From: Jason Riedy
Subject: Re: Back to the Future: Lisp as a Base for a Statistical Computing System
Date:
Message-ID: <87prk62r6t.fsf@sparse.dyndns.org>
And Francogrex writes:
> This below is an abstract of a seemingly very interesting article from
> Ross Ihaka the developer of the very popular and powerful software for
> statistical computing (R). He's going "back to the future" and using
> lisp to improve things.
In many ways, JavaScript is a much closer match to R, Octave, etc.
* Rewriting existing user code ain't gonna happen. Much of the
code is written by reluctant programmers, and companies /
funding agencies don't seem interested in directing them to
rewrite everything. So don't expect any solutions that
require rewriting working, tested code.
* The basic JS type system and conversion system is closer, and
the object system suffices for both R and Octave.
* A non-trivial amount of existing code really *does* associate
variables with values of different types within a loop.
TraceMonkey-style JIT compilation likely will work very well.
Building the same atop CL is possible, but if others already
have done the work...
So I'd expect the first successful reimplementation project to be
based off JS. It's not that CL can't handle these things, but
the JS implementations already seem to be going in a similar
direction as the scientific/statistical computing languages. And
while CL may give a better structure to new code, it won't help
with existing code (or existing texts, existing course work,
etc.).
Jason