 4.38 Re-entrancy 64-bit windows]

glpk xypron — 2012-05-11T16:00:37

Hello Nicholas,

your message had to be forwarded manually because you are not subscribed to the GLPK help list.

No version of GLPK is thread safe. The current version of GLPK is 4.47.

Best regards

Xypron

Re: Having trouble specifying a constraint: syntax error inliteral set

glpk xypron — 2012-05-11T15:57:50

Hello Joel,

the error must be in another line. Have a look at your definitions of the sets.

The following model does not create a syntax error:

set Campaigns:= {1..5};
set D:= {1..31};
set P:= { "A", "B", "C"};
set CampProd, dimen 2 := { (1,"A"), (2,"B"), (3,"C"), (4,"A"), (5,"B") };
set ProdDay, dimen 2 := { ("A",5), ("B",4) };

var Q{P,D}, >= 0;
var B{Campaigns,D}, >= 0;

s.t. quant_consumed{(p,d) in ProdDay}: Q[p,d] <= 9999999 * sum{c in Campaigns: (c,p) in CampProd} B[c,d];
end;

Best regards

Xypron

-------- Original-Nachricht --------

[Fwd: GLPK 4.38 Re-entrancy 64-bit windows]

Andrew Makhorin — 2012-05-10T22:28:47

-------- Forwarded Message --------
From: Nicholas Nash <nicholas.nash< at >gmail.com>
To: help-glpk< at >gnu.org
Subject: GLPK 4.38 Re-entrancy 64-bit windows
Date: Thu, 10 May 2012 20:12:13 +0100

Hi,

I have some legacy code that is using GLPK 4.38, despite much searching
online I can’t get definitive answers to some questions.

Is GLPK 4.38 re-entrant? The code I have is multi-threaded and running
in a Windows 64-bit environment, calling GLPK from C#.

A single GLPK problem is only ever accessed in a single thread; although
many threads create their own GLPK problems and solve them.

I can see code in glplib02.c that allocates thread local storage, so am
I correct that the usage I have described above is safe?

Despite using GLPK as above, I still occasionally see this error
reported:

xfree: memory allocation error

Error detected in file ..\src\glplib07.c at line 181

Since I’m not out of memory, guessing at the code makes it looks like
there was an unmatched free/delete (or there was a t

Having trouble specifying a constraint: syntax error inliteral set

joel mortyn — 2012-05-10T22:19:51

Hello,
I'm quite new to GLPK and am having difficulty specifying a constraint.  I am modelling production at a furniture plant.  The following constraint is trying to constrain the quantity (Q) of products (p) entering the plant each day (d).  In order for products to enter the plant, it must be set up in a compatible campaign mode (c) which consumes that product {(c,p) in CampProd}.  B[c,d] is a binary variable that specifies whether a campaign will be run on a given day.
The constraint produces a "syntax error in literal set" message:
s.t. quant_consumed{(p,d) in ProdDay}: Q[p,d] <= 9999999 * sum{c in Campaigns: (c,p) in CampProd} B[c,d];
Any help would be much appreciated.       _______________________________________________
Help-glpk mailing list
Help-glpk< at >gnu.org
https://lists.gnu.org/mailman/listinfo/help-glpk

Re: Generate data section files for model (in GNU MathProg or other languages)

glpk xypron — 2012-05-10T06:09:42

Hello Tony,

See example:

# Outputing results as GMPL data file
# Save model file as test.mod
# 1st run: glpsol -m test.mod
# 2nd run: glpsol -m test.mod -d test.dat

param f, symbolic := "test.dat";
set S, dimen 3 := { (0, 0, 1), (0, 1, 0), (1, 0, 0), (1, 1, 1) };
param p{(i, j, k) in S} := i + 2 * j + 4 * k;

set T, dimen 3;
param q{T};
param r;

printf "data;\n" > f;
# Output simple parameter
printf "data;\n" > f;
printf "param r := %f;\n", card(S) >> f;

# Output set
printf "set T :=\n" >> f;
for {(i,j,k) in S} {
  printf "%d, %d, %d%s", i, j ,k,
  if card(setof{(l, m, n) in S : l > i or m > j or n > k }(l, m, n)) > 0
  then ","
  else ";" >> f;
  printf "\n" >> f;
}

# Output indexed parameter
printf "param q :=\n" >> f;
for {(i,j,k) in S} {
  printf "[%d, %d, %d] %f%s", i, j ,k, p[i, j, k],
  if card(setof{(l, m, n) in S : l > i or m > j or n > k }(l, m, n)) > 0
  then ","
  else ";" >> f;
  printf "\n" >> f;
}

printf "end;\n" >> f;

display T;
display q;
display r;
end;

Anyway you will need some lo

Re: Generate data section files for model (in GNU MathProg or other languages)

Robbie Morrison — 2012-05-09T21:16:04

Hello again Tony

------------------------------------------------------------
To:           Robbie Morrison <robbie< at >actrix.co.nz>
Subject:      Re: [Help-glpk] Generate data section files for model (in GNU
Message-ID: 
<CABq29No6ameSfQ_L7FZ22S=Nj7NWq+5J8qvorma=MRF3eK6ANA< at >mail.gmail.com>
From:         Tony Khosravi <tony.khosravid< at >gmail.com>
Date:         Wed, 9 May 2012 22:29:12 +0200
------------------------------------------------------------

[snip : original problem statement]

It makes sense for GLPK to stay within strict ISO C for
maximum portability.  But I think that your kind of
software could afford to make Boost a dependency.
Besides many of the Boost libraries have or are
migrating to the C++ standard library.  In addition,
many are header-only and you could copy the relevant
headers into your codebase if you wish.

Moreover I would think that adopting generalized
libraries like Boost.Filesystem would actually improve
your portability, not lessen it.

Also, a special plug for reference-counte

Re: Generate data section files for model (in GNU MathProg or other languages)

Tony Khosravi — 2012-05-09T21:03:28

Hi,

Thanks for your answer!

For the use of the "printf" command, the fact is that I find it not very
flexible. Indeed, I have to write in plain text what I want to write as
param. The syntax is different when the param is a simple list of records
or when it is tabbing data.

When you mention the use of a database and the table statement, isn't it
when you have one model and you want to run the solver on one set of data?
I am interesting in running the solver many times (with my c++ program)
through many records of data with only on model.

The use of table statement here is limited because the table statement
cannot be used in the data section (not as far as I know). Or in that case,
I would have to generate as many models as the number of data that I want
to use by only changing the part related to table statement.As I want to
keep in very simple (that is being able to use the program on different
computer without having to copy whole bunch of data structure like a
database), maybe the use of a CSV file w

Re: Generate data section files for model (in GNUMathProgor other languages)

glpk xypron — 2012-05-09T20:41:52

Hello Tony,

GMPL data files can be easily created with any tool that can output text files, cf. 
http://en.wikibooks.org/wiki/GLPK/Scripting_plus_MathProg

You could use the printf command of GMPL to create data files.

When many data shall be used with the same model, I personally would use a SQL data base, which can be read with the table statement.

Best regards

Xypron
-------- Original-Nachricht --------

Re: Generate data section files for model (in GNU MathProg or other languages)

Tony Khosravi — 2012-05-09T20:29:12

Thanks for that Robbie

2012/5/9 Robbie Morrison <robbie< at >actrix.co.nz>

I knew the Boost library, but I never used them. I wanted to stay with the
standard library and not use any extra library (the code that I am
developing ought to be used by others people on other machine and there are
not computer science people, so I want to keep things simple and in the
standard c++ box).
I will have a look anyway on those libraries that Boost provides.

As you mentioned, I wanted to use some functors and/or template to achieve
this kind of translator to the MathProg format.
I guess this is what I wil try to do.

Thanks for the help anyway.

Regards,

Re: Generate data section files for model (in GNU MathProg or other languages)

Robbie Morrison — 2012-05-09T19:42:26

Hello Tony

------------------------------------------------------------
To:           Robbie Morrison <robbie< at >actrix.co.nz>
Subject:      Re: [Help-glpk] Generate data section files for model (in GNU
Message-ID: 
<CABq29NpOyf5szSeTvkvZDQE-is61Kfztf-v2sZua5MEfuv56Xg< at >mail.gmail.com>
From:         Tony Khosravi <tony.khosravid< at >gmail.com>
Date:         Wed, 9 May 2012 20:06:11 +0200
------------------------------------------------------------

d'Informations

I am not aware of any GLPK functions, official or
internal, that do what you suggest.  Someone else may
know for sure.

The following Boost C++ libraries may be of interest:

    http://www.boost.org/doc/libs/1_49_0/

    Boost.Smart_ptr
    Boost.Format
    Boost.Filesystem

Or STL std::ostringstream instead of Boost.Format.

Depending on your skill levels, maybe also C++ functors, custom
steam inserters, and the following:

   Boost.Bind
   Boost.Lambda

It would seem that fully fledged parsing is not
indicated, otherwise see Boost.Spirit.

Write tight

Re: Generate data section files for model (in GNU MathProg or other languages)

Tony Khosravi — 2012-05-09T18:06:11

Thank you for the answer Robbie,

2012/5/9 Robbie Morrison <robbie< at >actrix.co.nz>

I would say that the last link is partially the answer to my question.

I will try to explain better what I want to do.

So as I said, I would like to generate a file containing only data section
through a program in a way that I can generate many of those data files
through a program, and then run the GLPK solver on each data file according
to a model. So you understand that the data section and the model section
are in two different files.

I think I need to say that I program in C++.

Let's say that my program is split into three modules:
   - one which generates data (the nature of those data are not important)
through N iterations
   - one which would store those data into  N files (data section in a text
file)
   - one which read those data files one by one and run the solver (given a
file which contains the model section)
The first and third modules are implemented, so I am interested in
implementing the second one.

In

Re: Generate data section files for model (in GNU MathProg or other languages)

Robbie Morrison — 2012-05-09T17:04:18

Hello Tony

------------------------------------------------------------
To:           help-glpk< at >gnu.org
Subject:      [Help-glpk] Generate data section files for model (in GNU
MathProg or other languages)
From:         Tony Khosravi <tony.khosravid< at >gmail.com>
Date:         Wed, 9 May 2012 17:13:36 +0200
------------------------------------------------------------

Hello Tony

AFAIK, once your high-level MathProg model has been
translated into a low-level problem object, it is not
possible to detranslate it.

So I guess you are talking about doing this processing
before your MathProg model + data has been irrevocably
translated.

You may therefore need to do this work outside of GLPK?
That is my best guess anyway.

These links might be useful while thinking about your
problem:

  http://en.wikibooks.org/wiki/GLPK/Steps_of_GMPL_File_Processing
  http://en.wikibooks.org/wiki/GLPK/Interoperability
  http://en.wikibooks.org/wiki/GLPK/Scripting_plus_MathProg

best regards
---
Robbie Morrison
PhD student -- poli