Re: BUG #13908: Query returns too few rows

On Thu, Feb 4, 2016 at 10:17 PM, David G. Johnston <
david.g.johnston@gmail.com> wrote:

> On Thu, Feb 4, 2016 at 9:47 PM, David G. Johnston <
> david.g.johnston@gmail.com> wrote:
>
>> On this node's detail (the rows seems to appear consistently regardless
>> of node ordering in the plan...)
>> Hash  (cost=3D1440572.59..1440572.59 rows=3D872393 width=3D24) (actual
>> time=3D151981.919..151981.919 rows=3D*1275138* loops=3D1)
>>
>
>
>> Bad Plan: =E2=80=8B
>>
>> =E2=80=8B  =E2=80=8B
>> Buckets: 131072 (originally 131072)  Batches: 32 (originally 16)  Memory
>> Usage: 3073kB
>>
>
> Final result: =E2=80=8B415,967 / =E2=80=8B416,075 (108 missing)
>

=E2=80=8BSo just over 3 output records per bucket and 32 batches =E2=80=8Bg=
ives 96+


>
>> =E2=80=8BBad Plan:  =E2=80=8B  Buckets: 65536 (originally 65536)  Batche=
s: 32 (originally
>> 16)  Memory Usage: 3585kB
>>
>
> =E2=80=8BFinal result: 415,874 / 416,075 (201 missing)=E2=80=8B
>
>
=E2=80=8BJust over 6 output records per bucket and 32 batches yields 192+

Didn't notice any obvious numeric equalities using =E2=80=8B
1,275,138

David J.