Re: Commit within a PL/PGSQL procedure

Tom Lane writes:
> Harry Broomhall <harry.broomhall@uk.easynet.net> writes:
> > In fact my *test* function consists of a simple scan through a database,
> > and 3 INSERTs into a new table for each record in the original, where
> > the values inserted are derived from values in the original.
> >    The source file is about 6.4 ^6 records, and the system dies at about
> > the 1 million mark.
>
> Now that I think twice, do you have foreign key linkages in the
> tables being inserted into?

   Spot on!   The new table has a field that references an ID in the
original table.

>  If so, the most likely explanation is just
> that the list of deferred trigger events is getting too large.  This is
> a known deficiency on our TODO list (it should be possible to shove the
> list out to disk when it gets too large).

  Any way round this?

  Regards,
      Harry.