One more comment re group by: Don't know k or the connection machine (and struggle with take always, and goezinta's) but "group by" maybe helps me understand something. Implicit in a "group by" is that the grouping isn't by the referenced columns, it is by the unique, or distinct values in those columns. So it's neat that you can use ~ to throw the complete "column" at #/. , but it is only the unique values, the nub, that is defining the categories.
Or an I misunderstanding that too? ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm