Consider a hypothetical datasource with a dimension called "event_type" and various metrics (let's say 25 metrics).
Suppose a user runs a query on this datasource with a large number of filtered aggregators, two for each metric.
Half of these filtered aggregators share a filter on event_type = "A" and the other half share a filter on event_type = "B".
Because there is more than one filter condition, the filter cannot be pushed up to the query itself.
It could be a useful optimization if Druid can detect when filters are identical across multiple filtered aggregators and only apply these filters once per row, instead of once per aggregator.
Consider a hypothetical datasource with a dimension called "event_type" and various metrics (let's say 25 metrics).
Suppose a user runs a query on this datasource with a large number of filtered aggregators, two for each metric.
Half of these filtered aggregators share a filter on event_type = "A" and the other half share a filter on event_type = "B".
Because there is more than one filter condition, the filter cannot be pushed up to the query itself.
It could be a useful optimization if Druid can detect when filters are identical across multiple filtered aggregators and only apply these filters once per row, instead of once per aggregator.