On 02/12/2017 23:17, Jeff Janes wrote:
Right, so there is a cpu costing problem (which could only be fixed by
hacking postgresql and recompiling it), but it is much smaller of a
problem than the IO cost not being accurate due to the high hit rate.
Fixing the CPU costing problem is unlikely to m
On 03/12/2017 01:44, Tom Lane wrote:
I think it'd be a serious error to screw around with your cost settings
on the basis of a single case in which the rowcount estimates are so
far off. It's really those estimates that are the problem AFAICS.
The core issue in this example is that, the way the
On 03/12/2017 03:27, Jeff Janes wrote:
Due to that, when I disable bitmapscans and seqscans, I start getting
slow index scans on the wrong index, i2 rather than i1. I don't know
why he doesn't see that in his example.
When I increase effective_cache_size to 1024MB, I start getting the plan
wit
On Sat, Dec 2, 2017 at 8:04 PM, Justin Pryzby wrote:
> On Sat, Dec 02, 2017 at 05:27:51PM -0800, Jeff Janes wrote:
> > I think the non-extended stats code also has trouble with booleans.
> > pg_stats gives me a correlation of 0.8 or higher for the flag column.
>
> It's not due to the boolean tho
Jeff Janes writes:
> On Sat, Dec 2, 2017 at 8:04 PM, Justin Pryzby wrote:
>> It thinks there's somewhat-high correlation since it gets a list of x
>> and y values (integer positions by logical and physical sort order) and
>> 90% of the x list (logical value) are the same value ('t'), and the
>> C
On Dec 3, 2017 15:31, "Tom Lane" wrote:
Jeff Janes writes:
> On Sat, Dec 2, 2017 at 8:04 PM, Justin Pryzby
wrote:
>> It thinks there's somewhat-high correlation since it gets a list of x
>> and y values (integer positions by logical and physical sort order) and
>> 90% of the x list (logical val
I wrote:
> I tried creating multiple-column statistics using the v10 facility for
> that:
> regression=# create statistics s1 on num, flag from aaa;
> CREATE STATISTICS
> regression=# analyze aaa;
> ANALYZE
> but that changed the estimate not at all, which surprised me because
> dependency statisti
