Thanks. Implemented aggregator to emit the tuple as such and used the same fields of the old stream as the output fields also.
On Thu, Jan 29, 2015 at 6:00 PM, Brunner, Bill <[email protected]> wrote: > An aggregator just collects all the tuples in a batch. You don’t even > have to do anything with them inside the aggregator. When the batch has > been fully aggregated, then it emits whatever you tell it to from the > complete() method. Sure, the output field name is different, but it’s > still the same tuple. > > > > *From:* Pandian R [mailto:[email protected]] > *Sent:* Thursday, January 29, 2015 4:21 AM > *To:* user; [email protected] > *Subject:* Re: startBatch and finishBatch for Trident functions > > > > Thanks. Tried using an Aggregator first, but using aggregator loses the > original fields in the tuple and forwards only the new fields to the output > stream. > > > > On Wed, Jan 28, 2015 at 8:44 PM, Nathan Marz <[email protected]> > wrote: > > You should use an Aggregator if you want to process many tuples within a > batch, or a State and a StateUpdater if you want to process many tuples > across batches. > > > > On Wed, Jan 28, 2015 at 9:55 AM, Pandian R <[email protected]> wrote: > > Hi, > > I am implementing a trident topology and using functions at many steps(as > alternative to traditional bolts). But, I don't find a startBatch and > finishBatch function. I need to do some operations per batch within a > function. Function has prepare and cleanup which is called per tuple. Can > someone help how to implement startBatch and finishBatch functions in a > Function ? > > Thanks ! > > > -- > > Regards, > > Pandian > > > > > > -- > > Twitter: @nathanmarz > > http://nathanmarz.com > > > > > -- > > Regards, > > Pandian > ------------------------------ > This message, and any attachments, is for the intended recipient(s) only, > may contain information that is privileged, confidential and/or proprietary > and subject to important terms and conditions available at > http://www.bankofamerica.com/emaildisclaimer. If you are not the intended > recipient, please delete this message. > -- Regards, Pandian
