Jon Hough wrote:
>  As far as I know J pads the arrays if the size of the conjugacy class is 
>  smaller than the largest sized conjugacy class, i.e. it makes all the 
>  arrays the same size by padding with 0 0 0 0 when necessary.

OK, now I get it.

>  By the logic of group theory, it is inconceivable that there would be
>  two arrays with only some of the same elements, as all I'm doing is
>  partitioning the permutations into sets of "similar" permutations.

Ok.

In general, one prefers avoid padding and fills where that's possible. In
particular, the length of an array (the number of items along the first
dimension) usually carries critical information, and permitting padding
obscures this information.  

For that reason, usually we'll separate arrays of different lengths, often
by creating a boxed array, each box collecting and containing all arrays
with the same length (which you can take to mean "all arrays with the same
logical interpretation").

But that's just one approach, and it has its pros and cons.  If the zero
rows don't interfere with your logic, then you can keep the array unboxed,
which will probably have better performance than a boxed list of conjugacy
classes.

-Dan
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to