transaction behaviour with subselect - oracle

Question

I am wondering about the following strange behaviour.
This function should log the selected data to a table ps_cs_corr_data_tb (this table is empty):
create or replace function cs_corr_data(i_id in varchar2,
i_key1 in varchar2,
i_key2 in varchar2,
i_key3 in varchar2,
i_key4 in varchar2,
i_key5 in varchar2)
return number as pragma autonomous_transaction;
begin
insert into ps_cs_corr_data_tb
(descr,
cs_key_id_01,
cs_key_id_02,
cs_key_id_03,
cs_key_id_04,
cs_key_id_05)
values
(i_id, i_key1, i_key2, i_key3, i_key4, i_key5);
commit;
return 1; /* insert successful */
exception
when dup_val_on_index then
return 0;
end;
Test a)
The test with the following select statement is successful (as expected):
select b.id, b.key1, b.key2, b.key3, b.key4, b.key5
from (select a.id, a.key1, a.key2, a.key3, a.key4, a.key5
from ( -- test data
select '1' as id,'1' as key1,' ' as key2,' ' as key3,' ' as key4,' ' as key5 from dual union all
select '1' as id,'2' as key1,' ' as key2,' ' as key3,' ' as key4,' ' as key5 from dual union all
select '1' as id,'3' as key1,' ' as key2,' ' as key3,' ' as key4,' ' as key5 from dual union all
select '1' as id,'4' as key1,' ' as key2,' ' as key3,' ' as key4,' ' as key5 from dual union all
select '1' as id,'5' as key1,' ' as key2,' ' as key3,' ' as key4,' ' as key5 from dual
) a
-- some conditions
where a.id = '1'
and a.key1 = '4') b
-- log the results of selection
where cs_corr_data(b.id, b.key1, b.key2, b.key3, b.key4, b.key5) = 1;
result of selection:
ID KEY1 KEY2 KEY3 KEY4 KEY5
1 4
result in logging table:
select * from ps_cs_corr_data_tb d;
DESCR CS_KEY_ID_01 CS_KEY_ID_02 CS_KEY_ID_03 CS_KEY_ID_04 CS_KEY_ID_05
1 4
So far the expected result!
Explain Plan:
Plan hash value: 334628103
-------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
-------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 5 | 90 | 2 (0)| 00:00:01 |
| 1 | VIEW | | 5 | 90 | 2 (0)| 00:00:01 |
| 2 | UNION-ALL | | | | | |
|* 3 | FILTER | | | | | |
| 4 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
|* 5 | FILTER | | | | | |
| 6 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
|* 7 | FILTER | | | | | |
| 8 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
|* 9 | FILTER | | | | | |
| 10 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
|* 11 | FILTER | | | | | |
| 12 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
-------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
3 - filter(NULL IS NOT NULL AND "CS_CORR_DATA"('1','1',' ',' ',' ','
')=1)
5 - filter(NULL IS NOT NULL AND "CS_CORR_DATA"('1','2',' ',' ',' ','
')=1)
7 - filter(NULL IS NOT NULL AND "CS_CORR_DATA"('1','3',' ',' ',' ','
')=1)
9 - filter("CS_CORR_DATA"('1','4',' ',' ',' ',' ')=1)
11 - filter(NULL IS NOT NULL AND "CS_CORR_DATA"('1','5',' ',' ',' ','
')=1)
Test b)
Now the same test with different test data preparation (but the same test data):
select b.id, b.key1, b.key2, b.key3, b.key4, b.key5
from (select a.id, a.key1, a.key2, a.key3, a.key4, a.key5
from (select '1' as id,
to_char(level) as key1,
' ' as key2,
' ' as key3,
' ' as key4,
' ' as key5
from dual
connect by level <= 5) a
where a.id = '1'
and a.key1 = '4') b
where cs_corr_data(b.id, b.key1, b.key2, b.key3, b.key4, b.key5) = 1;
result of selection:
ID KEY1 KEY2 KEY3 KEY4 KEY5
1 4
result in logging table:
select * from ps_cs_corr_data_tb d;
DESCR CS_KEY_ID_01 CS_KEY_ID_02 CS_KEY_ID_03 CS_KEY_ID_04 CS_KEY_ID_05
1 1
1 2
1 3
1 4
1 5
Explain Plan:
Plan hash value: 2403765415
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 37 | 2 (0)| 00:00:01 |
|* 1 | VIEW | | 1 | 37 | 2 (0)| 00:00:01 |
|* 2 | CONNECT BY WITHOUT FILTERING| | | | | |
| 3 | FAST DUAL | | 1 | | 2 (0)| 00:00:01 |
--------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("CS_CORR_DATA"("A"."ID","A"."KEY1","A"."KEY2","A"."KEY3","A"."KE
Y4","A"."KEY5")=1 AND "A"."ID"='1' AND "A"."KEY1"='4')
2 - filter(LEVEL<=5)
Any ideas what is going on here?

Answer 1

Oracle (along with just about any relational database) is free to evaluate predicates in whatever order it expects would be most efficient. In either query, it is free to evaluate the function predicate first or to evaluate the a.id = '1' and a.key1 = '4' predicates first or to evaluate the function predicate between those two predicates. It appears that the actual plan the optimizer chose in the second case (at least this time) was to evaluate the function first while it chose to evaluate the function last in the first case. Of course, the optimizer is free to change its mind tomorrow in both cases so you shouldn't depend on a particular query plan.