Re: "Fast" Query of Multiple Columns?

Is the 'locator translation' file large or small compared to the 'updateable' table?

If it is relatively small (and has some hope of fitting into your buffer cache), then it's probably best to:

1. load the locator translator file into a table with an index on the old locator number
2. use PL/SQL to do a SINGLE pass (table scan) through the updateable table, updating each row based on the contents of each of the three old locator number columns
3. do commits after every n rows so as not to blow your rollback space into another galaxy (n should be chosen based on size of rows, size of rollback available, other concurrent jobs, etc)

You should position step 2 to make the job restartable - i.e. only select rows with a null "new" locator.

If the 'locator translation' file is relatively large, then you may be better off to reverse the process
- i.e. scan the 'locator translation' file and update the table as you go.

Steve Received on Wed Oct 17 2001 - 10:41:30 CDT