Josh,
* Josh Berkus (josh@agliodbs.com) wrote:
> I've been tinkering with a number of table bloat checks, and it occurred
> to me that the problm is that these are all approximations based on
> overall gross statistics, and as such highly inaccurate.
Greg Smith and I discussed some improvements around this area @Open.
I'd suggest you talk with him about the ideas that he has.
> It seems like would could have ANALYZE, while sampling from the table,
> also check the amount of dead space per page and use that as an estimate
> of the % of dead space overall. We'd still need something else for
> indexes, but this seems like it would be a good start.
>
> No?
Err, I thought that was one of the things like ANALYZE *did* collect
through the 'live tuples' number?
Thanks,
Stephen