On Tue, Jul 21, 2009 at 1:47 PM, David Kerr<dmk@mr-paradox.net> wrote:
> On Tue, Jul 21, 2009 at 01:13:18PM -0400, Tom Lane wrote:
> - David Kerr <dmk@mr-paradox.net> writes:
> - I tried it on a table with 899991 random values. It took frickin'
> - forever, but seemed to be willing to respond to cancels anywhere
> - along the line. I'm not sure why you're seeing differently.
>
> Hehe, yeah. For me I let it run 10 min and hit ^C so maybe i just
> hit it at a bad place.
>
> - (The reason it takes forever is that numeric is a variable-width
> - type, and access into a varwidth array is O(n), so the sorting
> - step you've got here is O(n^2). It might help to use unnest()
> - instead of this handmade version of it ...)
>
> unnest() is 8.4 only, right?
There is an undocumented version in earlier versions that's quite similar:
information_schema._pg_expandarray (it's an sql function, not C, so
you should be able examine the source in order to roll your own if you
need to).
merlin
