"Peter J. Holzer" <hjp-pgsql@hjp.at> writes:
> On 2015-07-31 20:03:41 -0400, Tom Lane wrote:
>> Only the last case produces use of the index. I agree that it'd be better
>> if they all did, but I'm disinclined to consider it a bug fix unless you
>> can show a specific case in which there's a performance regression from
>> older releases.
> "grossly incorrect plan choices are cause for a bug report" (from
> http://www.postgresql.org/docs/9.5/static/indexes-partial.html) ;-)
Sorry, I phrased that poorly. I meant that I'm disinclined to change
this behavior in stable branches unless it's a regression. I agree
that it's reasonable to fix it in 9.5/HEAD, and I've now done so.
regards, tom lane