Skip to main content
Last updated on

kll_sketch_agg_float aggregate function

Applies to: check marked yes Databricks Runtime 18.0 and later

Creates a KLL (K-Linear-Logarithmic) sketch for approximate quantile estimation on single-precision floating-point data with configurable accuracy.

Syntax​

kll_sketch_agg_float ( expr [, k] )

Arguments​

  • expr: A FLOAT expression to aggregate.
  • k: An optional INTEGER literal controlling sketch accuracy. Must be between 8 and 65535. The default is 200. Higher values provide better accuracy but use more memory.

Returns​

A BINARY value containing the serialized KLL sketch for FLOAT data.

Notes​

  • NULL values in expr are ignored during aggregation.
  • NaN values are handled according to IEEE 754 semantics.
  • The sketch provides approximate quantiles with a confidence level of about 99%.

Examples​

SQL
-- Create sketch from float values
> SELECT kll_sketch_agg_float(score) FROM VALUES (1.5), (2.3), (3.7) AS T(score)
  [binary data]