如何在对in操作使用变量绑定
如何在对in操作使用变量绑定
大家都知道在sql语句中变量绑定的重大意义,甚至有高人指出:一个应用想要它变的很糟糙的话,只要不使用变量绑定就可以了。这话的确不假。这时可能有人就会问:我该绑定的都绑定了,可是我实在无法想到好的方法在in操作符中使用变量绑定。下面的方法可以解决这个疑问。
我们知道in操作符接受两种list, 一个是由一个个item组成的list, 另一个是由另一个表中选出的list。第一种方式由于值个数不一定,变量绑定具有一定的困难。于是我们思路就集中到准备将一个字符串传入到sql语句中,然后使用一个方法将字符串parse成一个table, 再传回in operator:
首先我们来创建方法,及由此方法返回的table类型:
rudolf@test9i> create or replace type numTableType as table
2 of number
3 /
Type created.
rudolf@test9i> create or replace function str2numList( p_string in varchar2 ) return
2 numTableType
3 as
4 v_str long default p_string || ',';
5 v_n number;
6 v_data numTableType := numTableType();
7 begin
8 loop
9 v_n := to_number(instr( v_str, ',' ));
10 exit when (nvl(v_n,0) = 0);
11 v_data.extend;
12 v_data( v_data.count ) := ltrim(rtrim(substr(v_str,1,v_n-1)));
13 v_str := substr( v_str, v_n+1 );
14 end loop;
15 return v_data;
16 end;
17 /
Function created.
我们把下列语句:
select object_name from t where object_id in ( xx,xxx,xxx,...);
改为:
select object_name from t
where object_id in
( select * from THE ( select cast( str2numList(:variable ) as numtableType ) from dual )
);
现在我们来看看是否达到了我们的目的:
rudolf@test9i> var STR varchar2(3000)
rudolf@test9i> exec :STR := '5770,1810,4481'
PL/SQL procedure successfully completed.
rudolf@test9i> alter session set events= '10046 trace name context forever, level 4'
2 /
Session altered.
rudolf@test9i> select object_id,object_name from t where object_id in (
2 select * from THE ( select cast( str2numList(:STR ) as numtableType ) from
3 dual ) )
4
rudolf@test9i> /
OBJECT_ID OBJECT_NAME
---------- ------------------------------
1810 ALL_ALL_TABLES
4481 AGGXMLINPUTTYPE
5770 ALL_APPLY
检查dump文件,我们看到(注意星号的行):
PARSING IN CURSOR #1 len=146 dep=0 uid=81 oct=3 lid=81 tim=1036636837633718 hv=2833917919 ad='529b9f48'
select object_id,object_name from t where object_id in (
select * from THE ( select cast( str2numList(:STR ) as numtableType ) from
dual ) )
END OF STMT
PARSE #1:c=0,e=655,p=0,cr=0,cu=0,mis=1,r=0,dep=0,og=0,tim=1036636837633701
BINDS #1: *****
bind 0: dty=1 mxl=2000(200) mal=00 scl=00 pre=00 oacflg=03 oacfl2=10 size=2000 offset=0
bfp=406402fc bln=2000 avl=14 flg=05
value="5770,1810,4481"
EXEC #1:c=0,e=245,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=4,tim=1036636837645450 FETCH #1:c=20000,e=16739,p=0,cr=52,cu=0,mis=0,r=1,dep=0,og=4,tim=1036636837662283 FETCH #1:c=0,e=4644,p=0,cr=0,cu=0,mis=0,r=2,dep=0,og=4,tim=1036636837667979
看得出,它的确使用了变量绑定。
/******************the other article********************/
使用绑定变量带来的好处不用多说,但是遇到SQL条件中条件值不固定的这类SQL在普通方式下就不能使用绑定变量,看到Tom网站上提供了一种可行的办法。
网上有不少的地方也提到了这种方法,那就是,先定义一个表类型,将个数不定的参数转换成这种表类型,再将查询结果传给IN子句,这样就可以实现了SQL语句中的绑定变量方式。
例如:
Select * From 病人费用记录 Where NO IN(Select * From Table(Cast(zlTools.f_Str2List ('A01,A02,A03') as zlTools. t_StrList)));
Select * From 病人费用记录 Where NO IN(Select * From The(Select Cast(zlTools.f_Str2List ('A01,A02,A03') as zlTools. t_StrList) From Dual));
Select A.* From 病人费用记录 A,Table(Cast(zlTools.f_Str2List ('A01,A02,A03') as zlTools. t_StrList)) B Where A.NO=B.Column_Value;
Oracle提供了一个包dbms_utility.comma_to_table可以将参数列表转换为一个类型表,但是由于结果是通过参数传出的,无法用于应用程序的SQL语句中。
于是,我们仍然选择通过创建SQL表类型和相关函数来实现,Type:t_StrList, t_NumList, 函数:f_Str2List, f_Num2List,如下:
Create Or Replace type zltools.t_StrList as table of varchar2(4000)
/
Create Or Replace Function zltools.f_Str2List(Str_In In Varchar2) Return zltools.t_StrList As
v_Str Long Default Str_In || ',';
v_Index Number;
v_List zltools.t_StrList := zltools.t_StrList();
--功能:将由逗号分隔的不带引号的字符序列转换为数据表
--参数:Str_In如:G0000123,G0000124,G0000125...
Begin
Loop
v_Index := Instr(v_Str, ',');
Exit When(Nvl(v_Index, 0) = 0);
v_List.Extend;
v_List(v_List.Count) := Trim(Substr(v_Str, 1, v_Index - 1));
v_Str := Substr(v_Str, v_Index + 1);
End Loop;
Return v_List;
End;
/
Create Or Replace type zltools.t_NumList as table of number
/
Create Or Replace Function zltools.f_Num2List(Str_In In Varchar2) Return zltools.t_NumList As
v_Str Long Default Str_In || ',';
v_Index Number;
v_List zltools.t_NumList := zltools.t_NumList();
--功能:将由逗号分隔的数字序列转换为数据表
--参数:Str_In如:73265,73266,73267....
Begin
Loop
v_Index := Instr(v_Str, ',');
Exit When(Nvl(v_Index, 0) = 0);
v_List.Extend;
v_List(v_List.Count) := To_Number(Trim(Substr(v_Str, 1, v_Index - 1)));
v_Str := Substr(v_Str, v_Index + 1);
End Loop;
Return v_List;
End;
/
但是这种方法有一个问题:
如果用户使用CBO,使用这种方式的SQL语句,不加处理的话一般(要根据具体SQL分析)会错误的使用Hash连接来产生执行计划,不能正确使用索引从而引发性能问题,
因为Type Table不具备统计信息,在计算成本时,Oracle始终认为它有8168行,在基于成本的优化器模式下,多表连接时一般就会选择Hash连接,一般比使用索引的嵌套连接效率差。
目前,已知的解决办法有两种:
http://www.cnoug.org/viewthread.php?tid=1533&highlight=DROP%2B%2BCONSTRAINT%2Bindex
1. 对构造表的返回记录集加RowNum条件,通过子查询的实体化视图来改变执行计划,例如:
select * from table(Cast(zlTools.f_Str2List (v_str) as zlTools.t_StrList)) where rownum<5
2. 在SQL中加入优化器提示,例如:/*+ rule*/,或者/*+ ordered use_nl(t)*/,或者/*+ index(t ind_object_id)*/
第1种方式,需要对rownum的条件赋值,如果条件值采用动态的值,那么就达不到使用绑定变量的效果,如果采用固定值,有两种方案:
一种是使用一个固定的极限值,例如:99999,另一种是使用绑定变量。
试验结果见文末,这两种方式产生的执行计划采用了Hash连接,执行计划比不使用RowNum实体化视图更差。
第2种方式,如果使用指定嵌套连接提示或指定索引提示则需要对每个SQL针对性处理,需要分析具体的SQL。所以,一般情况建议在SQL中使用/*+ rule*/提示来固定执行计划。
--0.准备
--先创建表类型及转换函数: t_StrList,f_Str2List和t_NumList,f_Num2List
SQL> alter session set optimizer_mode=all_rows;
Session altered
SQL> var v_str varchar2(4000);
SQL> exec :v_str :='H0000005,H0000006,H0000007';
PL/SQL 过程已成功完成。
--1.当类型表不使用Rownum实体化视图时,错误地采用了Hash连接的执行计划,产生了6575块逻辑读+6909块物理读
SQL> Select a.No, a.结算方式, Nvl(b.性质, 1) As 性质, a.金额, a.摘要, a.结算号码
2 From (Select b.No, Decode(a.记录性质, 3, a.结算方式, Null) As 结算方式, a.摘要, a.结算号码, Sum
(1 * a.冲预交) As 金额
3 From 病人预交记录 a,
4 (Select Distinct No, 结帐id
5 From 病人费用记录
6 Where 记录性质 = 1 And No In (
7 Select * From Table(Cast(f_Str2List(:v_str) as t_StrList))
8 ) And 记录状态 In (1, 3)) b
9 Where a.结帐id = b.结帐id And a.记录性质 In (1, 11, 3) And Nvl(a.冲预交, 0) <> 0
10 Group By b.No, Decode(a.记录性质, 3, a.结算方式, Null), a.摘要, a.结算号码) a, 结算方式 b
11 Where a.结算方式 = b.名称(+);
执行计划
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=23 Card=1 Bytes=100)
1 0 NESTED LOOPS (OUTER) (Cost=23 Card=1 Bytes=100)
2 1 VIEW (Cost=22 Card=1 Bytes=75)
3 2 SORT (GROUP BY) (Cost=22 Card=1 Bytes=94)
4 3 VIEW (Cost=21 Card=1 Bytes=94)
5 4 SORT (UNIQUE) (Cost=21 Card=1 Bytes=149)
6 5 TABLE ACCESS (BY INDEX ROWID) OF '病人预交记录' (Cost=1 Card=1 Bytes=102)
7 6 NESTED LOOPS (Cost=19 Card=1 Bytes=149)
8 7 HASH JOIN (SEMI) (Cost=17 Card=4 Bytes=188)
9 8 INDEX (FAST FULL SCAN) OF '病人费用记录_UQ_NO' (UNIQUE) (Cost=4 Card=399 Bytes=17955)
10 8 COLLECTION ITERATOR (PICKLER FETCH) OF 'f_Str2List'
11 7 INDEX (RANGE SCAN) OF '病人预交记录_IX_结帐ID' (NON-UNIQUE)
12 1 TABLE ACCESS (BY INDEX ROWID) OF '结算方式' (Cost=1 Card=1 Bytes=25)
13 12 INDEX (UNIQUE SCAN) OF '结算方式_UQ_名称' (UNIQUE)
统计信息
----------------------------------------------------------
0 recursive calls
0 db block gets
6575 consistent gets
6909 physical reads
0 redo size
720 bytes sent via SQL*Net to client
372 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
3 rows processed
--2.当类型表的查询条件中Rownum使用固定值时,正确地采用了嵌套连接的执行计划,仅仅产生了29块逻辑读
SQL> Select a.No, a.结算方式, Nvl(b.性质, 1) As 性质, a.金额, a.摘要, a.结算号码
2 From (Select b.No, Decode(a.记录性质, 3, a.结算方式, Null) As 结算方式, a.摘要, a.结算号码, Sum
(1 * a.冲预交) As 金额
3 From 病人预交记录 a,
4 (Select Distinct No, 结帐id
5 From 病人费用记录
6 Where 记录性质 = 1 And No In (
7 Select * From Table(Cast(f_Str2List(:v_str) as t_StrList)) Where Rownum<4
8 ) And 记录状态 In (1, 3)) b
9 Where a.结帐id = b.结帐id And a.记录性质 In (1, 11, 3) And Nvl(a.冲预交, 0) <> 0
10 Group By b.No, Decode(a.记录性质, 3, a.结算方式, Null), a.摘要, a.结算号码) a, 结算方式 b
11 Where a.结算方式 = b.名称(+);
执行计划
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=23 Card=4 Bytes=400)
1 0 NESTED LOOPS (OUTER) (Cost=23 Card=4 Bytes=400)
2 1 VIEW (Cost=21 Card=4 Bytes=300)
3 2 SORT (GROUP BY) (Cost=21 Card=4 Bytes=376)
4 3 VIEW (Cost=20 Card=4 Bytes=376)
5 4 SORT (UNIQUE) (Cost=20 Card=4 Bytes=8596)
6 5 TABLE ACCESS (BY INDEX ROWID) OF '病人预交记录' (Cost=1 Card=1 Bytes=102)
7 6 NESTED LOOPS (Cost=17 Card=4 Bytes=8596)
8 7 NESTED LOOPS (Cost=13 Card=12 Bytes=24564)
9 8 VIEW OF 'VW_NSO_1' (Cost=11 Card=3 Bytes=6006)
10 9 COUNT (STOPKEY)
11 10 COLLECTION ITERATOR (PICKLER FETCH) OF 'f_Str2List'
12 8 INLIST ITERATOR
13 12 INDEX (RANGE SCAN) OF '病人费用记录_UQ_NO' (UNIQUE) (Cost=1 Card=4 Bytes=180)
14 7 INDEX (RANGE SCAN) OF '病人预交记录_IX_结帐ID' (NON-UNIQUE)
15 1 TABLE ACCESS (BY INDEX ROWID) OF '结算方式' (Cost=1 Card=1 Bytes=25)
16 15 INDEX (UNIQUE SCAN) OF '结算方式_UQ_名称' (UNIQUE)
统计信息
----------------------------------------------------------
0 recursive calls
0 db block gets
29 consistent gets
0 physical reads
0 redo size
720 bytes sent via SQL*Net to client
372 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
3 rows processed
--3.当类型表的查询条件中Rownum使用变量时,执行计划虽然有变化,但是仍然认为类型表返回的记录集数为8168,
-- 错误地采用了Hash连接的执行计划,产生了374369块逻辑读和11192块物理读
SQL> var v_count number;
SQL> exec :v_count:=4;
PL/SQL 过程已成功完成。
SQL> Select a.No, a.结算方式, Nvl(b.性质, 1) As 性质, a.金额, a.摘要, a.结算号码
2 From (Select b.No, Decode(a.记录性质, 3, a.结算方式, Null) As 结算方式, a.摘要, a.结算号码, Sum
(1 * a.冲预交) As 金额
3 From 病人预交记录 a,
4 (Select Distinct No, 结帐id
5 From 病人费用记录
6 Where 记录性质 = 1 And No In (
7 Select * From Table(Cast(f_Str2List(:v_str) as t_StrList)) Where Rownum<:v_count
8 ) And 记录状态 In (1, 3)) b
9 Where a.结帐id = b.结帐id And a.记录性质 In (1, 11, 3) And Nvl(a.冲预交, 0) <> 0
10 Group By b.No, Decode(a.记录性质, 3, a.结算方式, Null), a.摘要, a.结算号码) a, 结算方式 b
11 Where a.结算方式 = b.名称(+);
执行计划
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=16306 Card=10945 Bytes=1094500)
1 0 HASH JOIN (OUTER) (Cost=16306 Card=10945 Bytes=1094500)
2 1 VIEW (Cost=16264 Card=10945 Bytes=820875)
3 2 SORT (GROUP BY) (Cost=16264 Card=10945 Bytes=1028830)
4 3 VIEW (Cost=16148 Card=10945 Bytes=1028830)
5 4 SORT (UNIQUE) (Cost=16148 Card=10945 Bytes=23520805)
6 5 HASH JOIN (Cost=238 Card=10945 Bytes=23520805)
7 6 TABLE ACCESS (BY INDEX ROWID) OF '病人预交记录' (Cost=1 Card=1 Bytes=102)
8 7 NESTED LOOPS (Cost=164 Card=134 Bytes=19698)
9 8 INDEX (FAST FULL SCAN) OF '病人费用记录_UQ_NO' (UNIQUE) (Cost=4 Card=399 Bytes=17955)
10 8 INDEX (RANGE SCAN) OF '病人预交记录_IX_结帐ID' (NON-UNIQUE)
11 6 VIEW OF 'VW_NSO_1' (Cost=11 Card=8168 Bytes=16352336)
12 11 COUNT (STOPKEY)
13 12 COLLECTION ITERATOR (PICKLER FETCH) OF 'f_Str2List'
14 1 TABLE ACCESS (FULL) OF '结算方式' (Cost=2 Card=82 Bytes=2050)
统计信息
----------------------------------------------------------
28 recursive calls
0 db block gets
374369 consistent gets
11192 physical reads
0 redo size
720 bytes sent via SQL*Net to client
372 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
3 rows processed
--4.当类型表的查询条件中Rownum使用一个固定的极限值时(例如:99999),情况与使用变量一样遭,错误地采用了Hash连接的执行计划,产生了374354块逻辑读和10756块物理读
SQL> Select a.No, a.结算方式, Nvl(b.性质, 1) As 性质, a.金额, a.摘要, a.结算号码
2 From (Select b.No, Decode(a.记录性质, 3, a.结算方式, Null) As 结算方式, a.摘要, a.结算号码, Sum
(1 * a.冲预交) As 金额
3 From 病人预交记录 a,
4 (Select Distinct No, 结帐id
5 From 病人费用记录
6 Where 记录性质 = 1 And No In (
7 Select * From Table(Cast(f_Str2List(:v_str) as t_StrList)) Where Rownum<99999
8 ) And 记录状态 In (1, 3)) b
9 Where a.结帐id = b.结帐id And a.记录性质 In (1, 11, 3) And Nvl(a.冲预交, 0) <> 0
10 Group By b.No, Decode(a.记录性质, 3, a.结算方式, Null), a.摘要, a.结算号码) a, 结算方式 b
11 Where a.结算方式 = b.名称(+);
执行计划
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=16306 Card=10945 Bytes=1094500)
1 0 HASH JOIN (OUTER) (Cost=16306 Card=10945 Bytes=1094500)
2 1 VIEW (Cost=16264 Card=10945 Bytes=820875)
3 2 SORT (GROUP BY) (Cost=16264 Card=10945 Bytes=1028830)
4 3 VIEW (Cost=16148 Card=10945 Bytes=1028830)
5 4 SORT (UNIQUE) (Cost=16148 Card=10945 Bytes=23520805)
6 5 HASH JOIN (Cost=238 Card=10945 Bytes=23520805)
7 6 TABLE ACCESS (BY INDEX ROWID) OF '病人预交记录' (Cost=1 Card=1 Bytes=102)
8 7 NESTED LOOPS (Cost=164 Card=134 Bytes=19698)
9 8 INDEX (FAST FULL SCAN) OF '病人费用记录_UQ_NO' (UNIQUE) (Cost=4 Card=399 Bytes=17955)
10 8 INDEX (RANGE SCAN) OF '病人预交记录_IX_结帐ID' (NON-UNIQUE)
11 6 VIEW OF 'VW_NSO_1' (Cost=11 Card=8168 Bytes=16352336)
12 11 COUNT (STOPKEY)
13 12 COLLECTION ITERATOR (PICKLER FETCH) OF 'f_Str2List'
14 1 TABLE ACCESS (FULL) OF '结算方式' (Cost=2 Card=82 Bytes=2050)
统计信息
----------------------------------------------------------
0 recursive calls
0 db block gets
374354 consistent gets
10756 physical reads
0 redo size
720 bytes sent via SQL*Net to client
372 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
3 rows processed
--5.强制使用基于规则的优化器模式
SQL> Select /*+ rule*/ a.No, a.结算方式, Nvl(b.性质, 1) As 性质, a.金额, a.摘要, a.结算号码
2 From (Select b.No, Decode(a.记录性质, 3, a.结算方式, Null) As 结算方式, a.摘要, a.结算号码, Sum
(1 * a.冲预交) As 金额
3 From 病人预交记录 a,
4 (Select Distinct No, 结帐id
5 From 病人费用记录
6 Where 记录性质 = 1 And No In (
7 Select * From Table(Cast(f_Str2List(:v_str) as t_StrList))
8 ) And 记录状态 In (1, 3)) b
9 Where a.结帐id = b.结帐id And a.记录性质 In (1, 11, 3) And Nvl(a.冲预交, 0) <> 0
10 Group By b.No, Decode(a.记录性质, 3, a.结算方式, Null), a.摘要, a.结算号码) a, 结算方式 b
11 Where a.结算方式 = b.名称(+);
执行计划
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=HINT: RULE
1 0 NESTED LOOPS (OUTER)
2 1 VIEW
3 2 SORT (GROUP BY)
4 3 VIEW
5 4 SORT (UNIQUE)
6 5 TABLE ACCESS (BY INDEX ROWID) OF '病人预交记录'
7 6 NESTED LOOPS
8 7 NESTED LOOPS
9 8 COLLECTION ITERATOR (PICKLER FETCH) OF 'f_Str2List'
10 8 INDEX (RANGE SCAN) OF '病人费用记录_UQ_NO' (UNIQUE)
11 7 INDEX (RANGE SCAN) OF '病人预交记录_IX_结帐ID' (NON-UNIQUE)
12 1 TABLE ACCESS (BY INDEX ROWID) OF '结算方式'
13 12 INDEX (UNIQUE SCAN) OF '结算方式_UQ_名称' (UNIQUE)
统计信息
----------------------------------------------------------
0 recursive calls
0 db block gets
25 consistent gets
0 physical reads
0 redo size
720 bytes sent via SQL*Net to client
372 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
2 sorts (memory)
0 sorts (disk)
3 rows processed