On 18 October 2012 17:24, Peter Geoghegan <peter@2ndquadrant.com> wrote:
> On 18 October 2012 17:11, Thom Brown <thom@linux.com> wrote:
>> The estimate is down to almost a 10th of what it was before. What's going on?
>
> Even though the index isn't used, the pg_statistic entries that the
> expression index would have made available are. It's as if you
> materialised the expression into a column, analyzed and grouped by
> that.
D'oh, of course! Thanks Peter.
--
Thom