[S] scope rules and references

Jens Oehlschlaegel (oehl@Psyres-Stuttgart.DE)
Tue, 30 Jun 1998 14:04:27 +0200 (MET DST)

I can't resist adding to this discussion, as an experienced,
but not top5%-user:

I think Frank Harrell has demonstrated, what can be reached by
consequently designing an integrated modelling system, by including the
raw data into the model fit. I find it astonishing, how much an individual
has reached with library design, and I wonder why MATHSOFT did not do
similary yet.

In library design, everyone is free to include the raw data or not, so you
know, that at least at the expense of memory things are working.

Except for raw data, everything necessary should be included in the model
fit, at least optional (like y=T, x=T). If John Maindonald is right that
important things are not included in those objects, this should be
corrected rather sooner than later.

The possibility that the raw data be too big to be included regularly
(and some other occasions) clearly show that the daily use of S+
regularly requires breaking the usual scoping and getting things from
special frames/wheres .

S+ currently allows mechanisms for accessings stuff in special
frames/wheres. Those force each user reinventing the wheel (see Siem
Heisterkamp's "component called info" ) and are dangerous, as many
contributions to this list have shown. Some top1%-user might be able to
deal with this, but doesn't 'MathSoft wants to expand its market and
grow', citing [citing Steven P. Millard]?

Some people argue that implementing references into S+ would be too
dangerous. However, I have shown that using the even more dangerous legal
mechanisms of S+, a non-top5%-user like me can implement references in S+
(my library REF) which are *more* secure and easier to use than the
current mechanisms. E.g. assigning stuff to frame 1 *is* dangerous since
naming conflicts may occur. This e.g. would be prevented by aggreeing on
something like class 'static' from library REF. (REF was only developed
for access to memory frames, not to permanent objects)

I have argued that a similar kind of referencing should officially become
part of the S language, optimized for efficiency and documented, otherwise
people developing statistical procedures would not use it (and use more
dangerous mechanisms instead).

I have offered my library to everyone interested including MATHSOFT, but
during the first 4 months after posting, I only received requests from
people outside MATHSOFT, since January 98 someone at MATHSOFT has asked
for and received library REF, but I have never got any feedback from

Given all this, though not an official statement, I appreciate very much

On Mon, 29 Jun 1998, Charles Roosen wrote:

> Conceptually, there are two ways to get the data:
> * include a copy of the data in the model object
> * include a reference to the data in the model object
> [snipped]
> Another way to pass the data would be to add an orig.data argument to
> the various methods, so that the user could explicitely pass the
> original data used in fitting the model to the various methods. If this
> was omitted then the reference could be used. I think this would be my
> FAVORITE approach.

Might it be worth, taking a look at library REF ?
Are there still objections to official references in S+?

Best regards

Jens Oehlschlaegel-Akiyoshi

Jens Oehlschlaegel-Akiyoshi

CURRENT ADDRESS (until 12/98) | PERMANENT VIRTUAL ADDRESS ------------------------------------------------------------------------ | http://www.geocities.com/Tokyo/Bay/5663/ oehl@psyres-stuttgart.de | oehl@geocities.com +49 711 6781-408 (phone) | +49 711 6876902 (fax) |

Center for Psychotherapy Research | Christian-Belser-Strasse 79a | D-70597 Stuttgart Germany |

(general disclaimer)

----------------------------------------------------------------------- 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