Thread: tidscan not work ? Pg 8.4.5 + WinXP

Hello.

How to use tid scans? This below not works :-(
Always is used merge join.


DROP TABLE IF EXISTS test1;
CREATE TABLE test1 as select i,hashint4(i)::text from
generate_series(1,10000) as a(i);

DROP TABLE IF EXISTS test2;
CREATE TABLE test2 as select j,j%10000 as i,null::tid as ct from
generate_series(1,1000000) as a(j);

UPDATE test2 SET   ct=test1.ctid
FROM  test1 WHERE test2.i=test1.i;

VACUUM ANALYZE test1;
VACUUM ANALYZE test2;

SET enable_tidscan = true;

SELECT * FROM test1 join test2 on(test1.ctid=test2.ct)

------------------------
Explain analyze
------------------------

"Merge Join  (cost=249703.68..283698.78 rows=1999633 width=28) (actual
time=7567.582..19524.865 rows=999900 loops=1)"
"  Output: test1.i, test1.hashint4, test2.j, test2.i, test2.ct"
"  Merge Cond: (test2.ct = test1.ctid)"
"  ->  Sort  (cost=248955.55..253955.30 rows=1999900 width=14) (actual
time=7513.539..10361.598 rows=999901 loops=1)"
"        Output: test2.j, test2.i, test2.ct"
"        Sort Key: test2.ct"
"        Sort Method:  external sort  Disk: 23456kB"
"        ->  Seq Scan on test2  (cost=0.00..16456.80 rows=1999900
width=14) (actual time=0.551..2234.130 rows=1000000 loops=1)"
"              Output: test2.j, test2.i, test2.ct"
"  ->  Sort  (cost=748.14..773.14 rows=10000 width=20) (actual
time=54.020..2193.688 rows=999901 loops=1)"
"        Output: test1.i, test1.hashint4, test1.ctid"
"        Sort Key: test1.ctid"
"        Sort Method:  quicksort  Memory: 960kB"
"        ->  Seq Scan on test1  (cost=0.00..83.75 rows=10000 width=20)
(actual time=0.030..26.205 rows=10000 loops=1)"
"              Output: test1.i, test1.hashint4, test1.ctid"
"Total runtime: 21635.881 ms"


------------
pasman

pasman pasma*ski<pasman.p@gmail.com> wrote:

> How to use tid scans?

Write a query where they are the fastest way to retrieve the data,
and make sure your PostgreSQL installation is properly configured.

> This below not works :-( Always is used merge join.

> SELECT * FROM test1 join test2 on(test1.ctid=test2.ct)

You're reading through the entirety of two tables matching rows
between them.  What makes you think random access would be faster
than sequential?  If all this data is cached, then maybe random
access could win, but you would need to configure your PostgreSQL to
expect that.

Have you read this page:

http://wiki.postgresql.org/wiki/Tuning_Your_PostgreSQL_Server

-Kevin

"Kevin Grittner" <Kevin.Grittner@wicourts.gov> writes:
> pasman pasma*ski<pasman.p@gmail.com> wrote:
>> This below not works :-( Always is used merge join.

>> SELECT * FROM test1 join test2 on(test1.ctid=test2.ct)

> You're reading through the entirety of two tables matching rows
> between them.  What makes you think random access would be faster
> than sequential?

FWIW, it isn't going to happen anyway, because the TID scan mechanism
doesn't support scanning based on a join condition.  That hasn't gotten
to the top of the to-do list because the use case is almost vanishingly
small.  ctids generally aren't stable enough for it to be useful to
store references to one table's ctids in another table.

            regards, tom lane

>FWIW, it isn't going to happen anyway, because the TID scan mechanism
>doesn't support scanning based on a join condition.  That hasn't gotten
>to the top of the to-do list because the use case is almost vanishingly
>small.  ctids generally aren't stable enough for it to be useful to
>store references to one table's ctids in another table.

Thanks for explanation.


--
------------
pasman