Re: So, is COUNT(*) fast now?

On 10/24/11 12:35 PM, Tom Lane wrote:
> Your point about people trying to create wider indexes to exploit
> index-only scans is an interesting one, but I think it's premature to
> optimize on the basis of hypotheses about what people might do in
> future.

I don't think that this is hypothetical at all. I know *I'll* be doing
it, and we can expect users who are familiar with MySQL and Oracle to do
it as well.

No, it won't be the majority of our users, who are using ORMs and thus
don't really think about indexing at all. But it will be a significant
number of users who are performance-sensitive ... such as most or all of
our data warehousing users.

Mind you, we're pretty much talking exclusively about users whose tables
don't fit in memory ... usually tables which are 10X or more the size of
memory.

One case which is going to be critical to test is the "join" table, i.e.
the table which supports many-to-many joins and consists only of keys
from the respective two other tables.

--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com