Benchmark: Use arrays to store json properties

[macobo]

Changes

https://eng.uber.com/logging/ had an alternative way of storing properties - as parallel arrays.

This PR (not to be merged) benchmarks how this performs.

Data was populated via the following queries on the benchmark server:

ALTER TABLE events ADD COLUMN property_keys Array(String) DEFAULT arrayMap(x -> x.1, JSONExtractKeysAndValues(properties, 'String'));
ALTER TABLE events ADD COLUMN property_values Array(String) DEFAULT arrayMap(x -> x.2, JSONExtractKeysAndValues(properties, 'String'));
ALTER TABLE events UPDATE property_values = property_values, property_keys = property_keys WHERE 1=1;

ALTER TABLE person ADD COLUMN property_keys Array(String) DEFAULT arrayMap(x -> x.1, JSONExtractKeysAndValues(properties, 'String'));
ALTER TABLE person ADD COLUMN property_values Array(String) DEFAULT arrayMap(x -> x.2, JSONExtractKeysAndValues(properties, 'String'));
ALTER TABLE person UPDATE property_values = property_values, property_keys = property_keys WHERE 1=1;

[github-actions]

ClickHouse query benchmark results from GitHub Actions

Lower numbers are good, higher numbers are bad. A ratio less than 1
means a speed up and greater than 1 means a slowdown. Green lines
beginning with + are slowdowns (the PR is slower then master or
master is slower than the previous release). Red lines beginning
with - are speedups. Blank means no changes.

Significantly changed benchmark results (PR vs master)

       before           after         ratio
     [3d8bcdfa]       [6bf09777]
-       4331.0±61        3578.0±64     0.83  benchmarks.QuerySuite.track_session_recordings_list_person_property_filter

Click to view full benchmark results

All benchmarks:

     before           after         ratio
   [3d8bcdfa]       [6bf09777]
37481.5±6.6e+03  30301.5±1.6e+03    ~0.81  benchmarks.QuerySuite.track_correlations_by_event_properties
  12499.5±2e+02    14054.5±2e+02     1.12  benchmarks.QuerySuite.track_correlations_by_event_properties_materialized
10469.5±1.2e+02  11274.0±1.2e+02     1.08  benchmarks.QuerySuite.track_correlations_by_events
      5544.0±73        5872.5±90     1.06  benchmarks.QuerySuite.track_correlations_by_properties
      4835.5±41        5240.0±38     1.08  benchmarks.QuerySuite.track_correlations_by_properties_materialized
  764.0±1.6e+02    874.0±1.7e+02     1.14  benchmarks.QuerySuite.track_earliest_timestamp
      3150.0±95   3183.5±1.1e+02     1.01  benchmarks.QuerySuite.track_funnel_normal
28612.5±3.1e+02  30106.5±1.6e+03     1.05  benchmarks.QuerySuite.track_lifecycle
14024.5±1.7e+02       14529.5±72     1.04  benchmarks.QuerySuite.track_retention
  18254.0±2e+02       16888.5±66     0.93  benchmarks.QuerySuite.track_retention_filter_by_person_property
16493.0±1.5e+03  16951.0±1.7e+02     1.03  benchmarks.QuerySuite.track_retention_filter_by_person_property_materialized
54847.5±7.1e+02  53257.0±3.1e+02     0.97  benchmarks.QuerySuite.track_retention_with_person_breakdown
  789.0±2.2e+02    719.5±1.4e+02     0.91  benchmarks.QuerySuite.track_session_recordings_list
 4202.5±1.4e+02   3217.0±1.2e+02    ~0.77  benchmarks.QuerySuite.track_session_recordings_list_event_filter
-       4331.0±61        3578.0±64     0.83  benchmarks.QuerySuite.track_session_recordings_list_person_property_filter
      3937.0±20        3925.0±22     1.00  benchmarks.QuerySuite.track_stickiness
      5586.0±35        4743.5±18     0.85  benchmarks.QuerySuite.track_stickiness_filter_by_person_property
      4731.0±32        4690.5±57     0.99  benchmarks.QuerySuite.track_stickiness_filter_by_person_property_materialized
      4318.0±32        4320.5±45     1.00  benchmarks.QuerySuite.track_trends_dau
      5949.5±81        5263.0±69     0.88  benchmarks.QuerySuite.track_trends_dau_person_property_filter
      5105.0±40        5241.0±27     1.03  benchmarks.QuerySuite.track_trends_dau_person_property_filter_materialized
20359.5±7.6e+02  18810.0±3.2e+02     0.92  benchmarks.QuerySuite.track_trends_event_property_filter
      4208.5±71        4127.0±70     0.98  benchmarks.QuerySuite.track_trends_event_property_filter_materialized
  41310.0±2e+03    32029.0±6e+02    ~0.78  benchmarks.QuerySuite.track_trends_filter_by_action_current_url
 4305.5±1.4e+02   4431.0±1.2e+02     1.03  benchmarks.QuerySuite.track_trends_filter_by_action_current_url_materialized
10089.0±2.9e+02       10858.0±96     1.08  benchmarks.QuerySuite.track_trends_filter_by_action_with_person_filters
      5060.0±18        5278.5±61     1.04  benchmarks.QuerySuite.track_trends_filter_by_action_with_person_filters_materialized
      5737.0±43        5202.5±14     0.91  benchmarks.QuerySuite.track_trends_filter_by_cohort
      4956.0±41        5262.5±16     1.06  benchmarks.QuerySuite.track_trends_filter_by_cohort_materialized
 4445.5±1.1e+02   5954.0±1.4e+03    ~1.34  benchmarks.QuerySuite.track_trends_filter_by_cohort_precalculated
      1979.0±36       2075.5±3.5     1.05  benchmarks.QuerySuite.track_trends_no_filter
 5607.5±1.1e+02   5258.5±1.3e+02     0.94  benchmarks.QuerySuite.track_trends_person_property_filter
      4872.0±77        5246.5±64     1.08  benchmarks.QuerySuite.track_trends_person_property_filter_materialized

[macobo]

This was opened for analysis purposes only - see linked issue for details/conclusion!