Re: Fix picksplit with nan values

On Mon, Sep 16, 2013 at 4:13 PM, Andrew Gierth <andrew@tao11.riddles.org.uk> wrote:

>>>>> "Alexander" == Alexander Korotkov <aekorotkov@gmail.com> writes:

 Alexander> 2) NaN coordinates should be processed in GiST index scan
 Alexander> like in sequential scan.

postgres=# select * from pts order by a <-> '(0,0)' limit 10;
    a
----------
 (1,1)
 (7,nan)
 (9,nan)
 (11,nan)
 (4,nan)
 (nan,6)
 (2,1)
 (1,2)
 (2,2)
 (3,1)
(10 rows)

postgres=# set enable_indexscan=false;
SET

postgres=# select * from pts order by a <-> '(0,0)' limit 10;
   a
-------
 (1,1)
 (2,1)
 (1,2)
 (2,2)
 (3,1)
 (1,3)
 (3,2)
 (2,3)
 (4,1)
 (1,4)
(10 rows)

this data set was created by:
insert into pts
  select point(i,j)
    from (select generate_series(1,100)::float8 union all select 'nan') s1(i),
         (select generate_series(1,100)::float8 union all select 'nan') s2(j)
   order by random();

Thanks, Andrew! Good spot.

I didn't examine order by operators for work with NaNs.

I think this time problem is in GiST itself rather than in opclass. I'm going to fix it in a separate patch.