Lukas Reichlin wrote:
>> Lukas,
>> 
>> 
>> thanks for your comments. I'm not sure I understand your comment on
>>  "those automated changes just mess things up". The point of the
>> patch is that in some cases its better to use a manual change in
>> the source code, rather than applying an automated script.
> 
> IMO it doesn't make sense to partially convert the files when nobody
> does the manual changes afterwards, because they won't run on Matlab
> this way and it's possible they introduce errors. There are some
> issues like texinfo help strings and oct-files. AFAIK, Matlab even
> lacks a package manager.


Lukas,


of course it is desirable to have a perfect solution. Until this goal is
reached, Oct2mat is a great help and readly available for those who have 
the need of free toolboxes for matlab. And its not that nobody does the 
manual changes - I started doing it for functions that I needed, and 
extend the support and a few more. Others
might do so, too. Lack of support for texinfo help string and package
manager in matlab is not a crucial factor, I think. I'm happy with 
addpath/rmpath. But if there is a need for a package manager, someone 
might do it.


> If you are interested in porting control to Matlab, I urge you to
> copy the files to octave-forge/extra and name it something like
> control_ml. Then you could start porting the oct-files or running the
> conversion scripts.

Currently, I do not have plans to do this. But its good to see that you
are not opposed to the idea.

> 
> 
>> You are right Mathworks sells a control toolbox. But it does not 
>> mean that everyone with a Matlab license has also a license for the
>>  control toolbox. Just recently, I did a course at the university
>> in a computer lab with 20 people. Licenses for Matlab were 
>> available, and at the first glimpse it seems that also all
>> necessary toolboxes were available (we needed about 5 functions
>> from signal and statistics toolbox). However, when we tried to use
>> functions from these toolboxes, these licenses were locked to some
>> other department and we could not use them. Such a situation can
>> also happen to users of the control toolbox.
> 
> They should install Octave :-) There are even graphical frontends
> like QtOctave and XOctave for those who need such things.


Certainly. But at this event, only Windows machines were available and 
Matlab was already pre-installed. I had little options
about the setup. (and i do not want to say it loud, but M is still
faster, [1]. Admittingly, Octave has improved since 2006, and I see only 
difference by a factor of 2-3 instead of 4-5).

[1] http://arxiv.org/abs/cs/0603001v1


> 
>> Concering Oct-files: you are right, oct-files can not be converted.
>> For this reason, I'm prefering the use of the mex-interface. The
>> mex-interface is also open, and available to Octave. If you want to
>>  keep your the control toolbox to octave only, that's ok. But I can
>>  imagine that also matlab users might be interested in your work.
> 
> I prefer the OCT interface because it's object-oriented and fun to
> use. The MEX interface is neither of them. 


O-O and fun? not for me! I think that's a matter of taste, I prefer C
over C++.  Better control of memory management is also a good reason. 
But I do not want to start a discussion about C vs. C++, its good that 
both can be used simultaneously, and one can choose those concepts that 
are more suitable to solve a problem.

Following your argument, it would be of interest to make an 
OCT-interface for matlab (similar to the mex-interface for octave).


> Since control is licensed
> under GPL 3, I couldn't keep it Octave-only even if I wanted to. The
> reason is that I just have no motivation to put any effort into it.

That's fine. If there is a need for this, somebody else might step in.


Alois



------------------------------------------------------------------------------

_______________________________________________
Octave-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/octave-dev

Reply via email to