[Jprogramming] Speed of prefix and suffix

[Henry Rich]

See

http://code.jsoftware.com/wiki/Vocabulary/bslash#More_Information

Henry Rich

On 2/1/2016 5:51 AM, Ben Gorte - CITG wrote:
> Good morning,
>
> I stumbled into something I found funny, but perhaps is widely known: Sometimes, u/\. is much much faster than u/\  . In my application it makes the difference between 'perfect' and 'useless'. For example:
>
>     timespacex 'plus /\ i.1000'
> 0.209933 44928
>     timespacex 'plus /\. i.1000'
> 0.000542 19840
>
> Apparently J realizes that u/\. can be done incrementally (going from back to front) for any u, whereas the u/ part of u/\ is done from the beginning over and over. With a vplus that says what it is doing you see:
>
> vplus/ \ i.5
> 0+1
> 1+2
> 0+3
> 2+3
> 1+5
> 0+6
> 3+4
> 2+7
> 1+9
> 0+10
> 0 1 3 6 10
>     vplus/ \. i.5
> 3+4
> 2+7
> 1+9
> 0+10
> 10 10 9 7 4
>
> Indeed, appendix B of the dictionary contains the line
> u/\. 	monad 	special code
> whereas there is no such line for u/\        .
>
> In addition there is special code for many u/\ with built-in u. For user defined (associative, commutative?) u you can use u/\. &. |. instead of u/\ and it will be much faster.
>
> timespacex 'plus /\.&.|. i.1000'
> 0.000571 28928
>
>    vplus /\. &.|.i.5
> 1+0
> 2+1
> 3+3
> 4+6
> 0 1 3 6 10
>
> Ben
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm