I did add a null record, but I don't have a "TOTARGET NOT LOOKUP 1", I am still using the COUNT and using the alternate output as the mechanism to terminate the GATE that the records being fed to the primary input of LOOKUP flow through on their way there.
Look folks, I didn't mean to incite a religious war, I just needed a way to modify some records that didn't match the first time and then go look them up again. If you all think it would be safer/more-efficient to FANOUT the reference for the existing lookup and use it again as the reference for a second LOOKUP on the secondary output of the existing LOOKUP, I can do that (although I worry about how much storage that'll take, this is a seriously large chunk of data, more than half-a-million records). The only reason I thought of this approach was that I read the LOOKUP help and saw all the nice facilities for adding records to the reference on the fly and (naively) wondered if the same sort of thing might be doable for the detail records so that I didn't have to make a second copy of the reference. How about we forget whats gone on so far and start afresh. What's the very best way to do what I need to do? Can we get some sort of consensus from the experts? -- bc On Thu, Aug 6, 2009 at 5:56 PM, Rob van der Heij<[email protected]> wrote: > On Thu, Aug 6, 2009 at 11:20 PM, Glenn Knickerbocker<[email protected]> wrote: > >> Actually, because it delays the record, your solution has the same >> problem. The alternate output of TAKE LAST may be written anytime after >> the last record on the primary is released, just the same as the >> alternate output of COUNT. > > Sigh. This isn't trivial plumbing... so I guess we need a "pipeline > pig" append a null record and let it go through the lookup. That will > not match and be written to the secondary output where we can produce > the eof (with a "totarget not lookup 1") > > Rob >
