Ric,
Yes, I agree with your assessment. It is a little
disappointing to me that the tacit solutions are not more
efficient. It would seem as if the amends could be done in
place for each boxed case, there would be major improvement
in the tacit versions. I bet Roger has spent a great amount
of time thinking about ways to enhance "each" processing.
On Mon, 31 Mar 2008, Sherlock, Ric wrote:
+ OK here are some more comprehensive comparisons of the Looping algorithm and
the best of the Jish ones. (I've modified both to handle empty tables.)
+
+ The main parameters of interest are the Length of the boxed list, and the
Size of the tables so:
+
+ ]'Length Size'=: |:>,{500 40 5;300 100 20
+ 500 500 500 40 40 40 5 5 5
+ 300 100 20 300 100 20 300 100 20
+
+ >From the table below it seems that as the Length
+ increases the looping solution is slightly slower than the
+ Jish solution as long as the table Size remains small. If
+ the table Size is medium to large then the superiority of
+ the looping solution actually increases. I imagine this is
+ related to the amount of RAM available and that the Jish
+ solution takes a long time for big tables because it has
+ to use the Page file??
+
+
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
