A more general phrase is
   proposition # ]
Whether it merits support by special code is another matter.



----- Original Message -----
From: Dan Bron <[EMAIL PROTECTED]>
Date: Monday, November 5, 2007 7:08
Subject: [Jprogramming] Performance of retain/expunge
To: Programming forum <[email protected]>

> I observe the perfomance difference:
> 
>            
> x0  =:  #~   e.&'ABCDEFGHIJKLMNOPQRSTUVWXYZ'
>            
> x1  =:  -.&(a.-.'ABCDEFGHIJKLMNOPQRSTUVWXYZ')
>            
>            
> M   =:  4e6$a.
>       
>            (x0 
> -: x1) M
>       1
>            
>            
> ts  =:  6!:2 , 7!:2@:]
>            
>            %/ 
> 100 ts&> x0`x1 ,L:0 ' M'
>       3.80106 1.12498
>            
> The difference is even more pronounced with integers:
> 
>            
> w   =:  20?100
>            xi0 
> =:  #~ e.&w
>            xi1 
> =:  -.&((i.100)-.w)
>            
>            
> N   =:  ?1e6$100
>  
>            
> (xi0 -: xi1) N
>       1
> 
>             %/ 100 ts&> xi0`xi1 ,L:0 ' N'
>       7.4249 0.500099
>    
> I think the first form is the clear, obvious way to code this 
> verb.  It's also more general, in the case where we know 
> only the identities of items we want to retain, not the whole 
> universe.
> Could it be supported by special code (e.g. by avoiding the 
> intermediate boolean array)?
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to