ORACLE SQL效能最佳化系列 (十一) (轉)

ORACLE SQL效能最佳化系列 (十一) (轉)[@more@]

36.  用UNION替換OR (適用於列):namespace prefix = o ns = "urn:schemas--com::office" />

通常情況下, 用UNION替換WHERE子句中的OR將會起到較好的效果. 對索引列使用OR將造成全表掃描. 注意, 以上規則只針對多個索引列有效. 如果有column沒有被索引, 查詢可能會因為你沒有選擇OR而降低.

  在下面的例子中, LOC_ID 和REGION上都建有索引.

高效:

  LOC_ID , LOC_DESC , REGION

  FROM LOCATION

  WHERE LOC_ID = 10

  UNION

  SELECT LOC_ID , LOC_DESC , REGION

  FROM LOCATION

  WHERE REGION = “MELBOURNE”

低效:

  SELECT LOC_ID , LOC_DESC , REGION

  FROM LOCATION

  WHERE LOC_ID = 10 OR REGION = “MELBOURNE”

如果你堅持要用OR, 那就需要返回記錄最少的索引列寫在最前面.

注意:

WHERE KEY1 = 10  (返回最少記錄)

OR KEY2 = 20  (返回最多記錄)

內部將以上轉換為

WHERE KEY1 = 10 AND

((NOT KEY1 = 10) AND KEY2 = 20) 

譯者按:

下面的測試資料僅供參考: (a = 1003 返回一條記錄 , b = 1 返回1003條記錄)

> select * from unionvsor /*1st test*/

  2  where a = 1003 or b = 1;

1003 rows selected.

Execution Plan

----------------------------------------------------------

  0  SELECT STATEMENT Optimizer=CHOOSE

  1  0  CONCATENATION

  2  1  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  3  2  INDEX (RANGE SCAN) OF 'UB' (NON-UNIQUE)

  4  1  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  5  4  INDEX (RANGE SCAN) OF 'UA' (NON-UNIQUE)

Statistics

----------------------------------------------------------

  0  recursive calls

  0  block gets

  144  consistent gets

  0  physical reads

  0  redo size

  63749  bytes sent via SQL to client

  7751  bytes received via SQL*Net from client

  68  SQL*Net roundtrips to/from client

  0  sorts (memory)

  0  sorts (disk)

  1003  rows processed

SQL> select * from unionvsor /*2nd test*/

  2  where b  = 1 or a = 1003 ;

1003 rows selected.

Execution Plan

----------------------------------------------------------

  0  SELECT STATEMENT Optimizer=CHOOSE

  1  0  CONCATENATION

  2  1  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  3  2  INDEX (RANGE SCAN) OF 'UA' (NON-UNIQUE)

  4  1  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  5  4  INDEX (RANGE SCAN) OF 'UB' (NON-UNIQUE)

Statistics

----------------------------------------------------------

  0  recursive calls

  0  db block gets

  143  consistent gets

  0  physical reads

   0  redo size

  63749  bytes sent via SQL*Net to client

  7751  bytes received via SQL*Net from client

  68  SQL*Net roundtrips to/from client

  0  sorts (memory)

  0  sorts (disk)

  1003  rows processed

SQL> select * from unionvsor /*3rd test*/

  2  where a = 1003

  3  union

  4  select * from unionvsor

  5  where b = 1;

1003 rows selected.

Execution Plan

----------------------------------------------------------

  0  SELECT STATEMENT Optimizer=CHOOSE

  1  0  SORT (UNIQUE)

  2  1  UNION-ALL

  3  2  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  4  3  INDEX (RANGE SCAN) OF 'UA' (NON-UNIQUE)

  5  2  TABLE ACCESS (BY INDEX ROWID) OF 'UNIONVSOR'

  6  5  INDEX (RANGE SCAN) OF 'UB' (NON-UNIQUE)

Statistics

----------------------------------------------------------

  0  recursive calls

  0  db block gets

  10  consistent gets 

  0  physical reads

  0  redo size

  63735  bytes sent via SQL*Net to client

  7751  bytes received via SQL*Net from client

  68  SQL*Net roundtrips to/from client

  1  sorts (memory)

  0  sorts (disk)

  1003  rows processed

用UNION的效果可以從consistent gets和 SQL*NET的資料量的減少看出

37.  用IN來替換OR

下面的查詢可以被更有效率的語句替換:

低效:

SELECT….

FROM LOCATION

WHERE LOC_ID = 10

OR  LOC_ID = 20

OR  LOC_ID = 30

高效

SELECT…

FROM LOCATION

WHERE LOC_IN IN (10,20,30);

 

譯者按:

這是一條簡單易記的規則，但是實際的效果還須檢驗，在下，兩者的執行路徑似乎是相同的．　

38.  避免在索引列上使用IS NULL和IS NOT NULL

避免在索引中使用任何可以為空的列，ORACLE將無法使用該索引 ．對於單列索引，如果列包含空值，索引中將不存在此記錄. 對於複合索引，如果每個列都為空，索引中同樣不存在此記錄.　如果至少有一個列不為空，則記錄存在於索引中．

舉例:

  如果唯一性索引建立在表的A列和B列上, 並且表中存在一條記錄的A,B值為(123,null) , ORACLE將不接受下一條具有相同A,B值（123,null）的記錄(插入). 然而如果

所有的索引列都為空，ORACLE將認為整個鍵值為空而空不等於空. 因此你可以插入1000

條具有相同鍵值的記錄,當然它們都是空!

  因為空值不存在於索引列中,所以WHERE子句中對索引列進行空值比較將使ORACLE停用該索引.

舉例:

低效: (索引失效)

SELECT …

FROM DEPARTMENT

WHERE DEPT_CODE IS NOT NULL;

高效: (索引有效)

SELECT …

FROM DEPARTMENT

WHERE DEPT_CODE >=0;