Using TABLE() to improve performance... Am I on the right track?

Question

I have a situation where I read data into a set of collections (let's assume 10,000 records and an emp_no collection).
I then process each record in a for loop. Based on conditions, a subsequent query is issued to one of two tables:
For i in emp_no.first .. emp_no.last loop
<<processing>>
if <<some condition>> then
select emp_age into age from tab_a where employee_number=emp_no(i);
else
select spouse_age into age from tab_b
where employee_number=emp_no(i) and {other conditions};
end if;
age_array(i) := age;
<<processing>>
end loop;
after the additional fields are retrieved, processing continues using the retrived data.
<<additional processing>>
At the end of the processing I want to update a table's records given the values calculated during processing
ForAll i in emp_no.first .. emp_no.last
Update retirement Set age := age_array(i) ......
where employee_number = emp_no(i);
I imagine the single select queries in the loop structure will cause a lot of context switches between PL/SQL and SQL which will significantly decrease performance.
After some review of the Oraclewebsite I found the TABLE function. It appears I can use this to change my routine to a more efficient bulk processing structure. Something like:
-- In the loop build a collection of emp_no's associated to each query
For i in emp_no.first .. emp_no.last loop
<<processing>>
if emp_no(i) is even then
tab_a_emp_no_array.extend;
tab_a_emp_no_array(tab_a_emp_no_array.last) := emp_no(i);
else
tab_b_emp_no_array.extend;
tab_b_emp_no_array(tab_a_emp_no_array.last) := emp_no(i);
end if;
<<processing>>
end loop;
--After the loop use a Select... Bulk Collect Into statement with a where condition that references the collection values
Select emp_no, emp_age
bulk collect into emp_no_a, age_a
from tab_a
where employee_number in (select column_value from table(tab_a_emp_no_array));
Select emp_no, spouse_age
bulk collect into emp_no_b, age_b
from tab_b
where employee_number in (select column_value from table(tab_b_emp_no_array));
Using the emp_no_a and emp_no_b the age values can be reassociated with the correct employee for further processing.
I HAVE THREE CONCERNS:
1. Am I understanding and using the TABLE function correctly? I don't think "pipelined processing" would help in this situation, correct?
2. I may end up with an IN clause that has thousands of elements. Will this perform poorly and eliminate any performance gains obtained from the bulk collect? Would "where exists (select 1 from table(tab_a_emp_no_array) where column_value=employee_number)" work any better?
3. Is there a better way to solve this issue of optimizing performance when various tables are conditionally queried during a loop?
I hope my issue is clear (obviously the code isn't accurate) and I thank you in advance for any insights!
Peace,
Larry

Answer 1

No.
I will repeat one of Tom Kytes' mantras here
1 when you can do it in 1 SQL statement, you should do it in SQL
2 When you can not do it in SQL, you should do it in PL/SQL
3 When you can not do in in PL/SQL, you should do it in Java
Which means: You should things non-procedurally as often as possible. Quite often people resort too early to 3GL strategies.
update inside a loop raises a red flag, especially if there would have been a commit inside this loop. This means you are not only into slow-by-slow prtogramming, but also increases the possibility of ora-15555 errors.
Sybrand Bakker
Senior Oracle DBA