[Lens] Field stats endpoint does not need to use sampler aggregation

[wylieconlon]

Originally, we thought that the sampler aggregation would behave like a random sampling query, with improved performance across large datasets. This is not what the sampler aggregation actually does, which means that we are doing more work instead of less. This aggregation can be removed entirely.

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[flash1293]

@wylieconlon Are you sure not using sampler is better than using it in cases where it really matters (super large data sets)?

From the documentation you linked:

Example use cases

• Reducing the running cost of aggregations that can produce useful results using only samples e.g. significant_terms

Probably missing a nuance here.

[wylieconlon]

@flash1293 Because we don't use any of those aggregations when calculating the samples, that part is not relevant. The other example use case is potentially relevant, with caveats:

Tightening the focus of analytics to high-relevance matches rather than the potentially very long tail of low-quality matches

This part actually might be relevant if the user has added a rank-affecting query to the Lens editor before clicking on the preview. Exact match queries wouldn't have any effect, but a query with OR would affect the rank, as would a wildcard query.

Because it's such a narrow subset of queries that affect the results, I would say that the sampler is not useful.

[flash1293]

Ah, I forgot we aren't using aggregations for gathering the stats - makes total sense in that case, thanks.

[stratoula]

We have changed the implementation so this is not valid anymore