[Jprogramming] a J-ish way to generalize a suffix scan

[R.E. Boss]

If I understand you correctly, you say

	for_p. x do. for_q. y do.

is as (in)efficient as

	for_q. y do. for_p. x do.

Right?


I was thinking of something like doing some smart tricks on y , (x+/y) ,:
x*/y
but probably you are right and there is not much to be gained.
Thanks for pointing out.


R.E. boss


-----Oorspronkelijk bericht-----
Van: programming-bounces at jsoftware.com
[mailto:programming-bounces at jsoftware.com] Namens Raul Miller
Verzonden: dinsdag 5 januari 2010 21:56
Aan: Programming forum
Onderwerp: Re: [Jprogramming] a J-ish way to generalize a suffix scan

On Tue, Jan 5, 2010 at 11:27 AM, R.E. Boss <r.e.boss at planet.nl> wrote:
>   2 3 5 (4 : '(]`(]>.x+[)`(]<.x*[))@.(*@-)/\.y')"0 _  ?.#~10
> 10 10 11  4  8 11 0  9 0 4
> 12 12 12  6 12 12 0 10 0 4
> 10 10 10 10 14 14 0 12 0 4
>
> It does the job but this is a simplified example and actually x and y have
> much more items and the gerund is more complicated as well.
>
> So scanning the complete y for every atom of x does not seem very
efficient.

But that seems to be the essence of what you are doing.  In
the general case for each scan step, you are using a different
member from your  agenda's gerund for each different atom of x.

With a bit of work you could do this in a single scan, but your
overhead is from the many uses of the agenda which will have
to remain unchanged.  If anything pushing everything into a
single scan would make things less efficient since you would
have to iterate over the values of x for each scan element instead
of once per scan.

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