Oracle硬解析的几个例子

为了验证SQL硬解析的场景,设置了下面六个测试用的例子:

1、没有绑定变量下的普通查询

2、测试绑定变量下的查询

3、测试绑定变量下sql有变化的查询

4、测试DML非绑定变量的解析

5、测试在过程中执行插入的时候非绑定变量的SQL解析

6、使用了绑定变量之后的,过程中的SQL解析情况

[sql] 
/**  
测试例子1:  
   没有绑定变量下的普通查询  
**/  
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
SELECT * FROM foo WHERE x = 100;  
SELECT * FROM foo WHERE x =999;  
SELECT * FROM foo WHERE x=10000;  
  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';  


[sql] 
/**  
测试例子2:  
   测试绑定变量下的查询  
**/  
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
VARIABLE temp NUMBER;  
  
exec :temp :=99;  
SELECT * FROM foo WHERE X = :temp;  
  
exec :temp :=100;  
SELECT * FROM foo WHERE X = :temp;  
  
exec :temp :=101;  
SELECT * FROM foo WHERE X = :temp;  
  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';  
  
/**  




[sql] 
测试例子3:  
   测试绑定变量下sql有变化的查询  
**/  
  
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
VARIABLE temp NUMBER;  
  
exec :temp :=99;  
SELECT * FROM foo WHERE X = :temp;  
  
exec :temp :=100;  
SELECT * FROM FOO WHERE X = :temp;  
  
exec :temp :=101;  
SELECT * FROM foo WHERE X = :temp;  
  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';  



[sql] 
/**  
测试例子4:  
   测试DML非绑定变量的解析  
**/  
   
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
INSERT INTO FOO VALUES(100,200);  
INSERT INTO FOO VALUES(101,201);  
INSERT INTO FOO VALUES(103,203);  
  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';  



[sql] 
/**  
测试例子5:  
   测试在过程中执行插入的时候的SQL解析  
**/  
  
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
BEGIN  
  FOR I IN 1..3 LOOP  
      IF I=1 THEN  
        INSERT INTO FOO VALUES(1,1);   
      ELSIF I=2 THEN   
        INSERT INTO FOO VALUES(2,2);  
      ELSIF I=3 THEN  
        INSERT INTO FOO VALUES(3,3);  
      END IF;    
  END LOOP;  
END;  
/  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';  




[sql] 
/**  
测试例子6:  
   使用了绑定变量之后的,过程中的SQL解析情况  
**/  
  
drop table foo purge;  
  
CREATE TABLE foo AS   
SELECT LEVEL AS x,100000-LEVEL AS y FROM dual  CONNECT BY LEVEL<=100000;  
  
ALTER SYSTEM FLUSH SHARED_POOL;  
  
BEGIN  
  FOR I IN 1..200 LOOP  
    INSERT INTO FOO VALUES(I,100000-I);  
  END LOOP;  
END;  
/  
SELECT T.SQL_TEXT, T.SQL_ID, T.EXECUTIONS, T.PARSE_CALLS  
  FROM V$SQL T  
 WHERE UPPER(T.SQL_TEXT) LIKE '%FOO%';

通过上述六个情况的试验，我们最终可以得到如下结论：

Oracle进行软解析的SQL必须是完全相同的，所谓相同的SQL必须是大小写一致（测试例子3），甚至是不能多一个或者少一个空格，这个结论可以通过修改测试例子3增加一个空格得到，结果就得到了不同的SQL_ID。只有完全一致的SQL，才可以得到相应的HASH值，从而才可以进行软解析。对于在SQL池中，我们需要分析在SQL池中出现的只有参数部分不同的SQL，如果出现了很多次，我们就有必要对其进行相应的变量绑定，从而降低硬解析成本，提高性能。