I haven't done timings to check, but my recollection is that Roger 
implements the Boyer-Moore algorithm.

Henry Rich

On 1/5/2011 3:04 PM, Dan Bron wrote:
> [email protected]
>>   What is the fastest way to calculate the indices for all exact matches
> from
>>   k that exist in v.
>
> The naive approach in J would be  I.@:E.  (or just  E.  if we wanted a
> boolean mask instead of an index list):
>
>          ss=:I.@:E.
>
>          ('hello j';'software community') ss L:0 'hello j software
> community ...'
>       +-+-+
>       |0|8|
>       +-+-+
>
> Notationally speaking, J could run this code in parallel (the DoJ makes no
> promises about execution order), but currently J is single-threaded and will
> run the searches in series.
>
> Faster approaches are probably possible, such as looking for shared
> sub-sequences among the sought strings first, but I'll leave that for people
> who enjoy code optimization.
>
> -Dan
>
>
>
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to