ES|QL SAMPLE aggregation function #127629
Conversation
elasticsearchmachine
added
needs:triage
jan-elastic
force-pushed
the
esql-sample-agg-2
branch
from
55fe96f to
62de767
Compare
jan-elastic
added
>feature
:ml
elasticsearchmachine
removed
the
needs:triage
|
Pinging @elastic/ml-core (Team:ML) |
|
Hi @jan-elastic, I've created a changelog YAML for you. |
|
|
||
| public void add(int groupId, boolean value) { | ||
| try (BreakingBytesRefBuilder builder = new BreakingBytesRefBuilder(breaker, "sample")) { | ||
| ENCODER.encodeLong(new SplittableRandom().nextLong(), builder); |
There was a problem hiding this comment.
Is it more correct to put the new SplittableRandom into the ctor so we just keep calling nextLong on it?
Same for the BreakigBytesRefBuilder - you could clear it before every call here.
| "version" }, | ||
| description = "Collects sample values for a field.", | ||
| type = FunctionType.AGGREGATE, | ||
| examples = @Example(file = "stats_sample", tag = "doc") |
There was a problem hiding this comment.
I think we should have an example of the output in the docs. I'm not entirely sure the right way to hack that one up because it's non-deterministic. Maybe it's hand rolled.
I think we want that example because my first question when reading this is "can I get duplicates or do those count as distinct samples?" Mostly because I'm not good at statistics.
I do think it's interesting that SAMPLE(bool) is strictly more work than VALUES(bool). It feels like sampling shouldn't be, but it makes some sense.
There was a problem hiding this comment.
I think it makes sense that SAMPLE(bool) is more work. VALUES(bool) just keeps track of two boolean values: does true exist and does false exist. SAMPLE(bool) does more.
There was a problem hiding this comment.
Obv, I prefer some example output too. I didn't know how to achieve that, but I'll think of something. Should've left a TODO.
No description provided.