As far as performance goes, currently, folds should probably not be
expected to perform better than explicit code (since the current
implementation uses explicit code behind the curtains).
--
Raul
On Sat, Oct 22, 2022 at 10:49 AM Ed Gottsman
<edward.j.gotts...@gmail.com> wrote:
>
> Jan-Pieter,
>
> Interesting. I’m actually a bit surprised to learn that the “keyword”
> solution performs so well—keywords have apparently received more optimization
> love than I’d have expected.
>
> Only if you’re in the mood for measurement—I’m curious about the performance
> of Raul’s F: approach vis-a-vis Pascal’s approach. I don’t have the fold
> add-on yet so I can’t run the test myself…
>
> Thanks.
>
> Ed
>
> Sent from my iPhone
>
> > On Oct 22, 2022, at 9:33 AM, Jan-Pieter Jacobs <janpieter.jac...@gmail.com>
> > wrote:
> >
> > Don't get me wrong, I'm also a huge tacit fan. However, I learned on the
> > way that it is not always the fastest way, especially when having
> > "irregular" problems.
> >
> > Out of curiosity, I benchmarked the results presented so far:
> >
> > coords=: 1e6 3 ?@$0
> > pj=: }.^:(0=f@{.)^:_
> > jp=: {{for_val. y do. if. f val do. val_index{y return. end. end.}}
> > rm=: 0&(]F..(_2 Z: f@[))
> > +/2.95<+/"1 coords
> > 23
> > f=: 2.95 < +/
> > 100×pacex every 'rm coords';'pj coords';'jp coords'
> > 0.0261933 3136
> > 0.003927 2240
> > 0.00336666 1344
> >
> >
> > It turns out Pascal's solution is doing surprisingly well, likely due to
> > J's fantastic virtual block system; otherwise it'd be duplicating the
> > remainder of the coords at every step.
> >
> > That said, it is difficult to compare numbers without knowing what you'd
> > want to have returned. Raul's version returns no result, Pascale's returns
> > the remainder of the list, and mine the first match...
> >
> > Best regards,
> >
> > Jan-Pieter
> >
> >
> >
> >> On Fri, 21 Oct 2022, 21:53 Ed Gottsman, <edward.j.gotts...@gmail.com>
> >> wrote:
> >>
> >> Jan-Pieter,
> >>
> >> Thank you for the solution. I’m a huge tacit fan (or maybe the word is
> >> “dilettante”) when writing J…less so when reading it a week later :-). In
> >> this particular case the tacit solution wins on elegance and (at least at
> >> the moment) seems perfectly clear.
> >>
> >> Thanks again.
> >>
> >> Ed
> >>
> >> Sent from my iPad
> >>
> >>> On Oct 21, 2022, at 1:55 PM, Jan-Pieter Jacobs <
> >> janpieter.jac...@gmail.com> wrote:
> >>>
> >>> I don't know whether you'r only looking for a tacit function, but this
> >>> would also likely work if fold is unavailable for you (untested):
> >>>
> >>> {{ for_val. y do. if. f val do. val_index{y return. end. end.}} coords
> >>>
> >>> Jan-Pieter
> >>>
> >>>> On Fri, 21 Oct 2022, 19:51 Ed Gottsman, <edward.j.gotts...@gmail.com>
> >> wrote:
> >>>>
> >>>> Pascal,
> >>>>
> >>>> Your original solution is perfect for me—I don’t happen to need the
> >>>> index. And there may in fact be no items with f = 1.
> >>>>
> >>>> Very clever: it gets a place of honor in the flashcard deck :-).
> >>>>
> >>>> Many thanks.
> >>>>
> >>>> Ed
> >>>>
> >>>> Sent from my iPad
> >>>>
> >>>>> On Oct 21, 2022, at 12:46 PM, 'Pascal Jasmin' via Programming <
> >>>> programm...@jsoftware.com> wrote:
> >>>>>
> >>>>> if you want the index rather than the coordinate values, then
> >>>>>
> >>>>> (1 i.~ f) y
> >>>>>
> >>>>> will "short circuit" to the first 1
> >>>>>
> >>>>> and
> >>>>>
> >>>>> ({~ 1 i.~ f)
> >>>>>
> >>>>> may be faster to get first coordinates than my original solution,
> >> though
> >>>> errors if there are no items with 1 = f
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> On Friday, October 21, 2022 at 12:55:41 p.m. EDT, 'Pascal Jasmin' via
> >>>> Programming <programm...@jsoftware.com> wrote:
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> }.^:(0 = f@{.)^:_
> >>>>>
> >>>>> will return the list with the head being the first f not equal 0.
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> On Friday, October 21, 2022 at 12:19:41 p.m. EDT, Ed Gottsman <
> >>>> edward.j.gotts...@gmail.com> wrote:
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> Hi. I’m blanking on something basic and research isn’t helping. When
> >>>> the blindingly obvious solution is revealed I will blame a combination
> >> of
> >>>> sleep deprivation and senescence.
> >>>>>
> >>>>> I’ve got a long list of coordinates ordered by desirability. I want to
> >>>> invoke f (a custom script) with each coordinate in turn. f may return 0
> >>>> (failure), in which case I want to invoke it again with the next
> >>>> coordinate. When f (eventually) returns 1, I want to terminate—the
> >>>> remainder of the coordinates should not be processed. (Note that the
> >> very
> >>>> first invocation of f may return 1.)
> >>>>>
> >>>>> This *almost* feels like ^: in its Do While form. Almost. It also
> >>>> feels like SCs around list operations (though they seem to be limited to
> >>>> primitives).
> >>>>>
> >>>>> Help.
> >>>>>
> >>>>> Many thanks.
> >>>>>
> >>>>> Ed
> >>>>>
> >>>>> Sent from my iPad
> >>>>> ----------------------------------------------------------------------
> >>>>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>>>
> >>>>> ----------------------------------------------------------------------
> >>>>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>>> ----------------------------------------------------------------------
> >>>>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>> ----------------------------------------------------------------------
> >>>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>>
> >>> ----------------------------------------------------------------------
> >>> For information about J forums see http://www.jsoftware.com/forums.htm
> >> ----------------------------------------------------------------------
> >> For information about J forums see http://www.jsoftware.com/forums.htm
> >>
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
