Page 1 of 1

How to handle nullable columns in a reject link

Posted: Tue May 24, 2011 2:41 pm
by bamaxplant
Do you have any suggestions on how to handle nullable columns when the sequential file is a reject link? In that case, you don't have the option to handle nulls when you edit the row for the column. We have a large number of jobs with this situation, and a large number of columns in most of the related transformers, so options like putting a transformer or modify between the rejecting stage and the reject SF would be a pretty huge effort.

We had a suggestion to change the link from a reject to an 'otherwise', but we do have some situations where we generate output rows for every input row (I'm coding on one right now), so it seem that wouldn't work; and the situation is really errors that should cause the row to reject, like an error in a derivation on one of the output links.

Thanks for any ideas.

Posted: Tue May 24, 2011 6:49 pm
by chulett
Un-hijacked.

Posted: Wed May 25, 2011 12:04 am
by singhald
I believe , when you add reject link to sequential stage, you cann't handle null as it takes entire record as a single field "Rejected" (var binary notnull) you dont need to handle null here.

let me know if that is not a case

Posted: Wed May 25, 2011 1:32 pm
by bamaxplant
Yes, that is the case for SF stages. Sorry I wasn't clear, these rejects are on transformer stages.

Posted: Wed May 25, 2011 11:36 pm
by chandra.shekhar@tcs.com
is your reject target a seq File?

Posted: Thu May 26, 2011 1:47 pm
by bamaxplant
yes - it is, but we could put something in between or use some other stage if it would help resolve this issue - we need to have rejects end up somewhere where we can look at them.

Posted: Thu May 26, 2011 2:17 pm
by bamaxplant
D'oh. Mea culpa - problem was changing the input link metadata without updating the reject link - previous input had nullable columns but new link did not, so all the columns on the reject remained nullable until I deleted and re-added the stage. Then it got the new metadata. :oops: