Raul said:
> I have not found anything in J's dictionary conflicting with
> this behavior.
Well, the Dictionary explicitly forbids it:
NB. http://www.jsoftware.com/help/dictionary/d310n.htm
2605 174 qdoj ':'
2. The explicit result is the result of the last
non-test block sentence executed; that result
must be a noun in the 3 : and 4 : cases.
NB. http://www.jsoftware.com/help/dictionary/errors.htm
5684 239 qdoj 'errors'
syntax error ) the result of a sentence is not a
noun/verb/adverb/conjunction; a verb
attempting to produce a verb/adverb/
conjunction result
Because a verb returning a verb is currently an error, I think (but I am by no
means positive) that your change is backwards compatible (i.e. nothing else
*should* depend on the result of a verb attempting to return a verb, but since
the behavior in this case is explicitly defined, it *could*).
OTOH, I believe this change is large and subtle, and for those reasons requires
exceptional justification. IMO your example does not suffice:
> That would eliminate the trailing `'''' from explicit verbs
> like
> 3 :'+/,([-.-.)&y`'''''&>
You could achieve the same result now, without the quote characters:
3 : '+/,([-.-.)&y`(i.0)' NB. Or $~0 or ;a: etc.
3 : '+/,{.([-.-.)&y`]'
BTW, I don't quite understand your code's intentions, but did you mean:
3 : '([:+/ , [-.-.)&y`(i.0)'
instead? The way you've written it, it looks like you'll make a gerund of
([-.-.)&, then ravel that, then sum it. Since the ravel of the gerund is a
1-atom list, the sum does nothing.
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
