Apache Pinot

quick question: Does pinot support custom aggregation functions as UDFs? For example, when I'm aggregating results of a group, can I pick the latest record by timestamp?

so you want to pick the last msg per group or you want to do aggregation on that

the first part is already an aggregation function technically

pinot doesn’t support aggregation on top of aggregation for now

Example: SUM(col) is summing up all values in the group, LATEST(col) is just picking the latest value in that group

yes, LATEST is an aggregation. There is no aggregation of aggregations here

right, then query should be  something like `SELECT last(ordered_col, columns...) FROM myTable group by a`

I'm thinking like this:
`select id, latest(metric) FROM myTable group by id`
latest is an aggregation on only one column, i'm not thinking across columns, if that's what you're thinking

right, i was just thinking how to order those events

hmm... aren't these values already ordered based on timestamp in the raw segment?

<@UGRJA9TEH> what are the classes or modules in the source code to look at? I think `latest()` aggregation will be pretty useful and will unlock a bunch of use cases, at least for our open source project Hypertrace, where we are using Pinot.

it's ordered in one segment, but  we need to merge from multiple segments and from multiple servers :slightly_smiling_face:

I think we need to carry over the timestamp with it

you can check `org.apache.pinot.core.query.aggregation.function.AvgAggregationFunction`

It has the definition of intermediate results