SQL

December 30, 2010 (Modified January 1, 2011) When I read pages 101-103 of the book “Pro Oracle SQL” a couple of days ago, I was reminded of a couple of things.  This section of the book describes full outer joins, showing the ANSI syntax and Oracle syntax to perform a full outer join.  If you read [...]

December 29, 2010 I started reading the book “Pro Oracle SQL” a couple of days ago, and I am having trouble putting it down.  Other than a couple of minor word substitutions and intended, but not specified, qualifying words, the first 100 pages of the book are fantastic (I suspect that many of these cases [...]

December 26, 2010 Yesterday I started reading another book on the topic of Oracle SQL (for the moment I will keep the title of the book a mystery).  I am not much of a fan of ANSI style syntax – that syntax style is easy to read when there are just two tables involved, but [...]

December 24, 2010 Imagine that a developer approached you with the following SQL statement, and explained that he (or she) noticed that the execution plan showed a Cartesian merge join.  Even adding an ORDERED hint did not affect the execution plan, and the execution performance was the same with the ORDERED hint. SELECT X.TIME_PERIOD EXPOSURE_PERIOD,Y.TIME_PERIOD [...]

December 20, 2010 (Back to the Previous Post in the Series) The previous article in this series included a test case that demonstrated significantly different performance for a SELECT statement and an INSERT INTO statement that used the same SELECT statement as the data source.  This blog article includes the timing results and the execution [...]

December 18, 2010 (Modified December 19, 2010) (Back to the Previous Post in the Series) (Forward to the Next Post in the Series) The previous blog article in this series resulted in several interesting comments with very useful advice.  But was the performance issue just an isolated case, one that only happens in one database, with one Oracle Database release [...]

December 15, 2010  (Forward to the Next Post in the Series) I am impressed with the responses that were received from the most recent set of blog articles.  I recently encountered an interesting problem with an Oracle database that had the OPTIMIZER_FEATURES_ENABLE parameter set to 10.1.0.4 and the OPTIMIZER_MODE set to ALL_ROWS.  A somewhat complex SQL statement [...]

December 1, 2010 Another recent blog article forced me to Stop, Think, … and just about understand (in case you are wondering about the blog title, the definition of invalid). Consider the following table definition: CREATE TABLE T3(   V1 VARCHAR2(10),   D2 DATE,   N3 NUMBER); INSERT INTO T3 VALUES(   CHR(65),   TRUNC(SYSDATE)-65, [...]

November 29, 2010 (Updated November 30, 2010) I recently read a blog article that forced me to Stop, Think, … Understand (which happens to be the subtitle of this blog, so it is worth mentioning).  The blog article discussed an error found in the Oracle Database documentation regarding how to read an execution plan – more [...]

I’m in “nightmare weekend before presenting” mode. I’m up to my eyes at work (and have been for ages, thus the quiet blog) and my recent weekends have been full of normal {and abnormal} life.

As is the way, when up against it and putting together my proofs for wild claims, everything breaks subtly and makes my wild claims look a little, well, wild – even though they are real issues I’ve seen, worked through and fixed in the day job. *sigh*. It does not help when you come across little oddities you have never seen before and end up spending valuable time looking into them.

So here is one. I’m just putting together a very, very simple demo of how the number of rows the CBO expects to see drops off as you move outside the known range. In the below you can see the statement I am using (I keep passing in different days of the month and watching the expected number of rows drop until I hit 1 expected row), but look at how it progress to the last entry…

mdw11> select count(*) from date_test_flat where date_1=to_date('&day-02-2011','DD-MM-YYYY')
  2  /
Enter value for day: 01

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |    16 |   128 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE(' 2011-02-01 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))

mdw11> /
Enter value for day: 15

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |     2 |    16 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE(' 2011-02-15 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))

mdw11> /
Enter value for day: 21

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |     1 |     8 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE(' 2011-02-21 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))

mdw11> /
Enter value for day: 30

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |    99 |   792 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE('30-02-2011','DD-MM-YYYY'))

mdw11>

The expected number of rows drops, becomes and – and has shot up to 99 again (which is the expected number in the known range, as I have 10,000 rows spread over 100 days). My immediate thought is “Wow! Maybe Oracle have put some odd fix in where when you go well out of range it reverts to expecting an average number of rows”. Nope. It is because I asked for the data for 30th February. And I did not get an error.

I think it is because I have set autotrace traceonly explain. This causes the SQL statement not to be executed {if it is just a select, not an insert, update or delete}. It seems the costing section of the CBO is not so good at spotting duff dates, but it then gets the costing wrong.

I’ve spotted that the format of the filter also changes when the date is invalid, I really want to check that out – but I better continue failing to write the presentation!

I know, pretty pointless knowing this but it just amused me. Below is just a quick continuation to show that if the statment is to be executed you get an error and no plan and that utterly duff dates can be passed in.

mdw11> /
Enter value for day: 28

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |     1 |     8 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE(' 2011-02-28 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))

mdw11> SET AUTOTRACE ON
mdw11> /
Enter value for day: 20
any key>

  COUNT(*)
----------
         0

1 row selected.

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |     1 |     8 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE(' 2011-02-20 00:00:00', 'syyyy-mm-dd
              hh24:mi:ss'))

Statistics
----------------------------------------------------------
          1  recursive calls
          0  db block gets
        821  consistent gets
          0  physical reads
          0  redo size
        421  bytes sent via SQL*Net to client
        415  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

mdw11> /
Enter value for day: 30
select count(*) from date_test_flat where date_1=to_date('30-02-2011','DD-MM-YYYY')
                                                         *
ERROR at line 1:
ORA-01839: date not valid for month specified

mdw11> set autotrace traceonly explain
mdw11> /
Enter value for day: 30

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |    99 |   792 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE('30-02-2011','DD-MM-YYYY'))

mdw11> /
Enter value for day: 45

Execution Plan
----------------------------------------------------------
Plan hash value: 247163334

-------------------------------------------------------------------------------------
| Id  | Operation          | Name           | Rows  | Bytes | Cost (%CPU)| Time     |
-------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                |     1 |     8 |   215   (0)| 00:00:04 |
|   1 |  SORT AGGREGATE    |                |     1 |     8 |            |          |
|*  2 |   TABLE ACCESS FULL| DATE_TEST_FLAT |    99 |   792 |   215   (0)| 00:00:04 |
-------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("DATE_1"=TO_DATE('45-02-2011','DD-MM-YYYY'))