Comments on: Generating lots of rows using connect by – safely!

By: Multi-Row DML « Yet Another OCM

Multi-Row DML « Yet Another OCM — Fri, 27 Mar 2009 10:07:30 +0000

[...] Poder has a way of Generating lots of rows using connect by – safely!. Let’s load some data into this [...]

By: rc

rc — Tue, 30 Dec 2008 13:15:21 +0000

One can speed it up by using the with-clause.

with data1 as
(
select rownum r
from dual
connect by rownum <= 1000
),
data2 as
(
select rownum r
from data1 d1
, data1 d2
, data1 d3
)
select count(*)
from data2
/

By: tanelp

tanelp — Wed, 27 Aug 2008 02:41:59 +0000

Hi Peter,

Here it is:

http://tanelpoder.otepad.com/script:ses.sql

By: peter

peter — Tue, 26 Aug 2008 20:34:58 +0000

Hi Tanel,

Can you please post your “ses” script?

thanks
- peter

By: Mark J. Bobak

Mark J. Bobak — Fri, 20 Jun 2008 06:58:02 +0000

Hey Tanel,

Yes, that was my Oracle-L posting.

But you’re right, it doesn’t seem to be in the archive anymore.

Anyhow, cool idea/solution! Thanks for the tip!

-Mark

By: Asif Momen

Asif Momen — Sat, 14 Jun 2008 11:18:00 +0000

Good one !!!

By: Row-Generation without I/O auf maol Jeopardy!

Row-Generation without I/O auf maol Jeopardy! — Mon, 09 Jun 2008 05:31:56 +0000

[...] (Via) [...]

By: tanelp

tanelp — Sun, 08 Jun 2008 12:00:18 +0000

Hash join would probably be less efficient here than a cartesian merge join anyway (haven’t tested, but have reason to believe so as cartesian merge join is just a mulitplication loop, no data comparison is needed at all…)

Also, I suspect this query runs in serial anyway – as far as I know Oracle doesn’t parallelize connect by rowsources (as a self-connect by operation isn’t too suitable for automatic parallelization…)

By: H.Tonguç Yılmaz

H.Tonguç Yılmaz — Sun, 08 Jun 2008 11:29:11 +0000

wordpress cut SQLs in my previous comment, so here are just my results;

case 1 -
Elapsed: 00:00:43.22

case 2 -
Elapsed: 00:01:50.69

By: H.Tonguç Yılmaz

H.Tonguç Yılmaz — Sun, 08 Jun 2008 11:13:43 +0000

Thank you for this tip Tanel :)

SQL> select count(r) from (
2 select rownum r from
(select rownum r from dual connect by rownum <= 1000) a,
3 4 (select rownum r from dual connect by rownum <= 1000) b,
5 (select rownum r from dual connect by rownum <= 1000) c
6 where rownum select count(r) from (
2 (select rownum r from dual connect by rownum <= 100000000) )
3 /

COUNT(R)
———-
100000000

Elapsed: 00:01:50.69

I tried hash join and parallelism but didn’t make an impotant change.