On Fri, Mar 27, 2009 at 12:44 PM, Robert Dodier <robert.dodier at gmail.com> wrote:
> ...R is really terrific. On a tangential note, Ross Ihaka (one of the two original "R" developers) recently wrote a paper about using Lisp for statistical computing... it was about getting scripts to run faster....
Well, actually the paper ("Back to the Future: Lisp as a Base for a
Statistical Computing System") is about using a Lisp infrastructure
for implementing R so that the R implementors don't have to maintain
their own low-level systems (compiler, garbage collector, etc.). One
of the arguments for it is that there are several good Lisp compilers
-- R currently runs interpretively and gets its speed (which is often
quite impressive) from having its lower level routines operate on bulk
objects (vectors, matrices, ...) rather than individual items.
The irony of all this is that R has essentially displaced LispStat, a
Lisp-based statistical package which apparently had many good
features. (see http://lambda-the-ultimate.org/node/617) LispStat's
implementor, Luke Tierney, is now one of the core R developers.
And though it is often said that R is based on Scheme, as far as I can
tell, the only thing it shares with Scheme is lexical scoping
(besides of course things it shares with every other dynamic
language).
-s