hfstcnw=: ([:(# >:@i...@#) 2 , [: ; (0 <@; [: (<@{."0 _)~ >:@i.@>:)
(}:,>:@{:)@;@:{~&.>^:((<@])`(<@2:)) ])
2-:/\(HCF 18);(hc10kE 18);hfstcnw 19
1 1
('HCF';'hc10kE';'hfstcnw') dspl rnkng scores 'HCF 18';'hc10kE
18';'hfstcnw 19'
+-------+----+-----+----+----+
|verb |rank|et*sz|time|size|
+-------+----+-----+----+----+
|HCF | 2 |6.40 |6.40|1.00|
+-------+----+-----+----+----+
|hc10kE | 1 |4.60 |3.68|1.25|
+-------+----+-----+----+----+
|hfstcnw| 0 |1.00 |1.00|1.00|
+-------+----+-----+----+----+
R.E. Boss
> -----Oorspronkelijk bericht-----
> Van: [email protected] [mailto:programming-
> [email protected]] Namens Dan Bron
> Verzonden: zaterdag 13 maart 2010 22:38
> Aan: 'Programming forum'
> Onderwerp: Re: [Jprogramming] Fast(er) Hofstadter-Conway 10K $ sequence
> generation
>
> Aai wrote:
> > e.g. is it possible to avoid 'some' boxing.
>
> I responded:
> > Oh! Sure.
>
> The final form of the solution I posted to RC is:
>
> hc10kE =: 1 1 , exp...@tail
> expand =: 2+I.@;
> tail =: copies&.>^:(<@>:`(<@,@2:))
> copies =: >: |.@(#!.1 |.)~ 1 j. #;.1 #^:_1 ::1:~ ]~:{.
>
> I added it as an "alternative solution", noting that while it is slightly
> more complex, it is exponentially faster. Should I post the "true
> recursive" ($:) formulation as well?
>
> -Dan
>
> PS: I find it appealing that "copies" is a showcase for all the features
> of
> # : tally, copy in its full generality (i.e. with complex arguments), and
> expand, with fit even. The verb is also a good way to understand how
> expand
> differs from complex copy, and when you'd choose one over the other.
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
