[Lens] improve percentile agg optimizations

[drewdaemon]

Summary

Part of #135265

Unfiltered single-percentile agg configs were already optimized.

As of this PR, filtered single-percentile aggs that use the same percentile (and other args) are now collapsed into one agg config.

Unfortunately, esaggs doesn't currently support combining filtered single-percentile aggs with different percentiles into a single agg config, so that special optimization is currently only applied on non-filtered single-percentile aggs.

Testing

Filtered percentiles with same value now deduplicated

Add a dimension with the following formula

percentile(bytes, percentile=2, kql='geo.dest : "AL" ') + 
percentile(bytes, percentile=2, kql='geo.dest : "AL" ') + 
percentile(bytes, percentile=2, kql='geo.dest : "BA" ') + 
percentile(bytes, percentile=2, kql='geo.dest : "BA" ') + 
percentile(bytes, percentile=2, kql='geo.dest : "BA" ')

Check the request. Should only have two aggs, one for each filter.

Unfiltered percentiles with different values still collapsed

Add a dimension with the following formula

percentile(bytes, percentile=2) + 
percentile(bytes, percentile=3) +
percentile(bytes, percentile=4) +
percentile(bytes, percentile=4) +
percentile(bytes, percentile=5)

Check the request. Agg should look like this

"0": {
      "percentiles": {
        "field": "bytes",
        "percents": [
          2,
          3,
          4,
          5,
        ],
        "keyed": false
      }
    }

[drewdaemon]

@elasticmachine merge upstream

[elasticmachine]

Pinging @elastic/kibana-visualizations @elastic/kibana-visualizations-external (Team:Visualizations)

[drewdaemon]

@elasticmachine merge upstream

[drewdaemon]

This change prevents us wrapping this agg in a filter when the query is just an empty string. No difference on the elasticsearch query, but this does make sure that if someone removes a kql filter from a dimension, that agg can be collapsed into other matching aggs instead of being passed over because it looks like it has a query.

[kibana-ci]

💚 Build Succeeded

• Buildkite Build
• Commit: 04407c4

Metrics [docs]

Async chunks

Total size of all lazy-loaded chunks that will be downloaded as the user navigates the app

id	before	after	diff
lens	1.3MB	1.3MB	-239.0B

Unknown metric groups

ESLint disabled in files

id	before	after	diff
osquery	1	2	+1

ESLint disabled line counts

id	before	after	diff
enterpriseSearch	19	21	+2
fleet	59	65	+6
osquery	109	115	+6
securitySolution	442	448	+6
total			+20

Total ESLint disabled count

id	before	after	diff
enterpriseSearch	20	22	+2
fleet	68	74	+6
osquery	110	117	+7
securitySolution	519	525	+6
total			+21

History

• 💚 Build #90155 succeeded 1a7742d
• 💚 Build #89256 succeeded b465b68

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream

[dej611]

Tested locally on Safari 👍

I had a go also with a mixed formula like:

defaults(percentile(bytes, percentile=2, kql='extension.keyword : CSS'), 0) + 
defaults(percentile(bytes, percentile=2, kql='extension.keyword : CSS'), 0) +
percentile(bytes, percentile=4) +
percentile(bytes, percentile=4) +
percentile(bytes, percentile=5)

And both 4th and 2nd percentile were aggregated into each own bucket.
Adding an extra percentile(bytes, percentile=2) will add the 2nd percentile metric into the non-filtered bucket correctly and group the other similar ones into the filtered bucket.