As we all know. Flink sql will be translated into streaming api. Look at a sql : SELECT color, sum(id) FROM T GROUP BY color.
In the actual calculation, Flink will store the whole t-stream into the state, and a piece of data in the stream will trigger a full stream calculation. Or does the status save only the calculation results and a new piece of data comes in the same group by key (color) the result can be added or subtracted. Is there a document or regular summary of the specific operation mechanism of Flink?thx.
