symbolic toolbox in scilab 5
holgerna...@gmx.de
dcsd/idco/perso/Magni/toolboxes.html). It works quite well.
What are the plans for implementing a symbolic toolbox in scilab 5? I
have seen the symbolic directory in the subversion, but until now it
is quite empty.
thanks for answering my question.
ycollet
I've used this toolbox some months ago. I found some performances
problem but, in general, this toolbox is well defined and should be
supported by the scilab team (the socket interface to maxima should be
rewritten to bring some more performances).
Adding this toolbox to scilab will bring a real missing tool in the
current scilab.
YC
philippe
> I vote for this one !!!
me too!!
> I've used this toolbox some months ago. I found some performances
> problem but, in general, this toolbox is well defined and should be
> supported by the scilab team (the socket interface to maxima should be
> rewritten to bring some more performances).
I've discorered this toolbox quite recently. I only use the mexec()
function to add some symbolic computations in my scilab's scripts. Usage
of mexec is very easy although it is a bit slow, nevertheless this
toolbox is realy a great contribution to scilab!
Philippe.
Matthias Zenker
> hello, i'am using the maxima toolbox for scilab (http://www.cert.fr/
> dcsd/idco/perso/Magni/toolboxes.html). It works quite well.
I just searched the scilab toolboxes center http://www.scilab.org/contrib/
and did not find it. How can this be? Shouldn't it be part of the
scilab contributions as well as the other toolboxes mentioned on the
website above?
Matthias
Sylvestre Ledru
> hello, i'am using the maxima toolbox for scilab (http://www.cert.fr/
> dcsd/idco/perso/Magni/toolboxes.html). It works quite well.
>
> What are the plans for implementing a symbolic toolbox in scilab 5?
No, we don't have any plan for doing this.
However, this could be interesting but :
* there is a (mandatory ?) dependency on perl , we already have too much
deps in Scilab. Maybe this could be rewrite in Scilab or TCL ?
* We are not competant enough in the symbolic world to estimate this
toolbox.
* what about the "slow communication" between Scilab and Maxima ?
Prove/show me arguments for these points and we could do something. ;)
Sylvestre
ycollet
> * there is a (mandatory ?) dependency on perl , we already have too much deps in Scilab. Maybe this could be rewrite in Scilab or TCL ?
The perl is used only to send and get string from / to maxima via
socket.
> * We are not competant enough in the symbolic world to estimate this toolbox.
Maxima is really efficient in symbolic computation. Maxima looks like
R-project for statistics: a tool where each new method is implemented
for the community. So, a really up to date symbolic tool.
> * what about the "slow communication" between Scilab and Maxima ?
The communication is slow just because to send a string to maxima, you
need to first execute perl and then send the send via socket.
I think that this socket call can be easily replaced by a tcl call (I
am neither a perl specialist nor a tcl one).
Sylvestre Ledru
>> * there is a (mandatory ?) dependency on perl , we already have too
>> much deps in Scilab. Maybe this could be rewrite in Scilab or TCL ?
>
> The perl is used only to send and get string from / to maxima via
> socket.
>> * what about the "slow communication" between Scilab and Maxima ?
>
> The communication is slow just because to send a string to maxima, you
> need to first execute perl and then send the send via socket. I think
> that this socket call can be easily replaced by a tcl call (I am neither
> a perl specialist nor a tcl one).
As usually, thanks for the quality of your answer Yann !
If I understand well, Perl is just a detail and could be removed
quickly. We could open the socket in a native lib of Scilab and call
maxima from it. Am I right ?
Sylvestre
ycollet
> If I understand well, Perl is just a detail and could be removed
> quickly. We could open the socket in a native lib of Scilab and call
> maxima from it. Am I right ?
>
> Sylvestre
Yes. The perl script is called from 2 or 3 scilab files if I
remember ...
YC
ycollet
"interface"
The call to interface is done in:
- mexec.sci
- str2max2sym.sci
You can find in Symbolic/server another perl script: "serveur" which
is used to start maxima.
jfm
shall give some explanations.
1- I stopped the development of this toolbox because I expected some
help to replace the provisional Perl-based server, also, I received
almost no sign of interest for this work.
2- From now I cannot restart development of the toolbox, and for
implementation in SCilab someone else should lead this project (or a
team because a good knowledge of Maxima is also required, mine is
almost negligible). But a lot of work is still required for
"robustifying" the toolbox.
3- Some drawback are listed now.
- I wrote the core of this toolbox very quickly (a few days) so it
contains probably hundred of bugs.
- The Maxima version should be fixed, when I used the toolbox each new
version of Maxima introduced new bugs: a nightmare. So, I suggest to
choose a version and to join it to Scilab code. Maybe some features of
Maxima should be modified, for example the syntax of Maxima and Scilab
is not compatible for the function exp, even with regular expression,
I am not sure that the translation is feasible. I assume that Maple is
shipped with the Matlab symbolic toolbox for similar reasons.
- Why Perl: two reasons. First, it is very easy to implement server/
socket communication (platform independent) with Perl. I don't know
how to do that with C-code. It is probably easy because Scilab
communicates with TCL/TK. I suggest to write a function unix_e (e as
external program):
string_from_external_prgm =
unix_e(string_for_external_prgm,port_number,delay)
where the port number correspond here to the port used for launching
Maxima in server mode. Delay argument in case of problem in Maxima
(avoids freezing the system).
- Why Perl: second reason. I use Perl also for a preliminary treatment
of some strings returned by Maxima. This treatment should be performed
in Scilab just after a call to unix_e. But there is a new requirement:
implementation of regular expressions in Scilab (there are some GPL
packages for that). Matlab introduced probably regular expression for
treating Maple strings, see the functions "regexp, regexpi, regexprep,
regexptranslate" which might be ported to Scilab.
4- Conclusion: I am almost sure that it is not possible to include
this toolbox in Scilab 5 on account of big amount of work which is
still required.
ycollet
> this toolbox in Scilab 5 on account of big amount of work which is
> still required.
Yes, I agree. But it should be nice if the connexion to maxima is
robustified so as to work quietly of the scilab files and improve the
toolbox.
I am not a socket specialist. If someone can clean up the connexion
script, it should be easier to work on the sci files.
YC
Sylvestre Ledru
> I just discover this discussion. As I am the author of this toolbox I
> shall give some explanations.
Thanks for this. It is much appreciated !
> 1- I stopped the development of this toolbox because I expected some
> help to replace the provisional Perl-based server, also, I received
> almost no sign of interest for this work.
I am thinking about creating a page listing all the side project/idea
around Scilab on the Wiki...
I could add this idea in it.
> - Why Perl: second reason. I use Perl also for a preliminary treatment
> of some strings returned by Maxima. This treatment should be performed
> in Scilab just after a call to unix_e. But there is a new requirement:
> implementation of regular expressions in Scilab (there are some GPL
> packages for that). Matlab introduced probably regular expression for
> treating Maple strings, see the functions "regexp, regexpi, regexprep,
> regexptranslate" which might be ported to Scilab.
We have this feature in the trunk of Scilab and will be included by
default in the next version of Scilab.
Sylvestre
Francois Vogel
>> implementation of regular expressions in Scilab (there are some GPL
>> packages for that). Matlab introduced probably regular expression for
>> treating Maple strings, see the functions "regexp, regexpi, regexprep,
>> regexptranslate" which might be ported to Scilab.
> We have this feature in the trunk of Scilab and will be included by
> default in the next version of Scilab.
Regular expressions are included in Scilab as part of Tcl. And Tcl is
in Scilab for a looong time.
Now the op. team adds another regexp engine in Scilab, it's their choice.
F.