"Michael Paesold" <mpaesold@gmx.at> writes:
> To insert another 10562 rows takes about 12 minutes now!!!
> As I said I wrote a function to insert the rows (PL/pgSQL). All values were
> inserted inside a single function call; I always though that a function call
> would be executed inside a transaction block. Experience says it does.
Well, there's something fishy about your results. Using CVS tip I see
about a 4-to-1 difference between COPYing 10000 rows and INSERT'ing
10000 rows (as one transaction). That's annoyingly high, but it's still
way lower than what you're reporting ...
I used the contents of table tenk1 in the regression database for test
data, and dumped it out with "pg_dump -a" with and without -d. I then
just timed feeding the scripts to psql ...
regards, tom lane