On Wed, 10 Sep 2003 09:50:02 +0200
"Erwan DUROSELLE" <EDuroselle@seafrance.fr> wrote:
> This is much faster (x17 in my simple test) than without the begin/end,
> which will result in 1 transaction per insert.
I have to agree that is is the way to go. One of my colleagues went as
far as writing a Perl script that inserted the transaction code around X
number of operations. (Sorry I don't have that script now.)
What I would like to know is, how can you determine the optimum number
of lines to surround with a transaction? I guess at some point Postgres
will run out of memory or disk space if too many operations are done
within a single transaction.
It would be nice to have a rule of thumb to calculate roughly how many
operations to group together but I guess that this is almost impossible
to calculate.
Also, the bigger the dump the harder it is to edit by hand so you may
need to develop a script to do this at some point.