PMA <[email protected]> wrote:
> Yes, if the relatedness within each primitive set gets lost,
> won't a need for 3-to-6 times the number of J primitives
> virtually cancel any one-token-per-command advantage?

No, it won't cancel "any" (by which I mean "all") of the advantage.
But yes, it will lose SOME. It's a tradeoff.

By even CONSIDERING this we're making a different tradeoff -- right
now it's perfectly obvious what to type in order to get each J
primitive. We're talking about building an editor mode in which it's
no longer obvious. Tough choice!

Actually we could build a quick mockup using unicode of a graphics set
that makes a different tradeoff -- simply make the : operators mean
"place an umlaut above the preceding character", and the dot operators
mean "place a single dot above the preceding character." (Well, it's
ALMOST that simple -- obviously there'd have to be special cases for
dotted forms of i and tall characters.) The result would probably look
OK and be very clear on how to type. It might be worth a quick mockup
by someone who knows Unicode (and J) well enough to make a fakeup and
a screenshot. (Unicode can overstrike characters, and a decent or good
Unicode editor will display a good-looking glyph for a lot of
combinations that one might not expect if one didn't know how many
combinations are useful in different languages.)

We'd gain a single-glyph representation without losing the obvious
"how to type". We'd lose the iconicity -- but then, is it REALLY
iconic? Even relatively clear glyphs like grade-down are more
memorable than they are _obvious_.

My guess is that the ideal will be somewhere between the extremes:
(Unfortunately, different people will benefit differently, but that's
unavoidable -- but I also guess that everyone can get 80% of the
maximum benefit from a single choice.)

Extreme1: all exact APL characters.
Extreme2: all ASCII (as now).
Extreme3: all "joined" ASCII.
Extreme4: all newly invented characters (or animations, etc).

I strongly suspect Extreme4 would be unpleasant. I suspect Extreme3
wouldn't be bad. We all know Extreme2 is perfectly fine :-).

But I imagine that Extreme3 might make a nice starting point for some
*slight* adaptations, like using the less-or-equal sign for <:. Or...
Maybe Unicode has an easily available less-than-or-equal with the bar
on the top, or a broken bar on the top, and maybe that would work just
as well. I don't know.

-Wm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to