Re: Parallel Aggregate

On Thu, Dec 10, 2015 at 10:42 PM, Haribabu Kommi
<kommi(dot)haribabu(at)gmail(dot)com> wrote:
>
> Here I attached a POC patch of parallel aggregate based on combine
> aggregate patch. This patch contains the combine aggregate changes
> also. This patch generates and executes the parallel aggregate plan
> as discussed in earlier threads.

I tried this out using PostGIS with no great success. I used a very
simple aggregate for geometry union because the combine function is
just the same as the transfer function for this case (I also mark
ST_Area() as parallel safe, so that the planner will attempt to
parallelize the query)..

CREATE AGGREGATE ST_MemUnion (
basetype = geometry,
sfunc = ST_Union,
cfunc = ST_Union,
stype = geometry
);

Unfortunately attempting a test causes memory corruption and death.

select riding,
st_area(st_memunion(geom))
from vada group by riding;

The explain looks OK:

QUERY PLAN
---------------------------------------------------------------------------------------
Finalize HashAggregate (cost=220629.47..240380.26 rows=79 width=1189)
Group Key: riding
-> Gather (cost=0.00..807.49 rows=8792 width=1189)
Number of Workers: 1
-> Partial HashAggregate (cost=220628.59..220629.38 rows=79
width=1189)
Group Key: riding
-> Parallel Seq Scan on vada (cost=0.00..806.61
rows=8792 width=1189)

But the run dies.

NOTICE: SRID value -32897 converted to the officially unknown SRID value 0
ERROR: Unknown geometry type: 2139062143 - Invalid type

From the message it looks like geometry gets corrupted at some point,
causing a read to fail on very screwed up metadata.

P.