From koffice  Sat Nov 13 18:51:09 1999
From: John Zedlewski <zedlwski () Princeton ! EDU>
Date: Sat, 13 Nov 1999 18:51:09 +0000
To: koffice
Subject: Re: Mathematic library / KScript issues and questions
X-MARC-Message: https://marc.info/?l=koffice&m=94252016127713

> Two things you could consider:
> 1. there is goose (http://www.gnu.org/software/goose/goose.html)
> It's a statistics library under the LGPL. It's written in C++
> I remember that somebody wanted to integrate goose with KSpread. Has that
> happened already? If not, goose might be a good starting point for adding new
> functionality.

I looked into using Goose, and it's a great library (not too hard to integrate
either).  I should have time soon to take another crack at it (or at  least give
pointers to someone else), but there's only one problem: kscript (the parsing
engine for KSpread) can't parse a list of lists; everything shows up as a single
sequence.  This makes it pretty impossible to do, say, a linear regression, which
you'd like to have: "LINREG( (A1..A3), (B1..B3) )".  This shows up to
kscript/KSpread as: LINREG(A1,A2,A3,B1,B2,B3).  I got a bit discouraged trying to
find a good workaround for this and basically gave up.
I think this speaks to a real problem with the KSpread architecture that could be
easily fixed, I imagine.  You should be able to load a KScript library without
actually writing it into the KSpread code.  Then one would just create a KScript
lib that encapsulated Goose's functions.
Hmm... who's in charge of KScript?  Torben right?  I'll email him, and if he can
show me how to create and import a kscript lib, I'll go ahead and do it.  As for
the parsing a sequence of sequences, I just don't know anything about lex/yacc.
--JRZ