Re: [S] Splus 3.3 vs 4.0 4.2 4.5 ...

Z. Todd Taylor (
Fri, 22 May 1998 15:43:59 -0700

Tony Plate <> wrote:
> Z. Todd Taylor writes:
> >
> > 1) Arcane is only bad if the oddities are gratuitous. What
> > makes S arcane is what makes it powerful.
> In general, I'd agree, but not with everything.
> I'd argue that the variable scoping laws are
> arcane and subtract power. The way dataframes are stored in
> models as a name (and formulas too, from memory), combined
> with the fact that functions cannot easily access variables in
> intermediate higher frames, makes it very cumbersome to write
> some functions which call modeling functions.


> And I'm not sure if "arcane" is sufficiently negative to
> describe S+'s garbage collection (or lack thereof) and
> the resulting necessity for things like the For() function.

Agreed. This is where I'd like to see some S development focus.

> It's interesting to note that the R language, while based
> on S, differs in these two respects.

Yes. R's scoping is great. Its memory problem is worse,
though, at least for certain kinds of very large datasets.


Z. Todd Taylor
Pacific Northwest National Laboratory
Why don't stranger and anger rhyme?
This message was distributed by  To unsubscribe
send e-mail to with the BODY of the
message:  unsubscribe s-news