Horton

Re: path toward faster partition pruning

Amit Langote <langote_amit_f8@lab.ntt.co.jp>

View thread View on postgr.es
From: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>
To: Beena Emerson <memissemerson@gmail.com>, Rajkumar Raghuwanshi <rajkumar.raghuwanshi@enterprisedb.com>
Cc: David Rowley <david.rowley@2ndquadrant.com>, Dilip Kumar <dilipbalaut@gmail.com>, Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2017-10-25T06:47:46Z
Lists: pgsql-hackers

Attachments

Thanks a lot Rajkumar and Beena for the tests.

On 2017/10/24 1:38, Beena Emerson wrote:
> On Mon, Oct 23, 2017 at 3:24 PM, Rajkumar Raghuwanshi wrote:
>> Thanks for updated patches, I am able to compile it on head.
>>
>> While testing this, I got an observation, pruning is not scanning default
>> partition leading to wrong output. below is test to reproduce this.
>>
>> create table rp (a int, b varchar) partition by range (a);
>> create table rp_p1 partition of rp default;
>> create table rp_p2 partition of rp for values from (1) to (10);
>> create table rp_p3 partition of rp for values from (10) to (maxvalue);
>>
>> insert into rp values (-1,'p1');
>> insert into rp values (1,'p2');
>> insert into rp values (11,'p3');
>>
>> postgres=# select tableoid::regclass,* from rp;
>>  tableoid | a  | b
>> ----------+----+----
>>  rp_p2    |  1 | p2
>>  rp_p3    | 11 | p3
>>  rp_p1    | -1 | p1
>> (3 rows)
>>
>> --with pruning
>> postgres=# explain (costs off) select * from rp where a <= 1;
>>         QUERY PLAN
>> --------------------------
>>  Append
>>    ->  Seq Scan on rp_p2
>>          Filter: (a <= 1)
>> (3 rows)
>>
>> postgres=# select * from rp where a <= 1;
>>  a | b
>> ---+----
>>  1 | p2
>> (1 row)

Both this (wrong output)...

> I had noticed this and also that this crash:
> 
> tprt PARTITION BY RANGE(Col1)
>        tprt_1 FOR VALUES FROM (1) TO (50001) PARTITION BY RANGE(Col1)
>               tprt_11 FOR VALUES FROM (1) TO (10000),
>               tprt_1d DEFAULT
>        tprt_2 FOR VALUES FROM (50001) TO (100001)
> 
> EXPLAIN (COSTS OFF) SELECT * FROM tprt WHERE col1 BETWEEN 20000 AND 70000;
> server closed the connection unexpectedly
> This probably means the server terminated abnormally
> before or while processing the request.
> The connection to the server was lost. Attempting reset: Failed.
> !>

...and this (crash) were due to bugs in the 0005 patch.

Output with the updated patch for Rajkumar's test:

explain (costs off ) select * from rp where a <= 1;
        QUERY PLAN
--------------------------
 Append
   ->  Seq Scan on rp_p2
         Filter: (a <= 1)
   ->  Seq Scan on rp_p1
         Filter: (a <= 1)
(5 rows)

select tableoid::regclass, * from rp where a <= 1;
 tableoid | a  | b
----------+----+----
 rp_p2    |  1 | p2
 rp_p1    | -1 | p1
(2 rows)

-- moreover
select tableoid::regclass, * from rp where a < 1;
 tableoid | a  | b
----------+----+----
 rp_d     | -1 | p1
(1 row)


Should be fixed in the attached updated version.  While fixing the bugs, I
made some significant revisions to the code introduced by 0005.

No significant changes to any of the patches 0001-0004.

Thanks,
Amit

Commits

  1. Fix assorted partition pruning bugs

  2. Make gen_partprune_steps static

  3. Remove useless 'default' clause

  4. Reorganize partitioning code

  5. Use custom hash opclass for hash partition pruning

  6. Blindly attempt to fix sepgsql tests broken due to 9fdb675fc5.

  7. Attempt to fix endianess issues in new hash partition test.

  8. Faster partition pruning

  9. For partitionwise join, match on partcollation, not parttypcoll.

  10. Revise API for partition bound search functions.

  11. Revise API for partition_rbound_cmp/partition_rbound_datum_cmp.

  12. Fix possible crash in partition-wise join.

  13. Refactor code for partition bound searching

  14. New C function: bms_add_range

  15. Add extensive tests for partition pruning.

  16. Add null test to partition constraint for default range partitions.

  17. Remove BufFile's isTemp flag.

  18. Make OWNER TO subcommand mention consistent

  19. Fix index matching for operators with mixed collatable/noncollatable inputs.