Oracle SQL
  • LICENSE

Little script for finding tables for which dynamic sampling was used

Posted on October 7, 2014 by Sayan Malakshinov Posted in CBO, oracle, statistics 1,979 Page views 3 Comments

You can always download latest version here: http://github.com/xtender/xt_scripts/blob/master/dynamic_sampling_used_for.sql
Current source code:

col owner         for a30;
col tab_name      for a30;
col top_sql_id    for a13;
col temporary     for a9;
col last_analyzed for a30;
col partitioned   for a11;
col nested        for a6;
col IOT_TYPE      for a15;
with tabs as (
      select 
         to_char(regexp_substr(sql_fulltext,'FROM "([^"]+)"."([^"]+)"',1,1,null,1))  owner
        ,to_char(regexp_substr(sql_fulltext,'FROM "([^"]+)"."([^"]+)"',1,1,null,2))  tab_name
        ,count(*)                                                                    cnt
        ,sum(executions)                                                             execs
        ,round(sum(elapsed_time/1e6),3)                                              elapsed
        ,max(sql_id) keep(dense_rank first order by elapsed_time desc)               top_sql_id
      from v$sqlarea a
      where a.sql_text like 'SELECT /* OPT_DYN_SAMP */%'
      group by
         to_char(regexp_substr(sql_fulltext,'FROM "([^"]+)"."([^"]+)"',1,1,null,1))
        ,to_char(regexp_substr(sql_fulltext,'FROM "([^"]+)"."([^"]+)"',1,1,null,2))
)
select tabs.* 
      ,t.temporary
      ,t.last_analyzed
      ,t.partitioned
      ,t.nested
      ,t.IOT_TYPE
from tabs
    ,dba_tables t
where 
     tabs.owner    = t.owner(+)
 and tabs.tab_name = t.table_name(+)
order by elapsed desc
/
col owner         clear;
col tab_name      clear;
col top_sql_id    clear;
col temporary     clear;
col last_analyzed clear;
col partitioned   clear;
col nested        clear;
col IOT_TYPE      clear;

ps. Or if you want to find queries that used dynamic sampling, you can use query like that:

select s.*
from v$sql s
where 
  s.sql_id in (select p.sql_id 
               from v$sql_plan p
               where p.id=1
                 and p.other_xml like '%dynamic_sampling%'
              )
dynamic sampling dynamic_sampling

PRECOMPUTE_SUBQUERY hint

Posted on August 28, 2014 by Sayan Malakshinov Posted in CBO, hints, undocumented 2,735 Page views 3 Comments

I’ve just found out that we can specify query block for PRECOMPUTE_SUBQUERY: /*+ precompute_subquery(@sel$2) */
So we can use it now with SQL profiles, SPM baselines and patches.

SQL> select/*+ precompute_subquery(@sel$2) */ * from dual where dummy in (select chr(level) from dual connect by level<=100);

D
-
X

SQL> @last

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  c437vsqj7c4jy, child number 0
-------------------------------------
select/*+ precompute_subquery(@sel$2) */ * from dual where dummy in
(select chr(level) from dual connect by level<=100)

Plan hash value: 272002086

---------------------------------------------------------------------------
| Id  | Operation         | Name | E-Rows |E-Bytes| Cost (%CPU)| E-Time   |
---------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |        |       |     2 (100)|          |
|*  1 |  TABLE ACCESS FULL| DUAL |      1 |     2 |     2   (0)| 00:00:01 |
---------------------------------------------------------------------------

Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------

   1 - SEL$1 / DUAL@SEL$1

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

   1 - filter(("DUMMY"='' OR "DUMMY"='' OR "DUMMY"='♥' OR "DUMMY"='♦'
              OR "DUMMY"='♣' OR "DUMMY"='♠' OR "DUMMY"='' OR "DUMMY"=' OR
              "DUMMY"=' ' OR "DUMMY"=' ' OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"=' '
              OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"='►' OR "DUMMY"='◄' OR
              "DUMMY"='' OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"=''
              OR "DUMMY"='' OR "DUMMY"='↑' OR "DUMMY"='↓' OR "DUMMY"='' OR
              "DUMMY"=' OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"='' OR "DUMMY"=''
              OR "DUMMY"=' ' OR "DUMMY"='!' OR "DUMMY"='"' OR "DUMMY"='#' OR
              "DUMMY"='$' OR "DUMMY"='%' OR "DUMMY"='&' OR "DUMMY"='''' OR
              "DUMMY"='(' OR "DUMMY"=')' OR "DUMMY"='*' OR "DUMMY"='+' OR "DUMMY"=','
              OR "DUMMY"='-' OR "DUMMY"='.' OR "DUMMY"='/' OR "DUMMY"='0' OR
              "DUMMY"='1' OR "DUMMY"='2' OR "DUMMY"='3' OR "DUMMY"='4' OR "DUMMY"='5'
              OR "DUMMY"='6' OR "DUMMY"='7' OR "DUMMY"='8' OR "DUMMY"='9' OR
              "DUMMY"=':' OR "DUMMY"=';' OR "DUMMY"='<' OR "DUMMY"='=' OR "DUMMY"='>'
              OR "DUMMY"='?' OR "DUMMY"='@' OR "DUMMY"='A' OR "DUMMY"='B' OR
              "DUMMY"='C' OR "DUMMY"='D' OR "DUMMY"='E' OR "DUMMY"='F' OR "DUMMY"='G'
              OR "DUMMY"='H' OR "DUMMY"='I' OR "DUMMY"='J' OR "DUMMY"='K' OR
              "DUMMY"='L' OR "DUMMY"='M' OR "DUMMY"='N' OR "DUMMY"='O' OR "DUMMY"='P'
              OR "DUMMY"='Q' OR "DUMMY"='R' OR "DUMMY"='S' OR "DUMMY"='T' OR
              "DUMMY"='U' OR "DUMMY"='V' OR "DUMMY"='W' OR "DUMMY"='X' OR "DUMMY"='Y'
              OR "DUMMY"='Z' OR "DUMMY"='[' OR "DUMMY"='\' OR "DUMMY"=']' OR
              "DUMMY"='^' OR "DUMMY"='_' OR "DUMMY"='`' OR "DUMMY"='a' OR "DUMMY"='b'
              OR "DUMMY"='c' OR "DUMMY"='d'))

PS. I’m not sure, but as far as i remember, when I tested it on 10.2, it didn’t work with specifying a query block.
And I have never seen such usage.

cbo oracle undocumented behaviour precompute_subquery query optimization query optimizing undocumented oracle

How to find out session info about session that comes from remote database through dblink

Posted on July 25, 2014 by Sayan Malakshinov Posted in oracle, remote statements, SQL*Plus 2,679 Page views 6 Comments

It is well known thing and you can even find it on MOS, but I have a little more simple script for it, so I want to show little example.

First of all we need to start script on local database:

SQL>                                                                                                                                                                   
SQL> @transactions/global.sql
Enter filters(empty for any)...
Sid           :
Globalid mask :
Remote_db mask:

 INST_ID  SID    SERIAL# USERNAME REMOTE_DB REMOTE_DBID TRANS_ID         DIRECTION   GLOBALID                                           EVENT                      
-------- ---- ---------- -------- --------- ----------- ---------------- ----------- -------------------------------------------------- ---------------------------
       1  275       4469 XTENDER  BAIKAL     1742630060 8.20.7119        FROM REMOTE 4241494B414C2E63616336656437362E382E32302E37313139 SQL*Net message from client
                                                                                                                                                                  

Then we need to copy GLOBALID of interested session and run script on database that shown in column REMOTE_DBID, but with specifieng GLOBALID:

SQL>                                                                                                                                                                                                 
SQL> conn sys/syspass@baikal as sysdba
Connected.

======================================================================
=======  Connected to  SYS@BAIKAL(baikal)(BAIKAL)
=======  SID           203
=======  SERIAL#       38399
=======  SPID          6536
=======  DB_VERSION    11.2.0.4.0
======================================================================

SQL> @transactions/global.sql
Enter filters(empty for any)...
Sid           :
Globalid mask : 4241494B414C2E63616336656437362E382E32302E37313139
Remote_db mask:

INST_ID   SID    SERIAL# USERNAME  REMOTE_DB  REMOTE_DBID TRANS_ID   DIRECTION   GLOBALID                                            STATE                     
------- ----- ---------- --------- ---------- ----------- ---------- ----------- --------------------------------------------------  --------------------------
      1     9      39637 XTENDER   BAIKAL      1742630060 8.20.7119  TO REMOTE   4241494B414C2E63616336656437362E382E32302E37313139  [ORACLE COORDINATED]ACTIVE

It’s quite simple and fast.

dblink global transaction remote sessions

Standalone sqlplus script for plans comparing

Posted on July 25, 2014 by Sayan Malakshinov Posted in oracle 1,933 Page views Leave a comment

I have a couple scripts for plans comparing:

1. https://github.com/xtender/xt_scripts/blob/master/diff_plans.sql
2. http://github.com/xtender/xt_scripts/blob/master/plans/diff_plans_active.sql

But they have dependencies on other scripts, so I decided to create a standalone script for more convenient use without the need to download other scripts and to set up the sql*plus environment.
I’ve tested it already with firefox, so you can try it now: http://github.com/xtender/xt_scripts/blob/master/plans/diff_plans_active_standalone.sql

Some screenshots:
diff_plans.sql:
diff_plans

plans_active.sql:
plans_active

Usage:
1. plans_active:

SQL> @plans_active 0ws7ahf1d78qa 

2. diff_plans:

SQL> @diff_plans 0ws7ahf1d78qa 
 *** Diff plans by sql_id. Version with package XT_PLANS. 
Usage: @plans/diff_plans2 sqlid [+awr] [-v$sql] 

P_AWR           P_VSQL 
--------------- --------------- 
false           true 

Strictly speaking, we can do it sometimes easier: it’s quite simple to compare plans without first column “ID”, so we can simply compare “select .. from v$sql_plan/v$sql_plan_statistics_all/v$sql_plan_monitor” output with any comparing tool.

Bug with xmltable, xmlnamespaces and xquery_string specified using bind variable

Posted on July 24, 2014 by Sayan Malakshinov Posted in bug, oracle 2,367 Page views 3 Comments

Today I was asked about strange problem: xmltable does not return data, if xquery specified by bind variable and xml data has xmlnamespaces:

SQL> var x_path varchar2(100);
SQL> var x_xml  varchar2(4000);
SQL> col x format a100;
SQL> begin
  2      :x_path:='/table/tr/td';
  3      :x_xml :=q'[
  4                  <table xmlns="http://www.w3.org/tr/html4/">
  5                    <tr>
  6                      <td>apples</td>
  7                      <td>bananas</td>
  8                    </tr>
  9                  </table>
 10                  ]';
 11  end;
 12  /

PL/SQL procedure successfully completed.

SQL> select
  2        i, x
  3   from xmltable( xmlnamespaces(default 'http://www.w3.org/tr/html4/'),
  4                  :x_path -- bind variable
  5                  --'/table/tr/td' -- same value as in the variable "X_PATH"
  6                  passing xmltype(:x_xml)
  7                  columns i    for ordinality,
  8                          x    xmltype path '.'
  9                );

no rows selected

But if we comment bind variable and comment out literal x_query ‘/table/tr/td’, query will return data:

SQL> select
  2        i, x
  3   from xmltable( xmlnamespaces(default 'http://www.w3.org/tr/html4/'),
  4                  --:x_path -- bind variable
  5                  '/table/tr/td' -- same value as in the variable "X_PATH"
  6                  passing xmltype(:x_xml)
  7                  columns i    for ordinality,
  8                          x    xmltype path '.'
  9                );

         I X
---------- -------------------------------------------------------------------
         1 <td xmlns="http://www.w3.org/tr/html4/">apples</td>
         2 <td xmlns="http://www.w3.org/tr/html4/">bananas</td>

2 rows selected.

The only workaround I found is the specifying any namespace in the x_query – ‘/*:table/*:tr/*:td’

SQL> exec :x_path:='/*:table/*:tr/*:td'

PL/SQL procedure successfully completed.

SQL> select
  2        i, x
  3   from xmltable( xmlnamespaces(default 'http://www.w3.org/tr/html4/'),
  4                  :x_path -- bind variable
  5                  passing xmltype(:x_xml)
  6                  columns i    for ordinality,
  7                          x    xmltype path '.'
  8                );

         I X
---------- -------------------------------------------------------------------
         1 <td xmlns="http://www.w3.org/tr/html4/">apples</td>
         2 <td xmlns="http://www.w3.org/tr/html4/">bananas</td>

2 rows selected.

It’s quite ugly solution, but I’m not sure whether there is another solution…

bind variable xmlnamespace xmltable xpath xquery

REGEXP_LIKE: strange unspecified value in parameter “modifier”

Posted on July 23, 2014 by Sayan Malakshinov Posted in bug, CBO, curious, oracle 1,741 Page views Leave a comment

Today I noticed strange thing in predicate section of execution plan for simple query with regexp_like, where 3rd parameter “MODIFIER” was not specified:

SQL> select * from dual where regexp_like(dummy,'.');

D
-
X

SQL> select * from table(dbms_xplan.display_cursor);

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------
SQL_ID  97xuqf9cmjsta, child number 0
-------------------------------------
select * from dual where regexp_like(dummy,'.')

Plan hash value: 272002086

--------------------------------------------------------------------------
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |      |       |       |     2 (100)|          |
|*  1 |  TABLE ACCESS FULL| DUAL |     1 |     2 |     2   (0)| 00:00:01 |
--------------------------------------------------------------------------

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

   1 - filter( REGEXP_LIKE ("DUMMY",'.',HEXTORAW('F07FD85CFF0700006A1116
              45010000000000000000000000FC12164501000000000000000000000000000000000000
              0010000000000000001880D85CFF07000002000000000000000000000081000000') ))


20 rows selected.

It is particularly interesting that the values in HEXTORAW() are always different for different first parameters:

SQL> select * from dual where regexp_like(dummy,'x');
...
   1 - filter( REGEXP_LIKE ("DUMMY",'x',HEXTORAW('3895D330FF0700006A1116
              45010000000000000000000000FC12164501000000000000000000000000000000000000
              0011000000000000006895D330FF07000002000000000000000000000081000000') ))
SQL> select * from dual where regexp_like(dummy,'y');
...
   1 - filter( REGEXP_LIKE ("DUMMY",'y',HEXTORAW('00DA3C3FFF0700006A1116
              45010000000000000000000000FC12164501000000000000000000000000000000000000
              00110000000000000030DA3C3FFF07000002000000000000000000000081000000') ))
SQL> select * from dual where regexp_like(dummy||'','x')
...
   1 - filter( REGEXP_LIKE ("DUMMY"||'','x',HEXTORAW('70964F2FFF0700006A
              111645010000000000000000000000FC1216450100000000000000000000000000000000
              0000001100000000000000A0964F2FFF07000002000000000000000000000081000000')
               ))

I don’t know, what does it mean, but it looks like garbage from memory.
When I noticed this, I decided to check how regexp_like will work in function-based indexes:

SQL> create table xtest as
  2    select dummy||level as str
  3    from dual
  4    connect by level<=30;

Table created.

SQL> select * from xtest where case when regexp_like(str,'1') then 1 end = 1;
...
12 rows selected.

SQL> select * from table(dbms_xplan.display_cursor);

PLAN_TABLE_OUTPUT
-----------------------------------------------------------------------------------------
SQL_ID  7ztp0k8c1zn2h, child number 0
-------------------------------------
select * from xtest where case when regexp_like(str,'1') then 1 end = 1

Plan hash value: 4207139086

---------------------------------------------------------------------------
| Id  | Operation         | Name  | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------
|   0 | SELECT STATEMENT  |       |       |       |     3 (100)|          |
|*  1 |  TABLE ACCESS FULL| XTEST |    12 |   264 |     3   (0)| 00:00:01 |
---------------------------------------------------------------------------

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

   1 - filter(CASE  WHEN  REGEXP_LIKE
              ("STR",'1',HEXTORAW('68F9CB32FF0700006A111645010000000000000000000000FC1
              216450100000000000000000000000000000000000000110000000000000098F9CB32FF0
              7000002000000000000000000000081000000') ) THEN 1 END =1)

SQL> create index xtest_fbi on xtest(case when regexp_like(str,'1') then 1 end);

Index created.

SQL> select * from xtest where case when regexp_like(str,'1') then 1 end = 1;
...
12 rows selected.

SQL> select * from table(dbms_xplan.display_cursor);

PLAN_TABLE_OUTPUT
-----------------------------------------------------------------------------------------
SQL_ID  7ztp0k8c1zn2h, child number 0
-------------------------------------
select * from xtest where case when regexp_like(str,'1') then 1 end = 1

Plan hash value: 1479471124

-----------------------------------------------------------------------------------------
| Id  | Operation                   | Name      | Rows  | Bytes | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |           |       |       |     2 (100)|          |
|   1 |  TABLE ACCESS BY INDEX ROWID| XTEST     |    12 |   300 |     2   (0)| 00:00:01 |
|*  2 |   INDEX RANGE SCAN          | XTEST_FBI |    12 |       |     1   (0)| 00:00:01 |
-----------------------------------------------------------------------------------------

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

   2 - access("XTEST"."SYS_NC00002$"=1)

SQL> select column_expression from user_ind_expressions e where e.index_name='XTEST_FBI';

COLUMN_EXPRESSION
-----------------------------------------------------------------------------------------
CASE  WHEN  REGEXP_LIKE ("STR",'1') THEN 1 END

As you can see it works fine, although the predicate from first execution plan differs from the FBI expression.
Then I dumped 10053 trace and noticed that the HEXTORAW(…) function appeared in “Explain Plan Dump” only, so it looks just like plan output bug.

execution plan HEXTORAW regexp_line

Strange moving filter predicates from index to table

Posted on April 24, 2014 by Sayan Malakshinov Posted in CBO, curious, oracle 2,697 Page views 9 Comments

It seems strange to me:
When all needed columns are in the index, filter predicates are expectedly applied to the index

select a,b from xt_test where a=1 and
(:b is null or b = :b)

-----------------------------------------------------------------------------------------
| Id  | Operation        | Name       | Starts | E-Rows | A-Rows |   A-Time   | Buffers |
-----------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT |            |      1 |        |      1 |00:00:00.01 |       2 |
|*  1 |  INDEX RANGE SCAN| PK_XT_TEST |      1 |      1 |      1 |00:00:00.01 |       2 |
-----------------------------------------------------------------------------------------

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

   1 - access("A"=1)
       filter((:B IS NULL OR "B"=:B))

But if I add another column “PAD”, the filter moves to the table filters:

select a,b,pad from xt_test where a=1
and (:b is null or b = :b)

---------------------------------------------------------------------------------------
| Id  | Operation                   | Name       | Starts | E-Rows | A-Rows | Buffers |
---------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |            |      1 |        |      1 |       4 |
|*  1 |  TABLE ACCESS BY INDEX ROWID| XT_TEST    |      1 |      1 |      1 |       4 |
|*  2 |   INDEX RANGE SCAN          | PK_XT_TEST |      1 |     10 |     10 |       2 |
---------------------------------------------------------------------------------------

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

   1 - filter((:B IS NULL OR "B"=:B))
   2 - access("A"=1)

As workaround we can use something like that:

select--+ NO_ELIMINATE_JOIN(t) NO_ELIMINATE_JOIN(t2@sel$2) gather_plan_statistics
   a,b,pad
from xt_test t
where t.rowid in ( select t2.rowid
                   from xt_test t2
                   where a=1
                     and (:b is null or b = :b)
                 );

Plan hash value: 1464320522

--------------------------------------------------------------------------------------
| Id  | Operation                   | Name       | Starts | E-Rows | A-Rows |Buffers |
--------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |            |      1 |        |      1 |      3 |
|   1 |  NESTED LOOPS               |            |      1 |      1 |      1 |      3 |
|*  2 |   INDEX RANGE SCAN          | PK_XT_TEST |      1 |      1 |      1 |      2 |
|   3 |   TABLE ACCESS BY USER ROWID| XT_TEST    |      1 |      1 |      1 |      1 |
--------------------------------------------------------------------------------------

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

   2 - access("A"=1)
       filter((:B IS NULL OR "B"=:B))

Full test case

[sourcecode language=”sql”]
create table xt_test(a,b,pad,constraint pk_xt_test primary key(a,b))
as select
mod(rownum,10) a
,rownum b
,rpad(rownum,10) pad
from dual
connect by level<=100;
call dbms_stats.gather_table_stats(”,’XT_TEST’);
var b number;
exec :b:=1;
select/*+ gather_plan_statistics */
a,b,pad
from xt_test
where a=1 and (:b is null or b = :b);
select * from table(dbms_xplan.display_cursor(”,”,’allstats last’));

select/*+ gather_plan_statistics */
a,b
from xt_test
where a=1 and (:b is null or b = :b);
select * from table(dbms_xplan.display_cursor(”,”,’allstats last’));

select–+ NO_ELIMINATE_JOIN(t) NO_ELIMINATE_JOIN(t2@sel$2) gather_plan_statistics
a,b,pad
from xt_test t
where t.rowid in ( select t2.rowid
from xt_test t2
where a=1
and (:b is null or b = :b)
);
select * from table(dbms_xplan.display_cursor(”,”,’allstats last’));
[/sourcecode]

[collapse]

Update:
I just forgot to mention that there is another workaround – to force concatenation:

SQL> select--+ use_concat(or_predicates(2))
  2     a,b,pad
  3  from xt_test where a=1 and (:b is null or b = :b);

         A          B PAD
---------- ---------- ----------------------------------------
         1          1 1

Plan hash value: 3582916188

----------------------------------------------------------------------------------------------------------------------------
| Id  | Operation                     | Name       | Starts | E-Rows | Cost (%CPU)| A-Rows |   A-Time   | Buffers | Reads  |
----------------------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT              |            |      1 |        |     3 (100)|      1 |00:00:00.01 |       2 |      2 |
|   1 |  CONCATENATION                |            |      1 |        |            |      1 |00:00:00.01 |       2 |      2 |
|   2 |   TABLE ACCESS BY INDEX ROWID | XT_TEST    |      1 |      1 |     1   (0)|      1 |00:00:00.01 |       2 |      2 |
|*  3 |    INDEX UNIQUE SCAN          | PK_XT_TEST |      1 |      1 |     0   (0)|      1 |00:00:00.01 |       1 |      1 |
|*  4 |   FILTER                      |            |      1 |        |            |      0 |00:00:00.01 |       0 |      0 |
|   5 |    TABLE ACCESS BY INDEX ROWID| XT_TEST    |      0 |     10 |     2   (0)|      0 |00:00:00.01 |       0 |      0 |
|*  6 |     INDEX RANGE SCAN          | PK_XT_TEST |      0 |     10 |     1   (0)|      0 |00:00:00.01 |       0 |      0 |
----------------------------------------------------------------------------------------------------------------------------

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

   3 - access("A"=1 AND "B"=:B)
   4 - filter(:B IS NULL)
   6 - access("A"=1)
       filter(LNNVL("B"=:B))
cbo query optimization

To_char, Infinity and NaN

Posted on April 1, 2014 by Sayan Malakshinov Posted in curious, oracle 3,257 Page views Leave a comment

Funny that oracle can easily cast ‘nan’,’inf’,’infinity’,’-inf’,’-infinity’ to corresponding binary_float_infinity,binary_double_nan, but there is no any format models for to_char(binary_float_infinity,format) or to_binary_***(text_expr,format) that can output the same as to_char(binary_float_infinity)/to_binary_float(‘inf’) without format parameter:

If a BINARY_FLOAT or BINARY_DOUBLE value is converted to CHAR or NCHAR, and the input is either infinity or NaN (not a number), then Oracle always returns the pound signs to replace the value.

Little example:

SQL> select to_binary_float('inf') from dual;

TO_BINARY_FLOAT('INF')
----------------------
                   Inf

SQL> select to_binary_float('inf','9999') from dual;
select to_binary_float('inf','9999') from dual
                       *
ERROR at line 1:
ORA-01722: invalid number

SQL> select
  2     to_char(binary_float_infinity)         without_format
  3    ,to_char(binary_float_infinity,'99999') with_format
  4    ,to_char(1e6d,'99999')                  too_large
  5  from dual;

WITHOUT_F WITH_FORMAT        TOO_LARGE
--------- ------------------ ------------------
Inf       ######             ######

SQL> select to_char(0/0f) without_format, to_char(0/0f,'tme') with_format from dual;

WITHOUT_F WITH_FORMAT
--------- --------------------------------------------------------------------------
Nan       ################################################################

ps. it’s just crossposting of my old blog.

Deterministic functions, result_cache and operators

Posted on March 31, 2014 by Sayan Malakshinov Posted in deterministic functions, oracle, result_cache 8,808 Page views Leave a comment

In previous posts about caching mechanism of determinstic functions I wrote that cached results are kept only between fetch calls, but there is one exception from this rule: if all function parameters are literals, cached result will not be flushed every fetch call.
Little example with difference:

SQL> create or replace function f_deterministic(p varchar2)
  2     return varchar2
  3     deterministic
  4  as
  5  begin
  6     dbms_output.put_line(p);
  7     return p;
  8  end;
  9  /
SQL> set arrays 2 feed on;
SQL> set serverout on;
SQL> select
  2     f_deterministic(x) a
  3    ,f_deterministic('literal') b
  4  from (select 'not literal' x
  5        from dual
  6        connect by level<=10
  7       );

A                              B
------------------------------ ------------------------------
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal
not literal                    literal

10 rows selected.

not literal
literal
not literal
not literal
not literal
not literal
not literal

As you can see, ‘literal’ was printed once, but ‘not literal’ was printed 6 times, so it was returned from cache 4 times.

Also i want to show the differences in consistency between:
1. Calling a function with determinstic and result_cache;
2. Calling an operator for function with result_cache;
3. Calling an operator for function with deterministic and result_cache;

In this example I will do updates in autonomouse transactions to emulate updates in another session during query execution:

Tables and procedures with updates

[sourcecode language=”sql”]
drop table t1 purge;
drop table t2 purge;
drop table t3 purge;

create table t1 as select 1 id from dual;
create table t2 as select 1 id from dual;
create table t3 as select 1 id from dual;

create or replace procedure p1_update as
pragma autonomous_transaction;
begin
update t1 set id=id+1;
commit;
end;
/
create or replace procedure p2_update as
pragma autonomous_transaction;
begin
update t2 set id=id+1;
commit;
end;
/
create or replace procedure p3_update as
pragma autonomous_transaction;
begin
update t3 set id=id+1;
commit;
end;
/
[/sourcecode]

[collapse]

Variant 1

[sourcecode language=”sql”]
create or replace function f1(x varchar2) return number result_cache deterministic
as
r number;
begin
select id into r from t1;
p1_update;
return r;
end;
/
[/sourcecode]

[collapse]

Variant 2

[sourcecode language=”sql”]
create or replace function f2(x varchar2) return number result_cache
as
r number;
begin
select id into r from t2;
p2_update;
return r;
end;
/
create or replace operator o2
binding(varchar2)
return number
using f2
/
[/sourcecode]

[collapse]

Variant 3

[sourcecode language=”sql”]
create or replace function f3(x varchar2) return number result_cache deterministic
as
r number;
begin
select id into r from t3;
p3_update;
return r;
end;
/
create or replace operator o3
binding(varchar2)
return number
using f3
/
[/sourcecode]

[collapse]

Test:

SQL> set arrays 2;
SQL> select
  2     f1(dummy) variant1
  3    ,o2(dummy) variant2
  4    ,o3(dummy) variant3
  5  from dual
  6  connect by level<=10;

  VARIANT1   VARIANT2   VARIANT3
---------- ---------- ----------
         1          1          1
         2          1          1
         2          1          1
         3          1          1
         3          1          1
         4          1          1
         4          1          1
         5          1          1
         5          1          1
         6          1          1

10 rows selected.

SQL> /

  VARIANT1   VARIANT2   VARIANT3
---------- ---------- ----------
         7         11         11
         8         11         11
         8         11         11
         9         11         11
         9         11         11
        10         11         11
        10         11         11
        11         11         11
        11         11         11
        12         11         11

10 rows selected.

We can see that function F1 returns same results every 2 execution – it is equal to fetch size(“set arraysize 2”),
operator O2 and O3 return same results for all rows in first query execution, but in the second query executions we can see that they are incremented by 10 – it’s equal to number of rows.
What we can learn from that:
1. The use of the function F1 with result_cache and deterministic reduces function executions, but all function results are inconsistent with query;
2. Operator O2 returns consistent results, but function is always executed because we invalidating result_cache every execution;
3. Operator O3 works as well as operator O2, without considering that function is deterministic.

All tests scripts: tests.zip

consistency deterministic functions pl/sql pl/sql functions result_cache

When v$sesstat statistics are updated

Posted on March 21, 2014 by Sayan Malakshinov Posted in oracle, PL/SQL, runstats, statistics, troubleshooting, undocumented 2,914 Page views Leave a comment

Craig Shallahamer wrote excellent article “When is v$sesstat really updated?”.
And my today post just a little addition and correction about the difference of updating ‘Db time’ and ‘CPU used by this session’ statistics.

Test #1

In this test I want to show that the statistics will be updated after every fetch call.
I have set arraysize=2, so sql*plus will fetch by 2 rows:
(full script)

-- Result will be fetched by 2 rows:
set arraysize 2;
-- this query generates CPU consumption 
-- in the scalar subquery on fetch phase,
-- so CPU consumption will be separated 
-- into several periods between fetch calls:
with gen as (
            select/*+ materialize */
               level n, lpad(level,400) padding
            from dual
            connect by level<=200
            )
    ,stat as (
            select/*+ inline */
               sid,name,value 
            from v$mystat st, v$statname sn
            where st.statistic#=sn.statistic#
              and sn.name in ('DB time'
                             ,'CPU used by this session'
                             ,'user calls'
                             ,'recursive calls')
            )
--first rows just for avoiding SQL*Plus effect with fetching 1 row at start,
-- others will be fetched by "arraysize" rows:
select null rn,null cnt,null dbtime,null cpu,null user_calls, null recursive_calls from dual
union all -- main query:
select
   rownum rn
  ,(select count(*) from gen g1, gen g2, gen g3 where g1.n>g2.n and g1.n*0=main.n*0) cnt
  ,(select value from stat where sid*0=n*0 and name = 'DB time'                    ) dbtime
  ,(select value from stat where sid*0=n*0 and name = 'CPU used by this session'   ) cpu
  ,(select value from stat where sid*0=n*0 and name = 'user calls'                 ) user_calls
  ,(select value from stat where sid*0=n*0 and name = 'recursive calls'            ) recursive_calls
from gen main
where rownum<=10;
set arraysize 15;

Test results:

SQL> @tests/dbtime

        RN        CNT     DBTIME        CPU USER_CALLS RECURSIVE_CALLS
---------- ---------- ---------- ---------- ---------- ---------------

         1    3980000      12021      11989        200             472
         2    3980000      12021      11989        200             472
         3    3980000      12121      12089        201             472
         4    3980000      12121      12089        201             472
         5    3980000      12220      12186        202             472
         6    3980000      12220      12186        202             472
         7    3980000      12317      12283        203             472
         8    3980000      12317      12283        203             472
         9    3980000      12417      12383        204             472
        10    3980000      12417      12383        204             472

As you can see the statistics are updated after every fetch call.

Test #2

Now since we already tested simple sql query, I want to do a little bit more complicated test with PL/SQL:
I’m going to write single PL/SQL block with next algorithm:
1. Saving stats
2. Executing some pl/sql code with CPU consumption
3. Getting statistics difference
4. Starting query from first test
5. Fetch 10 rows
6. Getting statistics difference
7. Fetch next 10 rows
8. Getting statistics difference
9. Fetch next 10 rows
10. Getting statistics difference
And after executing this block, i want to check statistics.

Full script:

set feed off;

-- saving previous values
column st_dbtime      new_value prev_dbtime      noprint;
column st_cpu_time    new_value prev_cputime     noprint;
column st_user_calls  new_value prev_user_calls  noprint;
column st_recur_calls new_value prev_recur_calls noprint;

select   max(decode(sn.NAME,'DB time'                  ,st.value))*10 st_dbtime
        ,max(decode(sn.NAME,'CPU used by this session' ,st.value))*10 st_cpu_time
        ,max(decode(sn.NAME,'user calls'               ,st.value))    st_user_calls
        ,max(decode(sn.NAME,'recursive calls'          ,st.value))    st_recur_calls
from v$mystat st, v$statname sn
where st.statistic#=sn.statistic# 
  and sn.name in ('DB time','CPU used by this session'
                 ,'user calls','recursive calls'
                 )
/
-- variable for output from pl/sql block: 
var output varchar2(4000);

prompt Executing test...;
----- main test:
declare
   cnt int;
   st_dbtime      number; 
   st_cpu_time    number; 
   st_user_calls  number; 
   st_recur_calls number; 
   cursor c is 
      with gen as (select/*+ materialize */
                     level n, lpad(level,400) padding
                   from dual
                   connect by level<=200)
      select
          rownum rn
        , (select count(*) from gen g1, gen g2, gen g3 where g1.n>g2.n and g1.n*0=main.n*0) cnt
      from gen main
      where rownum<=60;
   
   type ctype is table of c%rowtype;
   c_array ctype;
   
   procedure SnapStats(descr varchar2:=null)
   is
      st_new_dbtime      number;
      st_new_cpu_time    number;
      st_new_user_calls  number;
      st_new_recur_calls number;
   begin
      select   max(decode(sn.NAME,'DB time'                 ,st.value))*10 st_dbtime
              ,max(decode(sn.NAME,'CPU used by this session',st.value))*10 st_cpu_time
              ,max(decode(sn.NAME,'user calls'              ,st.value))    st_user_calls
              ,max(decode(sn.NAME,'recursive calls'         ,st.value))    st_recur_calls
          into st_new_dbtime,st_new_cpu_time,st_new_user_calls,st_new_recur_calls
      from v$mystat st, v$statname sn
      where st.statistic#=sn.statistic#
        and sn.name in ('DB time','CPU used by this session'
                       ,'user calls','recursive calls'
                       );
      if descr is not null then
         :output:= :output || descr ||':'||chr(10)
                || 'sesstat dbtime:     ' || (st_new_dbtime      - st_dbtime      )||chr(10)
                || 'sesstat cputime:    ' || (st_new_cpu_time    - st_cpu_time    )||chr(10)
                || 'sesstat user calls: ' || (st_new_user_calls  - st_user_calls  )||chr(10)
                || 'sesstat recur calls:' || (st_new_recur_calls - st_recur_calls )||chr(10)
                || '======================================'||chr(10);
      end if;
      st_dbtime      := st_new_dbtime     ;
      st_cpu_time    := st_new_cpu_time   ;
      st_user_calls  := st_new_user_calls ;
      st_recur_calls := st_new_recur_calls;
   end;
   
begin
   -- saving previous stats:
   SnapStats;

   -- generating cpu load:
   for i in 1..1e7 loop
      cnt:=cnt**2+cnt**1.3-cnt**1.2;
   end loop;
   -- getting new stats:
   SnapStats('After pl/sql loop');
   
   open c;
   SnapStats('After "open c"');
   fetch c bulk collect into c_array limit 10;
   SnapStats('After fetch 10 rows');
   fetch c bulk collect into c_array limit 10;
   SnapStats('After fetch 20 rows');
   fetch c bulk collect into c_array limit 10;
   SnapStats('After fetch 30 rows');
   close c;
   SnapStats('After close c');
end;
/ 

prompt 'Delta stats after statement(ms):';
select   max(decode(sn.NAME,'DB time'                 ,st.value))*10
          - &&prev_dbtime      as delta_dbtime
        ,max(decode(sn.NAME,'CPU used by this session',st.value))*10
          - &&prev_cputime     as delta_cpu_time
        ,max(decode(sn.NAME,'user calls'              ,st.value))  
          - &&prev_user_calls  as delta_user_calls
        ,max(decode(sn.NAME,'recursive calls'         ,st.value))  
          - &&prev_recur_calls as delta_recur_calls
from v$mystat st, v$statname sn
where st.statistic#=sn.statistic# 
  and sn.name in ('DB time','CPU used by this session'
                 ,'user calls','recursive calls'
                 )
/
prompt 'Test results:';
col output format a40;
print output;
set feed off;

Output:

SQL> @tests/dbtime2

Executing test...
'Delta stats after statement(ms):'

DELTA_DBTIME DELTA_CPU_TIME DELTA_USER_CALLS DELTA_RECUR_CALLS
------------ -------------- ---------------- -----------------
       18530          18460                5                33

Test results:
OUTPUT
----------------------------------------
After pl/sql loop:
sesstat dbtime:     0
sesstat cputime:    4350
sesstat user calls: 0
sesstat recur calls:2
======================================
After "open c":
sesstat dbtime:     0
sesstat cputime:    20
sesstat user calls: 0
sesstat recur calls:4
======================================
After fetch 10 rows:
sesstat dbtime:     0
sesstat cputime:    4680
sesstat user calls: 0
sesstat recur calls:2
======================================
After fetch 20 rows:
sesstat dbtime:     0
sesstat cputime:    4680
sesstat user calls: 0
sesstat recur calls:2
======================================
After fetch 30 rows:
sesstat dbtime:     0
sesstat cputime:    4690
sesstat user calls: 0
sesstat recur calls:2
======================================
After close c:
sesstat dbtime:     0
sesstat cputime:    0
sesstat user calls: 0
sesstat recur calls:3
======================================

We can notice that “CPU time” is updated at the same time as recursive calls, but “DB time” is updated only with “User calls”. Although this difference is not so important(because in most cases we can use other statistics in sum), but i think, if you want to instrument some code, it gives reason to check out desirable statistics for update time.

oracle pl/sql v$mystat v$sesstat
photo Sayan Malakshinov

Oracle ACE Pro Oracle ACE Pro Alumni

DEVVYOracle Database Developer Choice Award winner

Oracle performance tuning expert

UK / Cambridge

LinkedIn   Twitter
sayan@orasql.org

Recent Posts

  • Oracle Telegram Bot
  • Partition Pruning and Global Indexes
  • Interval Search: Part 4. Dynamic Range Segmentation – interval quantization
  • Interval Search Series: Simplified, Advanced, and Custom Solutions
  • Interval Search: Part 3. Dynamic Range Segmentation – Custom Domain Index

Popular posts

Recent Comments

  • Oracle SQL | Interval Search: Part 4. Dynamic Range Segmentation – interval quantization on Interval Search: Part 3. Dynamic Range Segmentation – Custom Domain Index
  • Oracle SQL | Interval Search: Part 4. Dynamic Range Segmentation – interval quantization on Interval Search: Part 2. Dynamic Range Segmentation – Simplified
  • Oracle SQL | Interval Search: Part 4. Dynamic Range Segmentation – interval quantization on Interval Search: Optimizing Date Range Queries – Part 1
  • Oracle SQL | Interval Search Series: Simplified, Advanced, and Custom Solutions on Interval Search: Part 2. Dynamic Range Segmentation – Simplified
  • Oracle SQL | Interval Search: Part 2. Dynamic Range Segmentation – Simplified on Interval Search: Part 3. Dynamic Range Segmentation – Custom Domain Index

Blogroll

  • Alex Fatkulin
  • Alexander Anokhin
  • Andrey Nikolaev
  • Charles Hooper
  • Christian Antognini
  • Coskan Gundogar
  • David Fitzjarrell
  • Igor Usoltsev
  • Jonathan Lewis
  • Karl Arao
  • Mark Bobak
  • Martin Bach
  • Martin Berger
  • Neil Chandler
  • Randolf Geist
  • Richard Foote
  • Riyaj Shamsudeen
  • Tanel Poder
  • Timur Akhmadeev
  • Valentin Nikotin
  • Prev
  • 1
  • …
  • 10
  • 11
  • 12
  • 13
  • 14
  • …
  • 17
  • Next
©Sayan Malakshinov. Oracle SQL