Hi, I’m using the table/sql api and I’m trying to preserve the time attribute across a group by sum operation. In other words I want to emit a row for every row going into the sum, and I want to be able to preserve the event time of the row that went into the sum as a column. Is there a way to do this? I basically want a changelog stream of the sum over time, with the time of the update being the time the most recent event came in.

non-windowed aggregations like group-by don't maintain the time attribute (due to the fact that watermarks can no longer be guaranteed to be meaningful). Could window aggregation help with your use case? Row time attributes can be preserved using these functions.

Maybe, my use case is to have running all-time counts (since the beginning of the data) I wasn’t sure if that was possible with the window functions/whether that would cause there to be many windows kept open?