Nathan Bossart <nathandbossart@gmail.com> writes:
> I've recently committed some optimizations for dumping sequences and
> pg_class information (commits 68e9629, bd15b7d, and 2329cad), and I noticed
> that we are also executing a query per function in pg_dump. Commit be85727
> optimized this by preparing the query ahead of time, but I found that we
> can improve performance further by gathering all the relevant data in a
> single query. Here are the results I see for a database with 10k simple
> functions with and without the attached patch:
I'm a bit concerned about this on two grounds:
1. Is it a win for DBs with not so many functions?
2. On the other end of the scale, if you've got a *boatload* of
functions, what does it do to pg_dump's memory requirements?
I'm recalling my days at Salesforce, where they had quite a few
thousand pl/pgsql functions totalling very many megabytes of source
text. (Don't recall precise numbers offhand, and they'd be obsolete
by now even if I did.)
I'm not sure that the results you're showing justify taking any
risk here.
regards, tom lane
