Tom Lane wrote:
Thus the timing issue (at least as exhibited by this script) has nothing
whatever to do with the time to delete a file, but with the time to
create one. Since the part of DROP being timed has probably got no I/O
involved at all (the tuples being touched are almost surely still in
shared buffers), it's unsurprising that it is consistently fast.
In my original profiling, CREATE TEMPORARY TABLE/DROP TABLE wasn't much faster than TRUNCATE TABLE.
When I try it again now, I see that DROP TABLE is consistently fast, while the timings of CREATE
TEMPORARY TABLE vary as much as those of TRUNCATE TABLE. Your observations on the time needed to
open a file confirm that, I think.
In my test databases, autovacuum is off.
Regards,
Adriaan van Os
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
