count(*) or count(rowid) 1

[hubud]

Have been told that using count(*) is slower and uses more resources than count(rowid).
Having run a number of counts on a large table it appears that the timings are all over the place. This is probably due to other processes running when I do the tests.
What are the backgrounds reasons for this, and are they valid. I am trying to speed up selects as part of procedures and would be interested in any ideas as the whether they're worth changing.

simmo

 

[Turkbear]

Hi,
The AskTom site has what I consider to be the ultimate answer to this often asked question:

http://asktom.oracle.com/pls/ask/f?p=4950:8:::::F4950_P8_DISPLAYID:5374846604258

Enjoy..

 

[carp]

As long as there is an index on the table, I think you will get the same results:

09:01:00 SQL> select count(rowid) from gl_je_headers;

COUNT(ROWID)
------------
51421


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=138 Card=1 Bytes=7)
1 0 SORT (AGGREGATE)
2 1 INDEX (FULL SCAN) OF 'GL_JE_HEADERS_U1' (UNIQUE) (Cost=1
38 Card=51270 Bytes=358890)




09:01:16 SQL> select count(*) from gl_je_headers;

COUNT(*)
----------
51421


Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=138 Card=1)
1 0 SORT (AGGREGATE)
2 1 INDEX (FULL SCAN) OF 'GL_JE_HEADERS_U1' (UNIQUE) (Cost=1
38 Card=51270)

 

[hubud]

cheers turkbear, that is a very useful link

simmo