1 Oracle 隐式转换
Oracle中对不同类型的处理具有显式类型转换(Explicit)和隐式类型转换(Implicit)两种方式，对于显式类型转换，我们是可控的，但是对于隐式类型转换，当然不建议使用,
因为很难控制，有不少缺点，但是我们很难避免碰到隐式类型转换,如果不了解隐式类型转换的规则，那么往往会改变我们SQL的执行计划，从而可能导致效率降低或其它问题。
1.1 隐式转换发生场景
1.对于INSERT和UPDATE操作，oracle会把插入值或者更新值隐式转换为字段的数据类型。
例如：
SQL> create table text(id varchar2(32),name varchar2(10),age number);
Table created.
SQL> insert into text values ('1','Jack','18');
1 row created.
SQL> update text set age='19';
1 rows updated.
SQL> select * from text;
ID NAME AGE
---------- ---------- ----------
1 Jack 19
注：insert into text values ('1','Jack','18') 相当于 insert into text values(,’1’,’Jack’,to_number('18'))
update text set age='19'相当于update text set age=to_number('19')
2.当比较字符型和数值型的值时，oracle会把字符型的值隐式转换为数值型。例如：
SQL> explain plan for select * from text where age='19';
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 738342525
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2 | 66 | 2 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| TEXT | 2 | 66 | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("AGE"=19)
注：select * from text where age='19'相当于select * from text where age=to_number('19')
SQL> explain plan for select * from text where id=1;
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 738342525
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 38 | 2 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| TEXT | 1 | 38 | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(TO_NUMBER("ID")=1)
注：select * from text where id=1;相当于select * from text where to_number(id)=1
如果id列建有索引此时将失效
3.当比较字符型和日期型的数据时，oracle会把字符型转换为日期型。例如：
SQL> create table table_date(varchar_date varchar2(20),date_date Date);
Table created.
SQL> insert into table_date values(to_char(sysdate,'yyyy-mm-dd'),sysdate);
1 row created.
SQL> select * from table_date;
VARCHAR_DATE DATE_DATE
-------------------- ---------
2014-02-26 26-FEB-14
SQL> alter session set nls_date_format='yyyy-mm-dd hh24:mi:ss';
Session altered.
SQL> explain plan for select * from table_date where varchar_date
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 1510990824
--------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 21 | 2 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| TABLE_DATE | 1 | 21 | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(INTERNAL_FUNCTION("VARCHAR_DATE")
注：select * from table_date where varchar_date相当于
select * from table_date where to_date(varchar_date,’yyyy-mm-dd hh24:mi:ss’)
SQL> explain plan for select * from table_date where date_date>'2014-2-26 0:0:0';
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 1510990824
--------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 21 | 2 (0)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| TABLE_DATE | 1 | 21 | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("DATE_DATE">TO_DATE(' 2014-02-26 00:00:00', 'syyyy-mm-dd
hh24:mi:ss'))
注：select * from table_date where date_date>'2014-2-26 0:0:0'相当于
select * from table_date where date_date>to_date('2014-2-26 0:0:0’, ’yyyy-mm-dd hh24:mi:ss’
4. 隐式转换发正在字段列上时将使索引失效。例如：
1）当末发生隐式转换时索引有效
SQL> create table t1 as select OBJECT_ID as id ,to_char(OBJECT_ID) as vid from dba_objects;
Table created.
SQL> desc t1
Name Null? Type
----------------------------------------- -------- ----------------------------
ID NUMBER
VID VARCHAR2(40)
SQL> create index t1_ind_vid on t1(vid);
Index created.
SQL> explain plan for select * from t1 where vid='15612';
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 1215445203
------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 35 | 2 (0)| 00:00:01 |
| 1 | TABLE ACCESS BY INDEX ROWID| T1 | 1 | 35 | 2 (0)| 00:00:01 |
|* 2 | INDEX RANGE SCAN | T1_IND_VID | 1 | | 1 (0)| 00:00:01 |
------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("VID"='15612')
注：未发生隐式转换正常执行索引扫
2）当字段列发生隐式转换时索引将失效
SQL> explain plan for select * from t1 where vid=15612;
Explained.
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
-------------------------------------------------------------------------------
Plan hash value: 3617692013
--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 11 | 48 (5)| 00:00:01 |
|* 1 | TABLE ACCESS FULL| T1 | 1 | 11 | 48 (5)| 00:00:01 |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter(TO_NUMBER("VID")=15612)
注：select * from t1 where vid=15612 相当于select * from t1 where to_number(vid)=15612,vid列发生隐式转换执行计划显示全表扫描末使用索引。
1.2 隐式转换的缺点
1. 使用显示类型转换会让我们的SQL更加容易被理解,也就是可读性更强,但是隐式类型转换却没有这个优点。
2. 隐式类型转换往往对性能产生不好的影响，特别是左值的类型被隐式转为了右值的类型。这种方式很可能使我们本来可以使用索引的而没有用上索引,也有可能会导致结
果出错。
3. 隐式类型转换可能依赖于发生转换时的上下文环境，比如1中的to_date(sysdate,fmt),一旦上下文环境改变，很可能我们的程序就不能运行。
4. 隐式类型转换的算法或规则，以后Oracle可能改变，这是很危险的，意味着旧的代码很可能在新的Oracle版本中运行出现问题(性能、错误等)，显示类型转换总是有最高
的优先级，所以显示类型转换没有这种版本更替可能带来的问题。
5. 隐式类型转换是要消耗时间的,当然同等的显式类型转换时间也差不多,最好的方法就是避免类似的转换,在显示类型转换上我们会看到，最好不要将左值进行类型转换，到
时候有索引也用不上索引，还要建函数索引，索引储存和管理开销增大。
小结
Oracle使用数据类型的优先级来决定隐式类型转换，原则是将优先级低的转换为优先级高的（数据类型优先级为：Number>字符类型>日期类型）。隐式转换发生在字
段列上时将使索引失效。
DBA_建瑾
2014.2.27
来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/29324876/viewspace-1096741/，如需转载，请注明出处，否则将追究法律责任。
转载于:http://blog.itpub.net/29324876/viewspace-1096741/
617
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
