Re: WHICH FREEWARE MATLAB CLONE IS BETTER? Scilab or Octave

[John W. Eaton]

On  4-Feb-1997, Jim Van Zandt <address@hidden> wrote:

| Scilab has some symbolic capabilities the others lack.  I believe a
| polynomial is one of its primitive data types.  I don't think it is
| similar enough to matlab to justify the label "clone".
| 
| The Octave syntax is closer to matlab.  Octave uses well known
| numerical codes (lapack, for example).  I don't know about Scilab.

Looking at the Scilab source, it appears to support an interface to
Maple.  I don't think there are any technical reasons that Octave
couldn't support a similar interface.  So far though, no one has
volunteered to write it.

It also looks like Scilab uses dassl, lsode, minpack, etc., and that
it it uses lapack, but not for all linear algebra computations.  For
example, the lu, qr, qz, and svd factorizations don't appear to use
lapack (perhaps modified linpack subroutines?).

Also, if you compare the file routines/interf/matsvd.f from the Scilab
2.2 distribution to the file src/matfn3.f from the `classic' Matlab
sources, I think you will find some striking similarities.  Likewise
for most of the other functions in the routines directory and the
parser.  I expect that it is a fairly big job to add new features to
the Scilab language.

One of the goals for Octave has always been to make it reasonably easy
for users to modify and extend it by adding new functions or even by
experimenting with the language.  I'm not sure if I've really achieved
that goal yet, but there have been a few successes.  With the addition
of user-defined data types and dynamically linked functions in 2.x, I
think it should be even easier for users to experiment.  Now if I
could just finish the revisions to the manual, more people might be
able to take advantage of the new stuff.  I suppose I should get
back to work instead of reading mail...

:-)

jwe