PERCENTILE_APPROX_ARRAY

Description

PERCENTILE_APPROX_ARRAY calculates multiple approximate percentiles in one aggregation. It builds one T-Digest for the input values and evaluates every requested percentile from that shared state, which is more efficient than calling PERCENTILE_APPROX separately for each percentile.

note

This function is supported since 4.2.0.

Syntax

PERCENTILE_APPROX_ARRAY(<col>, <quantiles> [, <compression>])

Parameters

Parameter	Description
<col>	The numeric expression whose percentiles are calculated. Values are converted to DOUBLE. NULL values are ignored.
<quantiles>	A constant ARRAY<DOUBLE>. Each element must be finite, non-NULL, and in [0.0, 1.0]. Results follow the same order as the array, and duplicate levels are preserved. An empty array is allowed.
<compression>	Optional constant DOUBLE in [2048, 10000]. A larger value generally improves accuracy but uses more memory. The default is 10000; non-finite values and values outside the valid range also use 10000.

Return Value

Returns an ARRAY<DOUBLE> containing one approximate percentile for each element of <quantiles>. The function returns an empty array when <quantiles> is empty or when the group contains no non-NULL input values. Invalid percentile levels and NULL elements in <quantiles> produce an error.

Example

CREATE TABLE percentile_approx_array_example (
    id INT,
    value DOUBLE NULL
) DUPLICATE KEY(id)
DISTRIBUTED BY HASH(id) BUCKETS 1
PROPERTIES ("replication_num" = "1");

INSERT INTO percentile_approx_array_example VALUES
    (1, 1.0), (2, 2.0), (3, 3.0), (4, NULL),
    (5, 10.0), (6, 20.0), (7, 30.0), (8, 100.0);

Calculate several approximate percentiles with the default compression:

SELECT percentile_approx_array(value, [0.0, 0.25, 0.5, 0.75, 1.0]) AS percentiles
FROM percentile_approx_array_example;

+--------------------------+
| percentiles              |
+--------------------------+
| [1, 2.25, 10, 27.5, 100] |
+--------------------------+

Specify the compression explicitly:

SELECT percentile_approx_array(value, [0.25, 0.5, 0.75], 2048) AS percentiles
FROM percentile_approx_array_example;

+------------------+
| percentiles      |
+------------------+
| [2.25, 10, 27.5] |
+------------------+

All-NULL input returns an empty array:

SELECT percentile_approx_array(CAST(NULL AS DOUBLE), [0.0, 0.5, 1.0]) AS percentiles
FROM percentile_approx_array_example;

+-------------+
| percentiles |
+-------------+
| []          |
+-------------+

An empty percentile array also returns an empty array:

SELECT percentile_approx_array(value, []) AS percentiles
FROM percentile_approx_array_example;

+-------------+
| percentiles |
+-------------+
| []          |
+-------------+