Comments on: Insert into DUAL

By: Colin 't Hart

Colin 't Hart — Fri, 22 Feb 2008 11:33:33 +0000

One hopes that Oracle have marked dual as read-only in 11g now that we have the facility to make tables read-only!

By: Alex

Alex — Thu, 28 Sep 2006 13:32:49 +0000

TOAD shows rows inserted in dual without:
create table hr_dual asl select * from dual;
But I couldn’t get in in SQLPlus without this statement, not even with rownum > 0 as shown on asktom

By: Eddie Awad

Eddie Awad — Tue, 07 Mar 2006 05:01:38 +0000

Wont this impact a serious problem in all the process and applications server which uses dual table

Yes. So, my advice to you is to never insert/delete/update DUAL.

By: Arun Kumar

Arun Kumar — Mon, 06 Mar 2006 14:34:21 +0000

i have a interesting question...

say i do

delete from dual;

1 row deleted;

commit;

commited;

Now,

Select * from dual;

No rows returned;

select sysdate from dual;

No rows returned;

Wont this impact a serious problem in all the process and applications server which uses dual table for one or other purpose....

say now if i restart my database... I cant imagine what will happen....

can anybody say if i'm wrong in any of the things here....

i havent tried this as i dont want to spoil my server if at all there would be a problem.

By: gzhang

gzhang — Wed, 09 Nov 2005 16:31:34 +0000

Yes, dual is magic - see asktom [article](http://asktom.oracle.com/pls/ask/f?p=4950:8:::::F4950_P8_DISPLAYID:40476301944675#40756986348091)

By: Eddie Awad

Eddie Awad — Wed, 09 Nov 2005 15:35:51 +0000

Starting with 10g, Oracle does not do a full table scan on DUAL if it does not need to. I have just published a post about it.

So, following up on Sergio’s comment, it looks like in 10g, when I “select * from DUAL”, the “FAST DUAL” takes over and returns what it wants to return, one row.

By: Sergio

Sergio — Wed, 09 Nov 2005 14:56:22 +0000

Eddie,

I believe the difference is that when you use DUAL in a DDL statement, the fast dual optimizer path is not used. For example, I inserted an extra row into DUAL as you did. Then:

SQL> create table bar as select * from dual
  2  ;

Table created.

SQL> select * from bar;

D
-
X
A

Trace output show that the DDL statement resulted in this explain plan:

Rows     Execution Plan
-------  ---------------------------------------------------
      0  CREATE TABLE STATEMENT   MODE: ALL_ROWS
      0   LOAD AS SELECT OF 'BAR'
      0    TABLE ACCESS   MODE: ANALYZED (FULL) OF 'DUAL' (TABLE)

Note that full access of the regular DUAL table was used. "Normal" SELECTs against DUAL access use FAST DUAL these days (I believe since 9i):

Rows     Row Source Operation
-------  ---------------------------------------------------
      1  FAST DUAL  (cr=0 pr=0 pw=0 time=10 us)

Sergio

By: Rob Baillie

Rob Baillie — Wed, 09 Nov 2005 13:35:11 +0000

Aye, great way to at least temporarily cripple a database: add rows to dual, or remove all rows from dual. That and update all the VARCHAR2 columns in the database (including source code) so that:

 the letter l is replaced with the number 1

By: j

Wed, 09 Nov 2005 06:18:21 +0000

Back in 9i days we got strange exceptions with ejbs in 9i. Finally we got figured it out what was the problem: there were two rows in table dual..