Curve fitting (was Re: Python Plugins)

[Daniel Carrera <dcarrera math toronto edu>]

Hello,

I haven't written on this topic in a while.

I'd really like to see Gnumeric have a generic curve-fitting function
(it's sort of a "personal itch").

However, I don't want to commit to any timelines (like most of you, I have
other things to do).

That said, I have a few questions:


How difficult is it to integrate stuff into Gnumeric?
Can I do most of the work on my own and just come one day and say "here it
is"?

I figure that the final integration step will require us to coordinate.
However, I think that I should be able to do "the math part" on my own.

Are there any "guidelines" for making code that's useful to Gnumeric?

About the GPL:
The algorithms I need are well-known.  Much of the code I need is already
in the public domain  (_Numerical_Recipes_).  Is there any problem using
public-domain code in GPL code?


Cheers,
Daniel.



On Wed, 15 May 2002, Jody Goldberg wrote:

> On Tue, May 14, 2002 at 10:59:56PM -0400, Daniel Carrera wrote:
> > I thought that I'd write a generic least squares fit program.  I know that
> > Gnumeric already has a linear best fit and a few others.  But what if I
> > want to fit a sine function? (this happened to me not long ago).
> >
> > I want to fit to a user-defined function.  Allow the user to define the
> > paramters that will be tunned, and compute a fit with error estimates and
> > all.
> This is not currently possible from a plugin.  Your best bet for now
> is to add this directly to gnumeric in C.  The existence of multiple
> fitting routines will probably help clarify the interface we'd use
> to do it from a plugin.
>
> > I am a mathematician (grad student).  I'm sure I can do the mathematical
> > component without trouble.  However, I don't know how to integrate it with
> > Gnumeric.
> Drop by #gnumeric on irc.gnome.org and we can discuss it.