Update cursor is faster or normal update in a cursor

Question

Hi
oracle 10.2.0.1.0
Is update cursor faster or an update statement within a cursor ? Also , owing to fact that update cursor locks the rows unlike a norma update statement.
Regards
VK

Answer 1

804282 wrote:
ok. I need to update rows with in pl/ block only ( with in cursors l)If you can do that using a single SQL statement (cursor), so much better.
If you code an explicit cursor for updating, and use a loop fetch a row, check it, and then conditionally update it - you are generating a lot of work and I/O.
Let's say there are a 1000 rows output by that cursor. This means a 1000 loop iterations for fetching (ignoring PL/SQL internal optimisation). It means executing that condition in PL/SQL a 1000 times. If 500 of these rows match the condition, then it means executing 500 SQL update statements.
If the cursor in future returns 5000 rows, what happens to scalability and performance? You now have 5x the number of loop iterations! And how many SQL update commands will be fired off in that loop?
This approach simply does not scale.
The optimal approach is to create a single SQL update statement - add the conditional logic code in PL/SQL as predicates to the SQL statement. And then execute a single update cursor for updating the contents of that table. And not 100's or even 1000's of individual single-row update statements as the case is when using a PL/SQL loop.
SQL is always faster than PL/SQL. The less PL/SQL code and the more SQL code you use to crunch data, the faster it will be.
The rule is quite basic and simple: Maximise SQL and Minimise PL/SQL.