On 7/12/06, R&S HUI <[EMAIL PROTECTED]> wrote:
The prospects for more efficient computation improves if you say
name=: x u@:{`[`]} name NB. (0)
name=: x u`[`]} name NB. (1)
instead of
name=: x reverse amend name NB. (2)
name=: x |.@:{ amend name NB. (3)
It is easier and more productive for the prospective special code to
handle (0) and (1) than (2) or (3).
i do gather that special code does not like "cover functions". But
currently the flexibility/generality offered by the namings does not
seem to make a significant difference:
ii=:(1 2; 2 0)
xx=:i. 999 999
10 Ts 'ii reverse`[`]} xx'
0.0163818 4.1968e6
10 Ts 'ii ([:|.{)`[`]} xx'
0.0192198 4.19686e6
10 Ts 'ii reverse amend xx'
0.0194623 4.19674e6
10 Ts 'ii swapG xx'
0.0159129 4.19578e6
i had been hoping there might be a more efficient way of doing what i
was doing...
Is there some prospect for special coding for this more generic swap
functionality? i would certainly forgo the perspicuity and generality
if there were such a prospect - though i dislike such hard coding
which is hard to maintain and read. It is a significant time sink in
my app.
~greg
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
