Bug 100757 - Linear solver says 'model is infeasible', when real problem is non-linearity
Summary: Linear solver says 'model is infeasible', when real problem is non-linearity
Status: NEW
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: Calc (show other bugs)
Version:
(earliest affected)
Inherited From OOo
Hardware: All All
: low trivial
Assignee: Not Assigned
URL:
Whiteboard:
Keywords:
Depends on:
Blocks: Solver
  Show dependency treegraph
 
Reported: 2016-07-04 12:55 UTC by Jan Egil Kristiansen
Modified: 2023-05-03 19:43 UTC (History)
2 users (show)

See Also:
Crash report or crash signature:


Attachments
Simple but non-linear optimization problem. (29.27 KB, application/vnd.oasis.opendocument.spreadsheet)
2016-07-04 12:55 UTC, Jan Egil Kristiansen
Details
Solver status message (81.00 KB, image/png)
2017-10-22 21:07 UTC, Thomas Lendo
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Jan Egil Kristiansen 2016-07-04 12:55:20 UTC
Created attachment 126062 [details]
Simple but non-linear optimization problem.

Optimizing a model with the linear solver, but the variables are multiplied with each other, making the model non-linear.

Error message says  'model is infeasible', I think it should rather say that 'model is non-linear'.

(When staring at a feasible solution, the 'infeasible' message is rather confusing.)
Comment 1 Aron Budea 2016-07-05 03:06:43 UTC
Confirmed.
According to specification [1], this text comes from the solver engine, the appropriate one in this case would be "The model is not linear."

[1] https://www.openoffice.org/specs/calc/features/Solver.odt
Comment 2 QA Administrators 2017-09-01 11:17:56 UTC Comment hidden (obsolete)
Comment 3 Thomas Lendo 2017-10-22 21:07:58 UTC
Created attachment 137214 [details]
Solver status message

I tried the attachment and got following status message:
"Process stopped at iteration 338 if 2000.
Process stopped due to stagnation."
Comment 4 Julien Nabet 2018-06-24 20:04:41 UTC
On pc Debian x86-64 with master sources updated today, I don't reproduce this.
I got the same result as Thomas, except the number of iterations is slightly different.
Comment 5 Aron Budea 2018-06-24 23:00:44 UTC
Thomas, Julien, did you open Options... in the Solver dialog, and switch to LibreOffice (or LibreOfficeDev) Linear Solver?
Comment 6 Julien Nabet 2018-06-25 18:20:14 UTC
(In reply to Aron Budea from comment #5)
> Thomas, Julien, did you open Options... in the Solver dialog, and switch to
> LibreOffice (or LibreOfficeDev) Linear Solver?

Thank you Aron, I could reproduce the pb.
Comment 7 Julien Nabet 2018-06-25 18:44:05 UTC
It seems "infeasible" may have different reasons (see http://lpsolve.sourceforge.net/5.5/).
CoinMP also uses this expression.
Comment 8 Todor Balabanov 2019-05-01 22:07:26 UTC
I am not sure that linear solvers are capable to detect non-linearity. That is why I do not think that error messages should be changed.
Comment 9 QA Administrators 2021-05-01 03:56:35 UTC Comment hidden (obsolete)
Comment 10 QA Administrators 2023-05-02 03:18:08 UTC Comment hidden (obsolete)
Comment 11 Jan Egil Kristiansen 2023-05-03 19:43:57 UTC
Bug still here as of 7.4.6.2

As I get older and wiser, I think it may be unreasonable to expect a linear solver to detect nonlinearity in the model.

Maybe the error message should just hint that nonlinearity might be the problem?