SQL & PL/SQL

I didn't want to hijack this thread: {thread:id=2321156} so I thought I'd start a new one.

Scenario

Update a table, based on complex business conditions.

Approach 1

Single, monolithic UPDATE statement with multiple conditions (pseudo-code)

UPDATE some_table
SET    some_col = some_val
WHERE  col_a = val_x
AND    EXISTS( -- complex correlated query )
AND    NOT EXISTS ( -- complex correlated query )
...
AND    EXISTS( -- complex correlated query )

Approach 2

Cursor FOR LOOP (pseudo-code)

DECLARE
	t BOOLEAN;
BEGIN
	FOR r IN ( SELECT *
		   FROM   some_table
		   WHERE  col_a = val_x
		 )
	LOOP
		t := check_complex_condition_a(r.pk);

		t := check_complex_condition_b(r.pk);

		t := check_complex_condition_c(r.pk);

		t := check_complex_condition_d(r.pk);

		IF t
		THEN
			UPDATE some_table
			SET    some_col = some_val
			WHERE  pk = r.pk
			;
		END IF;
	END LOOP;
END;

Assumptions

1. Lets suppose the database design is satisfactory (3NF or BCNF)
2. Lets suppose in this example that the FOR LOOP approach runs significantly faster (as in wall clock time) than the single UPDATE assignment.
3. Approach 1 and 2 are both implemented as optimally as possible.

Questions

After one has validated that the FOR LOOP is faster do you move on? Or is there another approach that you use to avoid the FOR LOOP approach? If so, what?

I know cursor FOR LOOPs are not to be spoken of around here ;) so I thought I'd bring up the question to see what you've all done in the past.

Edited by: Centinul on Dec 14, 2011 11:09 AM