GNU Octave 4.0 Released, Includes A GUI & OpenGL

[valeriodean]

2k is bearable. The ad revenues aren't that small. Didn't Michael just buy a server with 96 GB of ECC RAM for this forum software?

I don't care if is bearable or not, the final result will be 2k money to a closed software and 0 to the open ones. This is more than enough to cancel that proposal instantly.

[DeepDayze]

Would be nice to see if anyone could try throwing Matlab programs at Octave to see how it handles them

[jabl]

Would be nice to see if anyone could try throwing Matlab programs at Octave to see how it handles them

Considering that it's over 20 years since Octave 1.0, and MATLAB compatibility is an explicit goal of the project, I'm not sure if you're trying to be sarcastic, or what are you trying to say?

So anyway, at the language level itself compatibility is pretty good. The problem is the toolboxes. But here's the rub: The MATLAB language itself is frickin horrible, and the only sane reason to use MATLAB is the toolboxes, many of which are actually good or offer unique functionality not available elsewhere. If there's no existing toolbox for the domain you're working in, you're better of with something like Scientific Python, R, or Julia.

[TheBlackCat]

Would be nice to see if anyone could try throwing Matlab programs at Octave to see how it handles them

Pretty much all Matlab code will run in Octave, if you have the functions the code calls. The opposite, however, is not true, since Octave fixes some of the stupider and more glaring omissions of the Matlab language. The language features Octave supports are basically a superset of those Matlab supports.

The issue, as jabl said, is toolboxes.

[TheBlackCat]

The problem is the toolboxes. But here's the rub: The MATLAB language itself is frickin horrible, and the only sane reason to use MATLAB is the toolboxes, many of which are actually good or offer unique functionality not available elsewhere. If there's no existing toolbox for the domain you're working in, you're better of with something like Scientific Python, R, or Julia.

Even if your language of choice (generally Python, R, or Julia) lacks the function you want (which is becoming less and less common every day, but still true in a few fields), you are still probably better off writing your code in another language and using something like pymatbridge to call the Matlab toolbox from your language.

[Naib]

I got matlab installed on a linux box at work. anyone got a testrun todo

The issue here however is ... Matlab vs Octave vs scilab isn't comparable, scilab is the outlier for one and matlab & octave might be compatible from a mathscript point of view, there will be implementation differences between calls and as such the results can easily be swayed purely from a script.

In all fairness though... matlabs power is with Simulink and there is literally nothing comparable. Modelica came along and overnight Mathworks came out with Simscape for Simulink to provide a similar modeling domain. Equally matlab has a verification & Validation toolbox that for big project's is indispensable.

These days unless it is simpowerSystems or simscape work, I usually mess around with my data in python+numpy+matplotlib... I even have a close-loop 3 control loop z-domain simulation running in python that actually outperforms matlab


Xcos is an attempt at s-domain modeling but in all fairness... its shite

[rdnetto]

Last I checked (2010), MATLAB was around 5x faster than Octave. However, since then MATLAB has started using LAPACK internally (which Octave also uses), so the differences might have reduced.
Realistically, Octave is only useful if you need to run MATLAB programs - Python with the Scipy and Matplotlib libraries is *much* nicer to work with, though I'm not sure if it has equivalents for all of the MATLAB toolboxes.