在Oracle中,给出下面语句的几种可能的优化思路。
CREATE TABLE T_YH_20170705_LHR( X INT);
BEGIN
FOR I IN 1 .. 100000 LOOP
EXECUTE IMMEDIATE 'INSERT INTO T_YH_20170705_LHR VALUES ( '||I||')';
COMMIT;
END LOOP;
END;
A 答案如下所示:
优化思路有:①采用绑定变量;②使用静态SQL;③采用批量提交或循环外提交;④根据功能,可以去掉PL/SQL块,采用直接一次性插入的方式来完成,SQL为“INSERT INTO T_YH_20170705_LHR SELECT ROWNUM FROM DUAL CONNECT BY LEVEL<=100000;”;⑤采用直接路径方式,例如,“CREATE TABLE T_YH_20170705_LHR AS SELECT ROWNUM X FROM DUAL CONNECT BY LEVEL<=100000;”;⑥采用NOLOGGING和PARALLEL的方式,例如,“CREATE TABLE T_YH_20170705_LHR NOLOGGING PARALLEL 8 AS SELECT ROWNUM X FROM DUAL CONNECT BY LEVEL<=100000;”。
实验如下所示:
DROP TABLE T_YH_20170705_LHR PURGE;
CREATE TABLE T_YH_20170705_LHR( X INT);
ALTER SYSTEM FLUSH SHARED_POOL;
CREATE OR REPLACE PROCEDURE PRO_YH_LHR
AS
BEGIN
FOR I IN 1 .. 100000 LOOP
EXECUTE IMMEDIATE 'INSERT INTO T_YH_20170705_LHR VALUES ( '||I||')';
COMMIT;
END LOOP;
END;
/
SET TIMING ON
EXEC PRO_YH_LHR; --Elapsed: 00:00:59.45
首次实验需要40秒完成,因为未用绑定变量:
SELECT T.SQL_TEXT, T.SQL_ID,T.PARSE_CALLS, T.EXECUTIONS
FROM V$SQL T
WHERE SQL_TEXT LIKE 'INSERT INTO T_YH_20170705_LHR VALUES%';
一、 采用绑定变量
DROP TABLE T_YH_20170705_LHR PURGE;
CREATE TABLE T_YH_20170705_LHR( X INT);
ALTER SYSTEM FLUSH SHARED_POOL;
CREATE OR REPLACE PROCEDURE PRO_YH_LHR2
AS
BEGIN
FOR I IN 1 .. 100000 LOOP
EXECUTE IMMEDIATE 'INSERT INTO T_YH_20170705_LHR VALUES ( :X )' USING I;
COMMIT;
END LOOP;
END;
/
SET TIMING ON
EXEC PRO_YH_LHR2; --Elapsed: 00:00:08.14
采用绑定变量后,8秒钟执行完成。SQL解析次数也变为了1次:
SELECT T.SQL_TEXT, T.SQL_ID,T.PARSE_CALLS, T.EXECUTIONS
FROM V$SQL T
WHERE SQL_TEXT LIKE 'INSERT INTO T_YH_20170705_LHR VALUES%';
二、 将动态SQL改造成静态SQL
DROP TABLE T_YH_20170705_LHR PURGE;
CREATE TABLE T_YH_20170705_LHR ( X INT);
ALTER SYSTEM FLUSH SHARED_POOL;
CREATE OR REPLACE PROCEDURE PRO_YH_LHR3
AS
BEGIN
FOR I IN 1 .. 100000 LOOP
INSERT INTO T_YH_20170705_LHR VALUES (I);
COMMIT;
END LOOP;
END;
/
SET TIMING ON
EXEC PRO_YH_LHR3; --Elapsed: 00:00:05.99
改造后6秒钟即可完成,再次查看解析情况:
SELECT T.SQL_TEXT, T.SQL_ID,T.PARSE_CALLS, T.EXECUTIONS
FROM V$SQL T
WHERE SQL_TEXT LIKE 'INSERT INTO T_YH_20170705_LHR VALUES%';
静态SQL会自动使用绑定变量。
三、 循环外提交
DROP TABLE T_YH_20170705_LHR PURGE;
CREATE TABLE T_YH_20170705_LHR ( X INT);
ALTER SYSTEM FLUSH SHARED_POOL;
CREATE OR REPLACE PROCEDURE PRO_YH_LHR4
AS
BEGIN
FOR I IN 1 .. 100000 LOOP
INSERT INTO T_YH_20170705_LHR VALUES (I);
END LOOP;
COMMIT;
END;
/
SET TIMING ON
EXEC PRO_YH_LHR4; --Elapsed: 00:00:04.75
5秒即可完成。
四、 去掉存储过程,采用插入表的写法
DROP TABLE T_YH_20170705_LHR PURGE;
CREATE TABLE T_YH_20170705_LHR( X INT);
ALTER SYSTEM FLUSH SHARED_POOL;
SET TIMING ON
INSERT INTO T_YH_20170705_LHR SELECT ROWNUM FROM DUAL CONNECT BY LEVEL<=100000;--Elapsed: 00:00:00.01
COMMIT;
SELECT COUNT(*) FROM T_YH_20170705_LHR;
用了0.01秒,太快了。由于要使用到Data Buffer,所以下面采用直接路径。
五、 直接路径
DROP TABLE T_YH_20170705_LHR PURGE;
ALTER SYSTEM FLUSH SHARED_POOL;
SET TIMING ON
CREATE TABLE T_YH_20170705_LHR AS SELECT ROWNUM X FROM DUAL CONNECT BY LEVEL<=100000;
六、 并行原理
DROP TABLE T_YH_20170705_LHR PURGE;
ALTER SYSTEM FLUSH SHARED_POOL;
SET TIMING ON
CREATE TABLE T_YH_20170705_LHR NOLOGGING PARALLEL 8 AS SELECT ROWNUM X FROM DUAL CONNECT BY LEVEL<=100000;