"Matthew Dennis" <mdennis@merfer.net> writes:
> My question is when is the "select into _data" query planned/replanned? I'm
> concerned that the query might only be planned once (e.g. the first time
> it's executed or when the function is first defined) and cached
> indefinitely. The table t is initially empty, but grows at a fairly steady
> rate over time. So if the table is essentially empty when the query is
> planned, a seqscan is certainly what the planner would do. However, as time
> goes on and table t is filled with data and the stats change, will the query
> eventually be replanned?
In 8.3, a replan will occur whenever VACUUM/ANALYZE update
pg_class.reltuples, so I think you don't need to worry --- at least not
unless you've disabled autovacuum and chosen a bad manual vacuuming
strategy.
regards, tom lane