I was "chatting" to someone off-list and they expressed the opinion that perhaps my insistence on implementing "filter" the way I described (with a boolean verb as one of the args) was influenced by the way Higher Order Procedures are implemented in most other common functional languages - which rely heavily on recursion - and that this may have skewed my view.
(Reading between the lines) it seems that the experts' view here is "don't do it that way" and that, in fact writing "(#~ odd) x" is no big deal, rather than insisting on the "traditional" "odd filter x" on which I was so intent. In fact, it would seem that the natural J way would be not to try shoehorn "filter" into a dyadic verb that way at all, which would account for some of the more experienced members' bemusement at my insistence. Would that be a fair comment? All comments welcome, hey, I _am_ trying to change gear here... On Tue, 2010-08-24 at 08:40 -0400, Raul Miller wrote: > On Tue, Aug 24, 2010 at 8:19 AM, Alex Gian <[email protected]> wrote: > > The way you did it with the fork was instructive to me, R.E., really > > nice to see all the different approaches! > > > > It's still not a "filter" verb on its own, though, as specified (left > > arg - boolean verb, right arg - list to filter) > > If you want the left arg to be a boolean verb, you have two > possibilities: > > 1) use an adverb > 2) require that the left arg be the gerund of the boolean verb. > > For an english example: if the boolean verb were "choose" > the gerund would be "choosing" and we could form sentences > where we talk of choosing, or we could have words which modify > "choose". J's grammar follow similar constraints. > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
