Re: [S] RE: Documentation

Tony Rossini (arossini@biostat.washington.edu)
Wed, 27 May 1998 21:18:07 -0700 (PDT)


On Wed, 27 May 1998, Duncan Murdoch wrote:

> I haven't used S 4, but I use a self-documentation system for Pascal
> code that I wrote a few years ago, and a development environment for
> Object Pascal code that has a more convenient but less extensive one
> built in. I can say from experience that this is a tremendous
> enhancement. I write a lot of code, and having documentation that is
> guaranteed to be up to date with complete coverage (if not complete
> depth) makes the job much easier.
>
> This is the real negative for me in writing S code: the development
> environment is so unhelpful.

I'd be interested in knowning what types of development environments
people think they'd find helpful (2 reasons: personal interest, and
second, to add to or re-prioritize the ESS (Emacs speaks statistics) TODO
list).

Here are directions that I've been leaning towards:
#1 - a more IDE type environment. For ESS, this implies:

- a speedbar mode for S (and the other ESS-supported languages), for
creating an on-the-fly table-of-contents for function definitions
and "marked" variables (via the tags mechanism). Status: needs a
bit more time for integration, but the basic code is there.

- object parsing and display of methods. For S4, this means parsing
the file buiding an object browser. For S3 (and hence, Splus 3.x,
4.x), this means recognizing that not everything is an object.
For XLispStat, this means chasing down prototypes and methods.

#2 -Literate Programming and Data Analysis. This is based on a noweb
interface, which allows for programming documentation and code at
the same time. In particular, the documentation can be sent to
Latex or HTML for later viewing, and the code can be dumped into files
for scripts (or, using ESS under Unix, straight into the running
process).

(In fact, I'll argue that Literate Programming for both programming
and data analysis, using Noweb, WILL save you time in the long run --
although it seems to extend initial coding time by 2x, has saved me
lots on debugging (for programming) and on chasing down
final assumptions as well as initial QA/QC and EDA investigation
decisions)

Anyway, I'd be happy to hear suggestions and thoughts on this (probably
personally, since this is more a "daily statistical practices" issue
rather than Splus specific), and would be happy to summarize and send to
interested parties.

best,
-tony

--
Anthony Rossini
Epimetrics Corporation / Manifold Graphics, Inc.

----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu. To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message: unsubscribe s-news