On Tue, Jul 14, 2009 at 11:22 PM, Glenn Knickerbocker<[email protected]> wrote:
> --and you're still stuck buffering at least all the unmatched records, > and fighting with input 3 of LOOKUP to delete the matched ones. But you still don't have the buffer the full input. Be aware that deleting the keys from lookup does not save you any storage. It only wipes out the master so that you can put a new master in with the same key (like I needed beause RSCS will re-use the spool file numbers eventually). If you want to free up memory for deleted keys, you will have to wash out the lookup stage. I think the cache in my Lookup paper does that. > Sure would be convenient to have an AUTODELete keyword like AUTOADD, to > delete master records once they're matched rather than having to figure > out the right combination of COPY and LOOKUP STRICT to feed them back > into input 3 without danger of stalling. In this case there wouldn't be > anything on input 2 to worry about, but in the more general case I'm not > sure how I would ensure that a matched record was fed back fast enough > to avoid deleting a newly added master record that happened to match it. The ONCE option (delete after matching) is certainly on my wish list, and there have been times where I wanted a REPLACE option to make duplicate masters delete the previous entry. And maybe a DELWHEN0 option to delete the master when the count goes to 0. But it gets hard to decide what other actions to associate with it (eg do we recent the counter). When you have to do it yourself, you make those decisions.
