When writing TQL, customers almost always want to aggregate either over the number of signals or the number of users in the query. Yet, both of these things require multiple lines of aggregation that are unintuitive:
For counting signals, customers have to sum up the count field:
Both of these are very unintuitive if you don't know how our Apache Druid backend works, and cumbersome even if you do. I'm proposing a new aggregation type that abstracts these away:
When writing TQL, customers almost always want to aggregate either over the number of signals or the number of users in the query. Yet, both of these things require multiple lines of aggregation that are unintuitive:
For counting signals, customers have to sum up the
count
field:For counting users, customers have to create a thetaSketch object over the userID field:
Both of these are very unintuitive if you don't know how our Apache Druid backend works, and cumbersome even if you do. I'm proposing a new aggregation type that abstracts these away:
and
These would then get compiled into the
longSum
andthetaSketch
aggregators respectively.