Raul wrote
> One of the appeals of J is that it supports the use of multiple values.

This is a pertinent criticism of TABULA as it stands.

Hitherto I've been resolutely setting my face against extending TABULA this
way – which is a glaring omission if you come to it fresh from being an
Excel user.

But simply to extend the list of scalar values (vquan_cal_ internally) to a
table will irretrievably break certain features I've been keen to hold
onto. But Raul's remark has broken my longstanding mindset (TABULA is over
15 years old, going back to my 2004 announcement in
http://archive.vector.org.uk/20/4 – it was then written in APL) and I've
suddenly spotted how to do it.

Recently I ported vquan_cal_ from floating point to rational, with
relatively few tears. But it did at least locate the code that breaks if
you change vquan_cal_ to an alien precision… e.g. Boxed. So maybe it
wouldn't be such a big job after all (Famous Last Words…!)

As it stands, TABULA does in fact extend vquan_cal_ to a vector in order to
plot a given "tree" of calculations.
See:
https://code.jsoftware.com/wiki/TABULA/samples/plot_test_using_Function_plotline
But the resulting table (DATA_cal_), used to construct the input data to
the Plot package (graphics/plot), is a passive cache and gets totally
recalculated in a not-awfully-smart way if a single item is changed.

I'm going to explore as a priority the introduction of proper multi-valued
items. Then TABULA will become fit for experiments with image processing,
cellular automata and neural nets. Thank you, Raul.


On Sat, 15 Jun 2019 at 13:15, Raul Miller <[email protected]> wrote:

> On Sat, Jun 15, 2019 at 3:25 AM Ian Clark <[email protected]> wrote:
> > But one of my points of concern was checking primary sources for the
> input
> > figures to the model. If these figures show different values to the ones
> > I've used, e.g. 415 ppm as against my >400, then should I use actual
> > figures, or nice tidy ones that make for an uncluttered display? I
> haven't
> > come to a decision about this. There are arguments both ways.
>
> One of the appeals of J is that it supports the use of multiple values.
>
> Thanks,
>
> --
> Raul
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to