Oracle Sql Query Tuning Hints

  • November 2019
  • PDF

This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA


Overview

Download & View Oracle Sql Query Tuning Hints as PDF for free.

More details

  • Words: 2,747
  • Pages: 12
Oracle SQL Query Tuning Hints WHERE Clause Try to avoid operations on database objects referenced in the WHERE clause. Given Query

Alterna

SELECT ename, hiredate, sal FROM emp WHERE SUBSTR(ename,1,3) = 'SCO';

SELECT ename, hiredate, sal FROM emp WHERE ename LIKE 'SCO%';

VARIABLE name VARCHAR2(20) exec name := 'SCOTT'

VARIABLE name VARCHAR2(20) exec name := 'SCOTT'

SELECT ename, hiredate, sal FROM emp WHERE ename = NVL (:name, ename);

SELECT ename, hiredate, sal FROM emp WHERE ename LIKE NVL (:name

SELECT ename, hiredate, sal FROM emp WHERE TRUNC (hiredate) = TRUNC (SYSDATE);

SELECT ename, hiredate, sal FROM emp WHERE hiredate BETWEEN TRUN AND TRUNC (SYSDATE) + .99

SELECT ename, hiredate, sal FROM emp WHERE ename || empno = 'SCOTT7788';

SELECT ename, hiredate, sal FROM emp WHERE ename = 'SCOTT AND empno = 7788;

SELECT ename, hiredate, sal FROM emp WHERE sal + 3000 < 5000;

SELECT ename, hiredate, sal FROM emp WHERE sal < 2000;

SELECT ename, hiredate, sal FROM emp WHERE sal != 0;

SELECT ename, hiredate, sal FROM emp WHERE sal > 0;

HAVING Clause

The HAVING clause filters selected rows only after all rows have been fetched. Using a WHERE clause h summing, etc. HAVING clauses should only be used when columns with summary operations applied to clause. Given Query SELECT d.dname, AVG (e.sal) FROM emp e, dept d WHERE e.deptno = d.deptno GROUP BY d.dname HAVING dname != 'RESEAECH' AND dname != 'SALES';

Alternati SELECT d.dname, AVG (e.sal) FROM emp e, dept d WHERE e.deptno = d.deptno AND dname != 'RESEAECH' AND dname != 'SALES' GROUP BY d.dname;

Combined Subqueries

Minimize the number of table lookups (subquery blocks) in queries, particularly if your statements includ

multicolumn UPDATEs. Separate Subqueries SELECT ename FROM emp WHERE sal = (SELECT MAX (sal) FROM lookup) AND comm = (SELECT MAX (comm) FROM lookup);

Combined Subqu

SELECT ename FROM emp WHERE (sal,comm) = (SELECT MAX (sal MAX(comm FROM lookup);

EXISTS, NOT IN, Table Joins

Consider the alternatives EXISTS, IN and table joins when doing multiple table joins. None of the depends on your data. SELECT ename FROM emp E WHERE EXISTS (SELECT FROM WHERE AND

'X' dept deptno = E.deptno dname = 'ACCOUNTING');

SELECT ename FROM emp E WHERE deptno IN (SELECT deptno FROM dept WHERE deptno = E.deptno AND dname = 'ACCOUNTING'); SELECT ename FROM dept D, emp E WHERE E.deptno = D.deptno AND D.dname = 'ACCOUNTING';

DISTINCT

Avoid joins that require the DISTINCT qualifier on the SELECT list in queries which are used to de end of a one-to-many relationship. The DISTINCT operator causes Oracle to fetch all rows satisf and filter out duplicate values. EXISTS is a faster alternative, because the Oracle optimizer realiz satisfied once, there is no need to proceed further and the next matching row can be fetched. Given Query SELECT DISTINCT d.deptno, d.dname FROM dept D, emp E WHERE D.deptno = E.deptno;

Alternative

SELECT d.deptno, d.dname FROM dept D WHERE EXISTS (SELECT 'X' FROM emp E WHERE E.deptno = D.d

ION ALL

Consider whether a UNION ALL will suffice in place of a UNION. The UNION clause forces all rows UNION to be sorted and merged and duplicates to be filtered before the first row is returned. A

including duplicates and does not have to perform any sort, merge or filter. If your tables are m duplicate records), or you don't care if duplicates are returned, the UNION ALL is much more effi UNION

UNION A

SELECT acct, balance FROM debit WHERE trandate = '31-DEC-95' UNION SELECT acct, balance FROM credit WHERE trandate = '31-DEC-95';

SELECT acct, balance FROM debit WHERE trandate = '31-DEC-95' UNION ALL SELECT acct, balance FROM credit WHERE trandate = '31-DEC-95';

DECODE

Consider using DECODE to avoid having to scan the same rows repetitively or join the same tabl not necessarily faster as it depends on your data and the complexity of the resulting query. Also, change your code when new values are allowed in the field. SELECT COUNT(*) FROM emp WHERE status = 'Y' AND ename LIKE 'SMITH%'; ---------SELECT COUNT(*) FROM emp WHERE status = 'N' AND ename LIKE 'SMITH%'; SELECT COUNT(DECODE(status, 'Y', 'X', NULL)) Y_count, COUNT(DECODE(status, 'N', 'X', NULL)) N_count FROM emp WHERE ename LIKE 'SMITH%';

Anti Joins An anti-join is used to return rows from a table that that are present in another table. It might be used EMP to return only those rows in DEPT that didn't join to anything in EMP; SELECT * FROM dept WHERE deptno NOT IN (SELECT deptno FROM EMP); SELECT FROM WHERE AND

dept.* dept, emp dept.deptno = emp.deptno (+) emp.ROWID IS NULL;

SELECT * FROM dept WHERE NOT EXISTS (SELECT NULL FROM emp WHERE emp.deptno = dept.deptno);

Outer Joins

Normally, an outer join of table A to table B would return every record in table A, and if it had a mate in as well. Every row in table A would be output, but some rows of table B might not appear in the result s ebery row in table A, as well as every row in table B. The syntax for a full outer join is new in Oracle 9i, it is possible to produce full outer joins sets using conventional SQL. update emp set deptno = 9 where deptno = 10; commit; Conventional SQL

New Syntax

SELECT empno, ename, dept.deptno, dname FROM emp, dept WHERE emp.deptno(+) = dept.deptno UNION ALL SELECT empno, ename, emp.deptno, NULL FROM emp, dept WHERE emp.deptno = dept.deptno(+) AND dept.deptno IS NULL ORDER BY 1,2,3,4;

SELECT empno, ename, NVL(dept.deptno,emp.deptno) FROM emp FULL OUTER JOIN dept ON (emp.deptno = dept.deptno) ORDER BY 1,2,3,4;

EMPNO ---------7369 7499 7521 7566 7654 7698 7782 7788 7839 7844 7876 7900 7902 7934

EMPNO ---------7369 7499 7521 7566 7654 7698 7782 7788 7839 7844 7876 7900 7902 7934

ENAME DEPTNO DNAME ---------- ---------- -------------SMITH 20 RESEARCH ALLEN 30 SALES WARD 30 SALES JONES 20 RESEARCH MARTIN 30 SALES BLAKE 30 SALES CLARK 9 SCOTT 20 RESEARCH KING 9 TURNER 30 SALES ADAMS 20 RESEARCH JAMES 30 SALES FORD 20 RESEARCH MILLER 9 10 ACCOUNTING 40 OPERATIONS

ENAME DEPTNO DN ---------- ---------- -SMITH 20 RE ALLEN 30 SA WARD 30 SA JONES 20 RE MARTIN 30 SA BLAKE 30 SA CLARK 9 SCOTT 20 RE KING 9 TURNER 30 SA ADAMS 20 RE JAMES 30 SA FORD 20 RE MILLER 9 10 AC 40 OP

VIEWS The inline view is a construct in Oracle SQL where you can place a query in the SQL FROM, clause, just

OK, so why use the complicated select in the first place? Why not just create the view? Well, one good gives you another database object to maintain, and adds more complexity to your system. By placing t you have all of the code needed to support the query in one place.

The Power of Inline Views

Overview The inline view is a construct in Oracle SQL where you can place a query in the SQL FROM, clause, just

OK, so why use the complicated select in the first place? Why not just create the view? Well, one good gives you another database object to maintain, and adds more complexity to your system. By placing t you have all of the code needed to support the query in one place. If you have a query as the following ... SELECT FROM WHERE AND

a table id = :id b = (SELECT MAX (b) FROM table WHERE id = :id)

... it can be worth to check if an inline view, instead of the subquery will be faster.

Example 1 (Replace Subquery for MAX) With Subquery CREATE TABLE test (id INT, height INT, acc_date DATE); INSERT INTO test (id, height, acc_date) SELECT MOD(ROWNUM,1000), DBMS_RANDOM.RANDOM, SYSDATE-1000+DBMS_RANDOM.VALUE(0,1000) FROM all_objects; 6357 rows created. COMMIT; CREATE INDEX test_idx on test (id, acc_date, height); Index created. ANALYZE FOR FOR FOR

TABLE test COMPUTE STATISTICS TABLE ALL INDEXES ALL INDEXED COLUMNS;

Table analyzed. alter session set timed_statistics=true; alter session set sql_trace=true; VARIABLE b1 NUMBER exec :b1 := 10

ALTER SESSION SET TIMED_STATISTICS=TRUE; ALTER SESSION SET SQL_TRACE=TRUE; SELECT from WHERE AND

max(height) test id = :b1 acc_date = (SELECT MAX(acc_date) FROM test WHERE id = :b1);

MAX(HEIGHT) ----------1480603530 Elapsed: 00:00:00.12

Execution Plan ---------------------------------------------------------0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2 Card=1 Bytes=17) 1 0 SORT (AGGREGATE) 2 1 INDEX (RANGE SCAN) OF 'TEST_IDX' (NON-UNIQUE) (Cost=2 Card=1 Bytes=17) 3 2 SORT (AGGREGATE) 4 3 FIRST ROW (Cost=2 Card=6 Bytes=60) 5 4 INDEX (RANGE SCAN (MIN/MAX)) OF 'TEST_IDX' (NON-UNIQUE) (Cost=2 Car tkprof gek1_ora_16520.trc gek1_ora_16520.out explain=scott/tiger sort=exeela sys=no call count ------- -----Parse 1 Execute 1 Fetch 2 ------- -----total 4

cpu elapsed disk query current -------- ---------- ---------- ---------- ---------0.00 0.00 0 0 0 0.00 0.00 0 2 0 0.00 0.00 0 2 0 -------- ---------- ---------- ---------- ---------0.00 0.00 0 4 0

With Inline View VARIABLE b1 NUMBER exec :b1 := 10 SELECT height FROM (SELECT FROM WHERE ORDER BY WHERE ROWNUM =

height test id = :b1 id DESC, acc_date DESC, height DESC) 1;

HEIGHT ---------1480603530 Execution Plan ---------------------------------------------------------0 SELECT STATEMENT Optimizer=CHOOSE (Cost=2 Card=1 Bytes=13)

rows ---------0 0 1 ---------1

1 2 3

0 1 2

COUNT (STOPKEY) VIEW (Cost=2 Card=6 Bytes=78) INDEX (RANGE SCAN DESCENDING) OF 'TEST_IDX' (NON-UNIQUE) (Cost=2 Card=6

tkprof gek1_ora_16521.trc gek1_ora_16521.out explain=scott/tiger sort=exeela sys=no call count ------- -----Parse 1 Execute 1 Fetch 2 ------- -----total 4

cpu elapsed disk query current -------- ---------- ---------- ---------- ---------0.03 0.06 2 41 0 0.00 0.00 0 0 0 0.00 0.00 0 2 0 -------- ---------- ---------- ---------- ---------0.03 0.06 2 43 0

rows ---------0 0 1 ---------1

Example 2 (Replace Subquery for MAX) Original Query from a trace session: SELECT switch_time,rat_id FROM tariff WHERE effdate = (SELECT MAX(effdate) FROM tariff WHERE effdate <= TRUNC(:b1) AND weekday = :b2 AND t_id = :b3) AND TO_CHAR(switch_time,'HH24:MI') <= TO_CHAR(:b1,'HH24:MI') AND weekday = :b2 AND t_id = :b3 ORDER BY TO_CHAR(switch_time,'HH24:MI') DESC

With Subquery alter session set timed_statistics=true; select value from v$parameter where name = 'user_dump_dest'; alter session set sql_trace=true; VARIABLE b1 VARCHAR2(19) exec :b1 := '07.04.1999:13:30:31' VARIABLE b2 NUMBER exec :b2 := 2 VARIABLE b3 NUMBER exec :b3 := 317

SELECT switch_time, rat_id FROM tariff WHERE effdate = (SELECT MAX(effdate) FROM tariff WHERE effdate <= TRUNC(TO_DATE(:b1,'DD.MM.YYYY:HH24:MI:SS')) AND weekday = :b2 AND T_ID = :b3) AND TO_CHAR(switch_time,'HH24:MI') <= TO_CHAR(TO_DATE(:b1,'DD.MM.YYYY:HH24:MI:SS'),'

AND weekday = :b2 AND t_id = :b3 ORDER BY TO_CHAR(switch_time,'HH24:MI') DESC; SWITCH_TI RAT_ID --------- ---------01-JAN-98 3 01-JAN-98 1

Execution Plan ---------------------------------------------------------0 SELECT STATEMENT Optimizer=CHOOSE (Cost=4 Card=1 Bytes=21) 1 0 SORT (ORDER BY) (Cost=4 Card=1 Bytes=21) 2 1 FILTER 3 2 TABLE ACCESS (FULL) OF 'TARIFF' (Cost=2 Card=1 Bytes=21) 4 3 SORT (AGGREGATE) 5 4 FILTER 6 5 INDEX (RANGE SCAN) OF 'PK_TARIFF' (UNIQUE) (Cost=2 Card=1 Bytes=1 TKPROF: tkprof xyz.trc xyz.out explain=user/pwd sort=exeela sys=no call count ------- -----Parse 1 Execute 1 Fetch 2 ------- -----total 4

cpu elapsed disk query current -------- ---------- ---------- ---------- ---------0.00 0.00 0 0 0 0.00 0.00 0 0 0 0.01 0.00 0 38 8 -------- ---------- ---------- ---------- ---------0.01 0.00 0 38 8

rows ---------0 0 4 ---------4

Misses in library cache during parse: 0 Optimizer goal: CHOOSE

With Inline View VARIABLE b1 VARCHAR2(19) exec :b1 := '07.04.2005:13:30:31' VARIABLE b2 NUMBER exec :b2 := 2 VARIABLE b3 NUMBER exec :b3 := 317

SELECT switch_time, rat_iD FROM (SELECT switch_time, rat_id FROM tariff WHERE effdate <= TRUNC(TO_DATE(:b1,'DD.MM.YYYY:HH24:MI:SS')) AND weekday = :b2 AND t_id = :b3 ORDER BY effdate DESC) WHERE TO_CHAR(switch_time,'HH24:MI') <= TO_CHAR(TO_DATE(:b1,'DD.MM.YYYY:HH24:MI:SS'),'

SWITCH_TI RAT_ID --------- ---------01-JAN-98 3 01-JAN-98 1 Execution Plan ---------------------------------------------------------0 SELECT STATEMENT Optimizer=CHOOSE (Cost=4 Card=1 Bytes=22) 1 0 VIEW (Cost=4 Card=1 Bytes=22) 2 1 SORT (ORDER BY) (Cost=4 Card=1 Bytes=21) 3 2 FILTER 4 3 TABLE ACCESS (BY INDEX ROWID) OF 'TARIFF' (Cost=2 Card=1 Bytes=21) 5 4 INDEX (RANGE SCAN) OF 'PK_TARIFF' (UNIQUE) (Cost=2 Card=1) TKPROF: tkprof xyz.trc xyz.out explain=user/pwd sort=exeela sys=no call count ------- -----Parse 1 Execute 1 Fetch 2 ------- -----total 4

cpu elapsed disk query current -------- ---------- ---------- ---------- ---------0.00 0.00 0 0 0 0.00 0.00 0 0 0 0.00 0.00 0 19 4 -------- ---------- ---------- ---------- ---------0.00 0.00 0 19 4

rows ---------0 0 4 ---------4

Misses in library cache during parse: 0 Optimizer goal: CHOOSE

Example 3 (cannot have join with CONNECT BY)

Have you ever tried to join to a hierarchical query (a query using CONNECT BY and PRIOR) only to get t ORA-01437: cannot have join with CONNECT BY

One of the limitations of hierarchical queries is that you cannot join to them. However, there are often t them anyway. For instance, if the hierarchy table only has surrogate keys, and you would like to display how you can use "Inline Views" to join tables to a hierarchical query. SELECT level, LPAD(' ',2*level-2)||ename ename, empno, mgr, dept.deptno, dept.dname FROM emp, dept WHERE emp.deptno = dept.deptno CONNECT BY PRIOr empno = mgr START WITH empno = 7839; ORA-01437: cannot have join with CONNECT BY

SELECT E.emplevel, SUBSTR(E.ename,1,15) "ENAME", E.empno, dept.deptno, dept.dname FROM dept, (SELECT level emplevel, LPAD(' ',2*level-2)||ename ename, empno, mgr, dep FROM emp CONNECT BY PRIOR empno = mgr START WITH empno = 7839) E

WHERE E.deptno = dept.deptno / EMPLEVEL ENAME EMPNO DEPTNO DNAME ---------- --------------- ---------- ---------- -------------1 KING 7839 10 ACCOUNTING 2 CLARK 7782 10 ACCOUNTING 3 MILLER 7934 10 ACCOUNTING 2 JONES 7566 20 RESEARCH 3 SCOTT 7788 20 RESEARCH 4 ADAMS 7876 20 RESEARCH 3 FORD 7902 20 RESEARCH 4 SMITH 7369 20 RESEARCH 2 BLAKE 7698 30 SALES 3 ALLEN 7499 30 SALES 3 WARD 7521 30 SALES 3 MARTIN 7654 30 SALES 3 TURNER 7844 30 SALES 3 JAMES 7900 30 SALES

Example 3 (ROWNUM 1 Problem) A rownum restriction starting with 1 works: ROWNUM does not work for ranges that don't start at 1. A ROWNUM restriction starting with 1 works: SELECT ROWNUM,ename from emp WHERE ROWNUM BETWEEN 1 and 3 / ROWNUM ---------1 2 3

ENAME ---------SMITH ALLEN WARD

However, if you try to use a range it will not work. For example: SELECT ROWNUM,ename from emp WHERE ROWNUM BETWEEN 2 and 3 / no rows selected Using an Inline View to get around this limitation: SELECT t1.rn, t1.ename FROM (SELECT ROWNUM rn, ename FROM emp) t1 WHERE t1.rn BETWEEN 2 and 3 / The main trick to this query is the "internal" select statement. This select statement in the from clause,

table, then returns the values (along with the psuedo-column ROWNUM) to the "outside" query. The ou the results of the internal query. In order to access the internal query's columns from the external quer query an alias ("t1" highlighted below): This allows you to refer to the columns using the "t1" (highlight psuedo-column and therefore a reserved word, you need to alias that column in the internal query in or query:

Example 4 (ROWNUM and ORDER BY Problem, TOP-N Queries) The following query form is almost wrong: select * from emp where ROWNUM <= 5 order by sal desc; EMPNO ---------7566 7499 7521 7654 7369

ENAME ---------JONES ALLEN WARD MARTIN SMITH

/* WRONG! */

JOB MGR HIREDATE SAL COMM DEPTNO --------- ---------- --------- ---------- ---------- ---------MANAGER 7839 02-APR-81 2975 20 SALESMAN 7698 20-FEB-81 1600 300 30 SALESMAN 7698 22-FEB-81 1250 500 30 SALESMAN 7698 28-SEP-81 1250 1400 30 CLERK 7902 17-DEC-80 800 20

The users intention was most likely to get the the top-five paid people - a top-N query. What the will ge five we happen to hit), sorted by salary. If you use an inline view with the ORDER BY inside the inline vi select * from (select * from emp order by sal desc) where rownum <= 5; EMPNO ---------7839 7788 7902 7566 7698

ENAME ---------KING SCOTT FORD JONES BLAKE

JOB MGR HIREDATE SAL COMM DEPTNO --------- ---------- --------- ---------- ---------- ---------PRESIDENT 17-NOV-81 5000 10 ANALYST 7566 09-DEC-82 3000 20 ANALYST 7566 03-DEC-81 3000 20 MANAGER 7839 02-APR-81 2975 20 MANAGER 7839 01-MAY-81 2850 30

Example 5 (Pagination with ROWNUM)

Pagination with ROWNUM can be used to get rows N thru M of a result set. The general form of this is a SELECT * FROM (SELECT a.*, ROWNUM rn FROM (enter your query here) a WHERE ROWNUM <= :MAX_ROW) WHERE rn >= :MIN_ROW; SELECT * FROM (SELECT a.*, ROWNUM rn FROM (SELECT * FROM emp) a WHERE ROWNUM <= 6) WHERE rn >= 2;

EMPNO ENAME JOB MGR HIREDATE SAL COMM DEPTNO ---------- ---------- --------- ---------- --------- ---------- ---------- ---------- -

7499 7521 7566 7654 7698

ALLEN WARD JONES MARTIN BLAKE

SALESMAN SALESMAN MANAGER SALESMAN MANAGER

7698 7698 7839 7698 7839

20-FEB-81 22-FEB-81 02-APR-81 28-SEP-81 01-MAY-81

1600 1250 2975 1250 2850

300 500 1400

30 30 20 30 30

Related Documents