Re: How to trap the row count from a Commit ?

"Alkos" <azerty_at_nospam.org> wrote in message news:<bntof1$m9i3_at_news.rd.francetelecom.fr>...
> What about using a implicit cursor within a PL/SQL proc ?
> Cursors have a fancy property called %ROWCOUNT
>
> On the other hand, what about enlarging your RBS (if you can) ?
>
> "Richard Elliott" <richard.a.elliott_at_williams.com> a écrit dans le message
> news: 8c132b3c.0310310519.2963db99_at_posting.google.com...
> > I know the data is there somewhere. I'm running an update loop (5k
> > rows/commit) to avoid the rollback segment too small issue. It works
> > fine, but I have to do a count of the remaining rows to see if I need
> > to loop again. It would be much faster if I could just see if the
> > previous commit did 5k rows or not. How do I trap the "# of rows
> > affected" from the commit ?
> >
> > Thanks !

Let me add some more detail to explain why I ask the question and haven't considered the obvious. What I am doing is setting a flag in each row based on other values in the row. The table has 48 Million rows, and I expect to update about 5 million of them. No rollback segment is going to allow that to happen as one transaction. I am using a pl/sql procedure, ,but not a cursor. I don't think, (but I will try it), I can cursor up that many rows (5 million). The column is a new one and will be maintained by the ins/upd trigger going forward. Just need to get it initilized to start with. Thanks for all the feedback !

Thanks for all the Received on Fri Oct 31 2003 - 14:20:24 CST