The future of genius/GEL

[George <jirka@5z.com>]

I spent the last few days thinking about this and tinkering with the
current codebase and some new ideas.  The results of this pondering are as
follows:

  1) The genius engine sucks
  2) while CLN is cool, it sucks inside C code, not to mention it
     is not packaged, etc... thus increasing the curve to install dr.genius
  3) guile is the official launguage of GNU
  4) guile is very embeddable and doesn't really suck too bad.
  5) finishing the current stuff would actually take more and not less time
     as I've realized

Thus I've actually started on writing a GEL->Guile translator.  This will
change some things in GEL actually as it will be a very thin layer over
Guile.  This will mean that you should be able to intermix guile and GEL
functions (as essentially GEL will be just different syntax for guile).

The problems with this

  1) Guile is still missing rationals  (I find this the biggest problem)
  2) We cannot get partial results
  3) Guile only supports double arithmetic, not arbitrary

There are bright sides however:

  1) There is a gmp port of guile in the works apparently which should solve
     #1 and #3
  2) guile does some better optimization then GEL.
  3) guile is maintained (or maintained more then old genius core anyway)
  4) There are quite a few math functions aready written (unfortunately
     matrix stuff will need to be done, but that was rudimentary at best
     in genius anyway)

So what are the plans.

I will try to get this working over christmas.  It doesn't sound like much
work for the translator, and I think the most will be spent on the user
interface since I want to use a similar readline hack like I did for previous
genius engine.

George

-- 
George <jirka@5z.com>
   You have the right to free speech,
   as long as you're not dumb enough to actually try it!
                                         -- The Clash