Re: [Texmacs-dev] [bug #24585] Plugins for scipy and matplotlib

[Joris van der Hoeven]

Let us continue this discussion on texmacs-dev.

On Thu, Mar 15, 2012 at 09:30:24AM +0100, Andrea Gamba wrote:
> Thank you for taking interest in this request of features. Please, let me
> further comment on it.
> 
> The feature of inline graphics is extremely useful and desirable, since it
> would create an open-source, notebook-like environment similar to the one
> used in Mathematica. Such an environment allows to keep a well-organized
> record of all the computations that have to be done, with formulae and the
> resulting graphics in the same place.

Adrian: how manageable is it for you to get communications with Python
working in Scheme format? It would be really great if all Python-based plug-ins
(Python, Sage, Scipy, matplotlib, etc.) could be done on a similar model
as what we did for Mathemagix. Unfortunately, I have no idea about how
difficult that would be.

It would probably also be useful to write a small TeXmacs library with
counterparts for the $-prefixed commands in Mathemagix, so as to easily
generate TeXmacs markup as output.

Introspection would also be a useful thing in combination with Python.
In principle, TeXmacs has everything needed in order to get that working.
What is needed are volunteers to do it.

> On the other hand, also "live" graphics in a separate window, that you can
> zoom, adjust, etc., are useful. In the best possible world, one should have
> inline graphics by default, and should be allowed to switch to graphics in
> separate windows that can be manipulated, and to switch back if so desired.

I agree.

> The people at ipython have developed a web-based notebook-like environment,
> I have not tried it yet, but it seems very promising and ideas can probably
> be taken from it.
> 
> The most useful feature of ipython is probably history browsing (with
> up-down arrows, etc.). I feel that this feature may be difficult to
> implement in TeXmacs, but I have no idea of the details. On the other hand,
> I think that it would be highly appreciated.

What do you mean by history browsing? Is it different from moving to
previous inputs in a session? For Mathemagix, I have been playing with
the thought to have a different kind of history browsing:
going back in the list of previously sent commands,
but while staying in the same input on the screen.
Is that what you want?

> My last comment is that TeXmacs has an enormous potential for the
> integration of different computational tools, and that further development
> of a set of uniformly designed plugins really deserves to be pursued.
> 
> The point about uniform design is that one can use TeXmacs to smoothly move
> between different tools, for instance, one can do symbolic computation with
> Axiom and then feed the result to Scipy for numerics. There may be many
> reason to do this kind of operation, for instance, one could simply want to
> check the result of a symbolic computation with different symbolic
> computational tools, such as Axiom, Mathematica, Maple, etc. (it may seem
> weird, but these highly sophisticated softwares are not 100% bug free, and
> in particular cases they have been observed to produce wrong results). Or,
> one may want to feed to a CAS a formula that is inline in the text, and
> then copy the result back to the text. Or, there may be features available
> or more easily accessible in one program, other that are more easily
> accessible in other programs, so one may want to use different programs
> interchangeably.
> 
> The possibility to keep an organized and readable record of computations
> performed with different programs makes TeXmacs absolutely unique, to my
> knowledge there is nothing similar in the world. But this possibility of
> integration is far from being fully exploited today because not all plugins
> are at the same level of efficiency, and their design is not completely
> uniform. I am therefore really happy that there is renewed activity here.

We need volunteers to switch to using native TeXmacs (scheme) format for
all communications. However, this may require hacking the CAS' sources,
implying to need to get this kind of work integrated in the CAS' themselves.

Best, --Joris


> On Thu, Mar 15, 2012 at 8:28 AM, Alvaro Tejero Cantero <address@hidden>wrote:
> > Adrian, could you comment on the perspectives for this. In this way we
> > could make the wish more precise or split it into the
> > not-yet-implemented wishes
> >
> > --
> > address@hidden
> >
> >
> >
> > On Tue, Mar 13, 2012 at 08:13, Andrea Gamba <address@hidden>
> > wrote:
> > > Yes.
> > >
> > > Scipy and Matplotlib are fantastic pieces of open source software that
> > allow
> > > to completely substitute Matlab for most jobs. The best way to use them
> > is
> > > through the ipython interface, that provides features similar to the
> > Matlab
> > > shell, such as history, autocompletion, etc.
> > >
> > > A perfect Matlab-like working environment (including scipy and
> > matplotlib)
> > > is invoked with ipython -pylab.
> > >
> > > http://www.scipy.org/
> > > http://matplotlib.sourceforge.net/
> > > http://ipython.org/
> > >
> > > It would be great to have the possibility to open a Scipy session within
> > > TeXmacs, that would invoke ipython -pylab and set the environment in
> > such a
> > > way that graphics are automatically included in the text, as it happens
> > with
> > > other TeXmacs plugins. For instance, a command like:
> > >
> > > x=linspace(0,10); plot(x,sin(x))
> > >
> > > would automatically produce the inclusion of the sin graphic in the text.
> > >
> > > Andrea
> > >
> > >
> > >
> > > On Mon, Mar 12, 2012 at 11:52 PM, Alvaro Tejero Cantero
> > > <address@hidden> wrote:
> > >>
> > >> Update of bug #24585 (project texmacs):
> > >>
> > >>                  Status:                    None => Need Info
> > >>
> > >>    _______________________________________________________
> > >>
> > >> Follow-up Comment #1:
> > >>
> > >> This is too unspecific since scipy and matplotlib are libraries for
> > python
> > >> and
> > >> can, as such, be used through the Python interface.
> > >>
> > >> Do you mean a facility to see postscript output from matplotlib when in
> > a
> > >> Python session?
> > >>
> > >>    _______________________________________________________
> > >>
> > >> Reply to this item at:
> > >>
> > >>  <http://savannah.gnu.org/bugs/?24585>
> > >>
> > >> _______________________________________________
> > >>  Message sent via/by Savannah
> > >>  http://savannah.gnu.org/
> > >>
> > >
> >