Lisp as a Base for a Statistical Computing System

Subject: Re: [R-SIG-Finance] R to common lisp translator
The paper by Ross Ihaka and Duncan Temple Lang titled "Back to the
Future: Lisp as a Base for a Statistical Computing System" discusses
building a new foundation for R on top of Common Lisp, but I don't know
if any work is being done in this direction.
Btw, when I read that paper a few years ago, it was interesting, but
it also sounded a lot like its authors largely picked their solution
(Lisp) first, and then thought about how to justify it.  Lisp probably
WOULD be a pretty good base for building a next-generation R-like
tool, but their discussion of possible alternatives seemed quite
cursory, as did their "Backward's compatibility?" section.

  http://lambda-the-ultimate.org/node/3726

If I had both the time (a lot of time!) and a real mandate to do R&D
on something like that, I'd also look into LuaJIT, Erlang, and SciDB,
and think hard about what R problems I really want to solve and what I
could learn about solving them from those other tools.  However,
despite these criticisms, I do hope that Ihaka and Lang proceed
further with their ideas, and look forward to reading more about it.

Of course, this is all pretty much off topic...

For the original poster who sparked this thread:  Trying to speed up
your R code by auto-translating it to Lisp is a crazy, totally
impractical idea.  You are confusing someone's brief dead-end R&D
exploration with something you could actually use for real.  Profile
your R code and figure out what is really slowing it down, then
proceed from there.  There isn't any magic Lisp bullet to solve things
for you.
Andrew Piskorski <atp at piskorski.com>
http://www.piskorski.com/
| Subject: Re: [R-SIG-Finance] R to common lisp translator
| 
| > The paper by Ross Ihaka and Duncan Temple Lang titled "Back to the
| > Future: Lisp as a Base for a Statistical Computing System" discusses
| > building a new foundation for R on top of Common Lisp, but I don't know
| > if any work is being done in this direction.
| 
| Btw, when I read that paper a few years ago, it was interesting, but
| it also sounded a lot like its authors largely picked their solution
| (Lisp) first, and then thought about how to justify it.  Lisp probably
| WOULD be a pretty good base for building a next-generation R-like
| tool, but their discussion of possible alternatives seemed quite
| cursory, as did their "Backward's compatibility?" section.
| 
|   http://lambda-the-ultimate.org/node/3726
| 
| If I had both the time (a lot of time!) and a real mandate to do R&D
| on something like that, I'd also look into LuaJIT, Erlang, and SciDB,
| and think hard about what R problems I really want to solve and what I
| could learn about solving them from those other tools.  However,
| despite these criticisms, I do hope that Ihaka and Lang proceed
| further with their ideas, and look forward to reading more about it.
| 
| Of course, this is all pretty much off topic...
| 
| For the original poster who sparked this thread:  Trying to speed up
| your R code by auto-translating it to Lisp is a crazy, totally
| impractical idea.  You are confusing someone's brief dead-end R&D
| exploration with something you could actually use for real.  Profile
| your R code and figure out what is really slowing it down, then
| proceed from there.  There isn't any magic Lisp bullet to solve things
| for you.

Well put.  I don't often feel the need for an old-fashioned AOL-style 'me
too' but this is one of those cases.

R gives you more analytical fire power than any other tool we know, and it
also gives you well-tested means to accelerate your code.  I gave a few
tutorials on introducing HPC with R that illustrate a some of those means.
Reimplementing in another language nobody on the street uses is not one of
them.

Dirk

| -- 
| Andrew Piskorski <atp at piskorski.com>
| http://www.piskorski.com/
| 
| _______________________________________________
| R-SIG-Finance at r-project.org mailing list
| https://stat.ethz.ch/mailman/listinfo/r-sig-finance
| -- Subscriber-posting only. If you want to post, subscribe first.
| -- Also note that this is not the r-help list where general R questions should go.
Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com
On Tue, Jan 25, 2011 at 11:17:45AM -0500, Dominick Samperi wrote:
Subject: Re: [R-SIG-Finance] R to common lisp translator

The paper by Ross Ihaka and Duncan Temple Lang titled "Back to the
Future: Lisp as a Base for a Statistical Computing System" discusses
building a new foundation for R on top of Common Lisp, but I don't know
if any work is being done in this direction.
Btw, when I read that paper a few years ago, it was interesting, but
it also sounded a lot like its authors largely picked their solution
(Lisp) first, and then thought about how to justify it. ?Lisp probably
WOULD be a pretty good base for building a next-generation R-like
tool, but their discussion of possible alternatives seemed quite
cursory, as did their "Backward's compatibility?" section.
The solution (Lisp/Scheme) was picked a long time ago and
led to what we call R today. At its core R is still Lisp-based,
with some optimizations like replacing lists with vectors. The
Lisp-based foundation is well-hidden by the R language, a kind
of syntactic sugar for Lisp (according to the original creators of R).

What this paper suggests is that statisticians (and Wall Street
quants) could spend more time solving problems if they didn't have
to spend so much time reinventing the wheel by building and
tweaking a home-grown Lisp-like platform.

They were arguing for an engine replacement, not for a new car.

Dominick
?http://lambda-the-ultimate.org/node/3726

If I had both the time (a lot of time!) and a real mandate to do R&D
on something like that, I'd also look into LuaJIT, Erlang, and SciDB,
and think hard about what R problems I really want to solve and what I
could learn about solving them from those other tools. ?However,
despite these criticisms, I do hope that Ihaka and Lang proceed
further with their ideas, and look forward to reading more about it.

Of course, this is all pretty much off topic...

For the original poster who sparked this thread: ?Trying to speed up
your R code by auto-translating it to Lisp is a crazy, totally
impractical idea. ?You are confusing someone's brief dead-end R&D
exploration with something you could actually use for real. ?Profile
your R code and figure out what is really slowing it down, then
proceed from there. ?There isn't any magic Lisp bullet to solve things
for you.

--
Andrew Piskorski <atp at piskorski.com>
http://www.piskorski.com/