Re: Opinions on Matlab compatibility, Octave development

[fork]

> I am wondering, how do people feel about this at the moment? Playing
> the compatibility game does seem quite boring, but it's a very
> frequent request from Octave users. Furthermore, Octave has been
> listed in GNU's high priority project list[2] for some time as a
> replacement for Matlab.

I think Octave should continue to *strive* for compatibility overall, but with
exceptions.  I also think we should ensure that Matlab code (at least the
tradition non-OOP core) should always run on Octave, but compatibility in the
other direction is not as important.

Why?

1.  I actually think the core matlab language is the most elegant way of
expressing numerics out there today.  I am not much of a fan of numpy or perl;
matlab code looks and feels FAR MORE like the corresponding mathematics
(including 1-indexing) -- this is a huge strength.

2.  (VERY IMPORTANT) Part of the recent uptake (I think) in Octave has been
professors mentioning it as a cheap/ free alternative to a student license in
engineering courses.  It should remain possible to teach Matlab programming
through Octave, at least in the near term.

Exceptions I would like to see:

1.  Named parameters a la Python like  "z = f(x=3, y=[1 2 3])"

2.  A pointer/ reference notation for structs so that they can modified inline.
 Otherwise I am fine with pass by value.

3.  Some sort of coherent syntax for heterogeneous tables like datasets in SAS
or data.frames in SPLUS/R. (I wish I had a good idea what these might look 
like).

4.  Continued work on running octave scripts in an integrated Unix environment.

Other than that I am pretty happy with both Matlab and Octave.  Note that I DONT
actually want an extensive OOP system in Octave; I think it adds more complexity
than its worth (but we all choose our poisons,...).

> Further down he laments:
> 
> > Octave has never developed a strong core of dedicated and competent
> > developers.
> 
> This certainly seems to have changed since 2001, hasn't it? I would
> have to generate a code swarm or look at it with gource, but at least
> in the old one from a couple of years ago[3], there does seem to be
> quite an explosion approximately in the middle the 2000's. I think
> there's even a greater speedup around the time when the sources moved
> to hg.

Indeed -- I think Octave turned a corner in the last few years.  This is the
beauty of open source projects -- they never go away, and each incremental
improvement adds to the momentum.  We may use it as a prototype language at work
for some simulations that will (supposedly) be done in "real" Matlab; we may
never need to use the "real" thing, if Octave development continues apace.

THanks to all!