To improve the timing of my |.@:|:@:}. ("peeling") solution, I
decided to replace ^:a: ("until there's nothing left") with an
explicit calculation of iterations ("N skins"):
spiral =. 1 : '| $ /:@:;@:u@:i. f.'
bronion0 =. [: ({.&.> #~ 0 ~: */@:$&>) |.@:|:@:}.&.>^:a: @: <
bron0 =: bronion0 spiral
skins =. ((0 ~: *) * (2 * <.) - <: + 1 = [)/@:$&.>
bronion1 =. [: {.&.> |.@|:@:}.&.>^:skins @:<
bron1 =: bronion1 spiral
(bron1 -: bron0) 4 5
1
10 (7!:2@:] , 6!:2) 'bron0 400 500'
3.47534e8 2.01377
10 (7!:2@:] , 6!:2) 'bron1 400 500'
3.47471e8 1.97628
Unfortunately, the speed up is negligible. I have concluded the
reason is since f^:a: y stops when y -: f y , then I was only
saving N invocations of -: (where N =. skins y ) . In this case,
-: is trivially false, because |.@:|:@:}. is a shape-changing
operation, so no atoms need be compared.
Oh well.
-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
