Thanks for this explanation. The Matlab symbolic toolkit sounds like
something that would be of limited interest to me. Calling Octave from
Maxima would probably be much more interesting to me (and to the average
Maxima user.)
Of course, if someone else is interested in implementing such a thing
for Maxima, he/she should feel free...
--Jim
On Thu, 2003-11-20 at 15:02, Richard Fateman wrote:
> The matlab symbolic toolkit roughly works like this.
> I am working from memory so I may have details wrong.
>
> declare some symbolic variables like
>
> x= symbol("x")
> y= symbol("y")
>
> then the matlab expression x+y is (by virtue of
> generic, object-oriented blah blah) converted to
> a symbolic version of addition.
>
> There are some hacks to allow symbols in matrices, making
> the whole enterprise much slower.
>
> The genericity is only partial, and stuff like
>
> x+1
> fails.
> You need to do x+symbol("1").
> I count this, and some other features of the matlab symbolic
> toolkit as "bugs".
>
> So: you could do it all with strings + changing Octave
> to have generic symbol types that go off to maxima and
> occasionally return back to Octave data structures.
>
> Why do it at all?
> A person who loves Octave may see this as a very low-cost
> introduction to symbolic computing.
>
> Is it going to be better than a real computer algebra system
> interface that links to high-quality back-end math?
>
> I doubt it.
>
> RJF
>
>
> James Amundson wrote:
>
> >On Thu, 2003-11-20 at 10:02, Richard Fateman wrote:
> >
> >
> >>On the topic of Octave, which I gather is a GPL version of Matlab...
> >>How hard would it be to make Octave be a front end for Maxima,
> >>similar to Matlab with symbolic toolkit being a front end for
> >>Maple? I have only given this 15 seconds of thought, but maybe
> >>someone familiar with Octave can say something.
> >>RJF
> >>
> >>
> >
> >I use Octave every day. I definitely think that some sort Maxima-Octave
> >link would be useful. Even though I have thought about linking the two
> >quite a bit, I still have conceptual questions about how the link should
> >work. Unfortunately, I have not seen the Matlab symbolic toolkit.
> >
> >Somebody has already presented some Maxima-Octave code on the Octave
> >list. It allows Octave to send a string to Maxima, which evaluates it
> >and returns another string. I don't see how that would be helpful at all
> >-- why not just have one window running Octave and another running
> >Maxima? The basic problem is that Octave doesn't really deal with
> >symbolic data. Perhaps we could have Maxima return expressions in Octave
> >form in such a way that Octave would evaluate them, but I'm still not
> >really sure how helpful that would be.
> >
> >Since Maxima deals with numerical *and* symbolic data, it might make
> >more sense to have Maxima operate as a front end to Octave. There are
> >issues with this approach also: doing purely numerical operations in
> >Maxima is annoying. Expressions tend to end up partially symbolic, e.g.,
> >(C1) 1.1 + sqrt(2);
> >
> >(D1) SQRT(2) + 1.1
> >
> >How would you imagine the interface working?
> >
> >--Jim
> >
> >P.S. There are also a set of technical issues involved in linking Maxima
> >and Octave. I see multiple ways of solving them, so I don't think there
> >should be any insurmountable problems there.
> >
> >
> >_______________________________________________
> >Maxima mailing list
> >Maxima@www.math.utexas.edu
> >http://www.math.utexas.edu/mailman/listinfo/maxima
> >
> >
>
> _______________________________________________
> Maxima mailing list
> Maxima@www.math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima