mirror of
https://github.com/postgres/postgres.git
synced 2026-04-15 22:10:45 -04:00
need be returned. We keep a heap of the current best N tuples and sift-up new tuples into it as we scan the input. For M input tuples this means only about M*log(N) comparisons instead of M*log(M), not to mention a lot less workspace when N is small --- avoiding spill-to-disk for large M is actually the most attractive thing about it. Patch includes planner and executor support for invoking this facility in ORDER BY ... LIMIT queries. Greg Stark, with some editorialization by moi. |
||
|---|---|---|
| .. | ||
| logtape.c | ||
| Makefile | ||
| tuplesort.c | ||
| tuplestore.c | ||