> I've found it easier to select everything into another
> table, truncate
> the original table, then insert the rows as:
that takes 50 seconds of pure sorting and 8GB of ram to sort; my method doesn't require more memory than the size of
theheap table, and no sorting, since the index is already sorted. Basically the cluster operation would be:
A) time it takes to do a full scan of the heap
+ B) time it takes to do a full scan of the index
+ C) time it takes to rewrite ordered heap and index
of course C) is no different than any other method I guess.
plus: with the "create as" method indexes, foreign keys etc have to be recreated on the tab (I'm not talking about
timing:it's just that you have to "remember" to re-create whatever was in the old table...). Plus: if a table has a
foreignkey to the table you're clustering, I guess the "create as" method doesn't work (I guess you can't drop a table
thatis the foreign key of another one).