I see your point Andrey, and I agree that having an opportunity not to bring results back, but process them in place is cool.
Sergi, can you please file a ticket and put your design suggestions to it? --Yakov 2015-12-15 20:51 GMT+03:00 Andrey Kornev <[email protected]>: > Or, alternatively it would be nice to have a way submitting a query from a > single node. But instead of bringing the results back, the user would be > able to specify a closure that would be sent along with the query to the > nodes where the query will be executed. The closure would then be invoked > by the query executor for each matching entry locally. With this approach > the user would not have to worry about the state of partitions since Ignite > would hopefully be able to take care of that (as it presumably already does > while executing the SQL query). > > I believe the good ol' GridGain product used to have an API like that, but > then, in a misguided effort to dumb down the APIs, this useful feature was > dropped. > > Thanks > Andrey > > ------------------------------ > Date: Tue, 15 Dec 2015 20:26:39 +0300 > Subject: Re: Computation on NodeEntries > From: [email protected] > To: [email protected] > > Ah, now I see your point. It looks like we need to add an ability to run > an SQL query against an individual partition - this would have worked for > your use-case the same way a Scan query works. > > I wonder of anybody in the community with a deeper knowledge of query > processing can estimate the complexity of this feature and create a ticket > with a proper description. > >
