On 7 Apr 2013, at 6:56pm, Richard Hipp <[email protected]> wrote: > It uses a btree as a priority queue. If you have LIMIT N, then it > initially starts putting results into the btree until the btree contains N > entries. Then for each additional row, it first adds the value to the > btree, then removes the largest entry in the btree. Hence the btree always > holds the smallest N entries seen so far. Once all is finished, it walks > the btree to output the results. > > So with limit N, it never stores more than N results at one time.
That's neat. Simon. _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
