Dynamic pl sql cursor

Question

Hi.
I've written a package with the intent of updating a number of fairly large tables.
I'm using Oracle : 10.2.0
the package has a main function called to update a set of tables.
It loops through a table containing table names and column names (one on each row).
It calls a second function for each (tablename,columnname) to ask it to update all the values in that table.column.
This, in turn, calls my third 'calculate' function to recalculate the input table.column.
It is running really slowly, as I wrote it in a 'find out how to do it' mode.
The tables I am updating can have hundreds of thousands of rows.
I reviewed some examples which recommended using bulk collect and forall to efficiently update a table.
My question - it isn't clear if I can do this if the update of a field is accomplished by calling my third pl/sql function which
resides in the package I am working in.
summary:
have a table with 3 columns: tablename, columnname and an 'in use' column (not previously mentioned).
I loop through this table, and with each tablename/columname, I feed the data into a second function.
this second function updates the table via a cursor, invoking a third function that calculates the new value.
it is so slow.
Is bulk collection/forall a workable approach to repopulating such a set of tables. If so I'll go pursue the details.
If not, what is a recommended approach.
I can post some code but it is just gross.
Thanks
Jeff

Answer 1

Thanks for the suggestions.
This is a one time system modification, part of a larger reorganization. Because the rest of the reorganization is
going to take alot of time, I need to minimize the time this update process takes.
My requirement is that I update a number of columns in a number of tables - which tables / columns are due
to be updated are determined via a complex analysis.
I already have finished the analysis and code that figures out what tables and columns need updating.
This code populates a table that contains a column for tablename and a column for columname.
Each row in this table represent a table and column that need recalculating.
So now at runtime, I need to iterate through this table, and get the target table/column which required updating.
For each iteration I need to 'open' the table, read all the rows and update the target column.
The update is a fairly complex piece of parsing and analysis, so it is encapsulated in a function in my package.
Therefore, my quest was to figure out if bulk collect was a reasonable/possible way to update the tables at runtime.
I saw this code on the internet at
{noformat}
http://www.dba-oracle.com/plsql/t_plsql_dynamic.htm
{noformat}
as an example, and noticed that I seem to need to know the target table in a 'static' manner,
i.e. the bulk collect into 'parameter' which is of a type declared for a given table
so I can have a rowtype variable to stuff the data into from the bulk collect.
DECLARE
  TYPE t_object_id_tab IS TABLE OF bulk_collect_test.object_id%TYPE;
  l_tab   t_object_id_tab;
BEGIN
  -- Populate collection use in forall.
  SELECT object_id
  BULK COLLECT INTO l_tab
  FROM   bulk_collect_test
  WHERE  rownum < 101;
  FORALL i IN l_tab.first .. l_tab.last
    EXECUTE IMMEDIATE
      'UPDATE bulk_collect_test
       SET    object_id = object_id
       WHERE  object_id = :1'
       USING l_tab(i);
END;
/jeff