I wrote:
> There are also ways to use u without explictly inspecting or switching on
> its nameclass, but these are use-case specific, obviously.
>
> If you post your fuller use-case or current adverb, we might be able
> to help you minimize bookkeeping on the argument's nameclass.
I re-read your example:
> '+: y' (1 : 'if. isNoun ''u'' do. u =. (13 : u) end. u ') 3
Here, you'd like to allow the user to specify either a verb directly, or
the body of an explicit definition (a-la '12 * y') which you'll try to
normalize to a tacit verb (using 13 :). In other words, you want to accept
verbs or nouns and normalize all inputs to tacit verbs. That makes sense.
Unfortunately, the tip I posted:
> One common approach is 1 : 'stuff =. u"_ data' which will apply u to 'data'
> if u is
> a verb, or give 'stuff' the same value as u if the user passed in a noun.
> Then you just deal with 'stuff' and forget about u . Pretty nice way to
> let a user provide either a specific value (passing in a noun) or a way to
> produce a desired value (passing in a verb).
is not applicable, because you want to produce a verb instead of a noun,
and your (potential) noun-processor is a conjunction, :, not a verb.
Conjuctival phrases are harder to abstract than verbal phrases in J
(because verbs are the most common kind of word and the focus of the
language). That is, in the phase change from programming to
meta-programming we must shed some weight to rise higher. A few of our
favorite tools, though worthy and dependable, must be left behind.
So, in this case, I think we will have to test u's nameclass (at least, all
the ways I can think of test u's nameclass directly or indirecty [e.g. by
catching syntax errors raised by 13 : ]). But maybe we can make the test
a little more fluid; tied a little tighter to the consequences:
lenient =: 1 : 'u =. u 1 :(''u'' ,~ ''13 : '' #~ noun=nc<''u'')'
+: lenient
+:
'+: y' lenient
+:
;)
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
