Jackie-Jiang commented on issue #5259: Support Aggregation functions with multiple arguments. URL: https://github.com/apache/incubator-pinot/pull/5259#issuecomment-614832081 > > Recommend directly passing TransformBlock to the AggregationFunction, where AggregationFunction can get the column needed, instead of fetching the data on caller side and creating the map. > > Yeah, I thought about that as well. I feel having a map passes only the relevant content and makes the api more crisp and unit testable, and hence chose to create a Map. @mayankshriv Passing TransformBlock is still unit testable, where you just need to mock the TransformBlock. Here what I'm actually proposing is to wrap the column fetch logic inside aggregation function. Only the aggregation function knows how each column should be processed, and in what order, either the column is identifier or constant etc. Another concern is about the overhead and extra garbage of creating a map. For very high QPS use case, this might cause higher latency.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
