Explain Plan Questions

Question

Hi All,
Q1. Performance wise are the following queries both same.
Which one will yield a better throughput and why?
Approach 1
SELECT
e.desig "Emp Desig",
(select count(1) from emp where empid = e.empid and dt_created between sysdate and sysdate -1) "Emp added since yday",
(select count(1) from emp where empid = e.empid and case_solved_count > 10 ) "Emp with case solved counts > 10",
FROM EMP e
Approach 2
SELECT
e.desig "Emp Desig",
Func1(e.empid ),
Func2(e.empid )
FROM emp e
Q2. If we have three tables T1, T2 and T3 with following record counts - Will their sequence of placement in WHERE clause of the query affect performance? (My understanding is it should not)
Table Record Count
T1 5000
T2 500000
T3 50
Need expert advise....
Regards,
Chinmay
Happy Working

Answer 1

Where did you get these questions from?
Q1. Performance.. is the same. Which one will yield a better throughput and why?
What is the difference between performance and throughput in this context? How in the world throughput could be better when "performance is the same"?
Performance cannot be the same even when Func1 and Func2 contain no more than exactly the same sub-queries as in the 1st SELECT. Functions are called for each row in EMP and each call leads to so called context switch between SQL executor and PL/SQL. This is expensive and may result in substantial increase of elapsed time given the number of rows in EMP is relatively large
On the other hand if the functions contain no more than dummy NULL statement then 2nd query most certainly would perform better
Q2. It depends. With correct statistics and CBO - no, not a chance. With RBO it depends on
1. What is your understanding of "their [tables] sequence of placement in WHERE clause". There are no tables in WHERE clause, there are predicates that mostly deal with table [pseudo]columns
2. Are there any indexes on the tables
3. Which expressions do you use in WHERE clause
4. etc..
Consider following:
DROP TABLE T1;
DROP TABLE T2;
DROP TABLE T3;
CREATE TABLE T1(C1 NUMBER CONSTRAINT T1_PK PRIMARY KEY);
CREATE TABLE T2(C2 NUMBER CONSTRAINT T2_PK PRIMARY KEY);
CREATE TABLE T3(C3 NUMBER CONSTRAINT T3_PK PRIMARY KEY);
INSERT INTO T1 (SELECT OBJECT_ID FROM ALL_OBJECTS);
INSERT INTO T2 (SELECT OBJECT_ID FROM ALL_OBJECTS WHERE ROWNUM <= 1);
INSERT INTO T3 (SELECT OBJECT_ID FROM ALL_OBJECTS WHERE ROWNUM <= 1000);
SET AUTOTRACE TRACEONLY
SELECT --+ RULE
          C1
     FROM T1, T2, T3
     WHERE C3 = C1
          AND C1 = C2
Execution Plan
   0      SELECT STATEMENT Optimizer=HINT: RULE
   1    0   NESTED LOOPS
   2    1     NESTED LOOPS
   3    2       TABLE ACCESS (FULL) OF 'T3'
4 2 INDEX (UNIQUE SCAN) OF 'T1_PK' (UNIQUE)
5 1 INDEX (UNIQUE SCAN) OF 'T2_PK' (UNIQUE)
Statistics
       1008 consistent gets
But
SELECT --+ RULE
          C1
     FROM T1, T2, T3
     WHERE C3 = C2
          AND C1 = C2
Execution Plan
   0      SELECT STATEMENT Optimizer=HINT: RULE
   1    0   NESTED LOOPS
   2    1     NESTED LOOPS
   3    2       TABLE ACCESS (FULL) OF 'T3'
4 2 INDEX (UNIQUE SCAN) OF 'T2_PK' (UNIQUE)
5 1 INDEX (UNIQUE SCAN) OF 'T1_PK' (UNIQUE)
Statistics
       8 consistent getsFortunately RBO is depricated and unsupported in 10g, so the answer should be NO.