Pular para o conteúdo principal
Última atualização em

kll_sketch_get_rank_bigint function

Applies to: check marked yes Databricks Runtime 18.0 and later

Estimates the normalized rank (0.0 to 1.0) of a given value in an integer KLL sketch.

Syntax​

kll_sketch_get_rank_bigint ( sketch, value )

Arguments​

  • sketch: A BINARY expression containing a serialized integer KLL sketch.
  • value: A BIGINT expression or ARRAY<BIGINT> of values to find ranks for.

Returns​

  • If value is BIGINT: returns a DOUBLE between 0.0 and 1.0 representing the normalized rank.
  • If value is ARRAY<BIGINT>: returns ARRAY<DOUBLE> with ranks for each value.

Notes​

  • The rank represents the fraction of values in the sketch that are less than or equal to the given value.
  • Returns 0.0 if all sketch values are greater than the input value.
  • Returns 1.0 if all sketch values are less than or equal to the input value.

Examples​

SQL
-- Find what percentile a value falls into
> WITH sketch_data AS (
    SELECT kll_sketch_agg_bigint(value) AS sketch
    FROM VALUES (1), (2), (3), (4), (5) AS T(value)
  )
  SELECT kll_sketch_get_rank_bigint(sketch, 3) FROM sketch_data
  0.6

-- Check ranks for multiple values
> WITH sketch_data AS (
    SELECT kll_sketch_agg_bigint(value) AS sketch
    FROM VALUES (10), (20), (30), (40), (50) AS T(value)
  )
  SELECT kll_sketch_get_rank_bigint(sketch, array(15, 25, 35)) FROM sketch_data
  [0.2, 0.4, 0.6]

-- SLA compliance: what fraction of requests are under threshold
> WITH sketch_data AS (
    SELECT kll_sketch_agg_bigint(response_time_ms) AS sketch FROM requests
  )
  SELECT kll_sketch_get_rank_bigint(sketch, 100) AS under_100ms_fraction FROM sketch_data
  0.87