[Zope] Zope and R

A.J. Rossini rossini@u.washington.edu
04 Feb 2002 08:39:04 -0800


>>>>> "sb" == seb bacon <seb@jamkit.com> writes:

    sb> On Mon, 2002-02-04 at 15:42, A.J. Rossini wrote:
    >> >>>>> "RB" == Ragnar Beer <rbeer@uni-goettingen.de> writes:
    >> 
    RB> Yep, that's exactly what I mean. It would be very nice if you
    RB> could comment some more.
    >> So, I (and another group, independently) have discussed
    >> integration, with the idea of having Zope be a WWW-front end
    >> for statistical analysis, both report generation as well as
    >> interactive style.
    sb> <snip>
    >> Version 1: treat R as batch.  This is ugly, and works
    >> (i.e. shell-out in CGI-style, constructing image files for the
    >> graphics and use LocalFS to read them back in.

    sb> What's wrong with that approach?  It doesn't have to be ugly.
    sb> You can distribute the work to other computers, the batch will
    sb> run as a separate process which gives you a certain number of
    sb> advantages, and *it works*. I use imagemagick like this and
    sb> I'm very happy with it.  The only issue is writing x-platform
    sb> code can be a bit tricksy.

Nothing is wrong with this approach.  As I stated, it's ugly and
works.  I've used variants on it (using R) since 1996.  Of course,
beauty is in the eye of the beholder.  Since my life isn't based on
writing functional software (though that's an important component),
I've got a different aesthetic scale.

There happen to be more elegant but much less stable approaches.  The
critical point is that R was designed for interactive (statistical)
data analysis, which is what I mostly care about.  The above is fine
for static scripts (not necessarily static results, since databases
can change), however, I'm much more interested more in a
discussion-style flexible UI for statistical analysis than an "embed
results within an object by reading from the updated database".  The
CGI approach is just not terribly useful for building an effective UI
for data analysis scripts, though (by effective, I mean "the average
statistician or student can use it with minimal instructions", not web
programmers who are comfortable with semi-arcane tools).  It doesn't
enhance rapid construction based on general principles.  Been there,
done that, need to move on.

Note that by CGI approach, I'm thinking of the crude "R or other image
processing programs as batch" appoach.  Of course, we can also think
of Zope as a CGI approach unto itself.

best,
-tony

-- 
A.J. Rossini				Rsrch. Asst. Prof. of Biostatistics
U. of Washington Biostatistics		rossini@u.washington.edu	
FHCRC/SCHARP/HIV Vaccine Trials Net	rossini@scharp.org
-------------- http://software.biostat.washington.edu/ ----------------
FHCRC: M: 206-667-7025 (fax=4812)|Voicemail is pretty sketchy/use Email
UW:   Th: 206-543-1044 (fax=3286)|Change last 4 digits of phone to FAX
--- I'm 40% time until March 1st.  Try email the other 3 days.... -----