How to use a collection type of bind variable for execute dynamic statement

Question

Hi,
We have a case where we copy selective data from Schema A To Schema B in one oracle database. The copy is achieved by using
execute immediate 'insert into '||target_schema||'.tablea select * from '||from_schema||'.table a where a.id in (select test_id from '||from_schema||'.table c);';
This works fine it takes an average of 10 seconds to copy around 14 tables. We have a requirement to bring this time to 2 seconds. One observation has been the clause
"select test_id from '||from_schema||'.table c" in the above sql statement repeats for many inserts . Thus we were thinking to bulk fetch this set of tests ids and use a bind vatiable of collection type for the execute immediate clause. Any suggestions on how to achieve it?
Thanks,
Chandana

Answer 1

>
One observation has been the clause
"select test_id from '||from_schema||'.table c" in the above sql statement repeats for many inserts
>
So what? Constructing a string for a table level insert and parsing it can't possibly be a performance problem. If you were creating a string in a loop to insert rows into a table by getting the data FROM a collection - that's a problem that keeps showing up in the forums.
I'm with bravid and Nikolay on this one. First find out which side, select/insert, the problem is on.
As they said you need to provide more information about the process.
And using collections for your use case is definitely not the thing to do.
1. How many rows are we talking about?
2. Are the rows being inserted into an empty table?
3. Are you running these queries during peak production hours or in a batch windows?
Tune the SELECT if the problem is on that side.
Post an execution plan for the SELECT part of a query you think should run faster.
SET SERVEROUTPUT ON
SET AUTOTRACE TRACEONLY
SQL> select * from emp;
Execution Plan
Plan hash value: 3956160932
| Id  | Operation         | Name | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT  |      |    14 |   546 |     3   (0)| 00:00:01 |
|   1 |  TABLE ACCESS FULL| EMP  |    14 |   546 |     3   (0)| 00:00:01 |
SQL>On the INSERT side you might be able to use bulk inserts
ALTER TABLE myTable NOLOGGING;
INSERT /*+ append */ INTO myTable . . .
ALTER TABLE myTable LOGGINGIf the insert tables are always reused you could just leave them as NOLOGGING. Since you can't recover bulk data you need to make sure to get a good backup after the loads if you need to recover the data from the logs rather than just reload it yourself.