Tamar,

Now I wonder why Calvin never just built B+Tree indexes on each array column? *g*

>Finally, I tested searching for data in each case. Here, the cursor was a clear winner. For larger data sets, SEEKing in a cursor was an order of magnitude faster than searching in an array with ASCAN(). The cursor’s advantage was even bigger when the search item wasn’t in the data set. For 10,010 items, searching for a non-existent item in an array took 50 times as long as SEEKing the same non-existent item in a cursor.