Re: Index problem.... GIST (tsearch2)

Greg Stark <gsstark(at)mit(dot)edu> writes:
> 8.0 will have statistics on how often is_null() will return false. But that
> isn't really going to solve your problem since it still won't have any idea
> how many rows the full text search will find.

> I don't even know of anything you can do to influence the selectivity
> estimates of the full text search.

Write some code ;-) ?

Seriously, we desperately need some people thinking about how to do
statistics and selectivity estimates for these sorts of complex
indexable conditions. Even crude estimates would be better than none
at all, which is where we're at now. I think that as of 8.0 there is
sufficient infrastructure in place to collect datatype-specific stats
and do something with them --- but *what* to do is now the pressing
problem.

regards, tom lane