Re: Review: listagg aggregate

Tom Lane <tgl@sss.pgh.pa.us>

From: Tom Lane <tgl@sss.pgh.pa.us>
To: Pavel Stehule <pavel.stehule@gmail.com>
Cc: Robert Haas <robertmhaas@gmail.com>, Takahiro Itagaki <itagaki.takahiro@oss.ntt.co.jp>, "David E. Wheeler" <david@kineticode.com>, "pgsql-hackers@postgresql.org Hackers" <pgsql-hackers@postgresql.org>
Date: 2010-01-28T17:16:59Z
Lists: pgsql-hackers
Pavel Stehule <pavel.stehule@gmail.com> writes:
> in 99.99% the second argument will be a constant. Can we use this
> information and optimize function for this case?

> The detoast on every row can take some percent from a performance.

What detoast?  There won't be one for a constant, nor even for a
variable in any sane situation --- who's going to be using
multi-kilobyte delimiter values?  And if they do, aren't they likely
to run out of memory for the result long before the repeated detoasts
become an interesting problem?  You're arguing about a case that
seems quite irrelevant to the real world.

			regards, tom lane