---

{ “title”: “QUANTILE_UNION”, “language”: “en” }

Description

The QUANTILE_UNION function is used to merge multiple intermediate results of quantile calculations. This function is usually used together with QUANTILE_STATE, especially suitable for scenarios requiring multi-stage quantile calculation.

Syntax

QUANTILE_UNION(<query_state>)

Parameters

Parameter	Description
<query_state>	The data to be aggregated, type QuantileState supported.

Return Value

Returns an aggregation state for further quantile calculation, type QuantileState. Returns NULL if there is no valid data in the group.

Example

-- setup
CREATE TABLE response_times (
    request_id INT,
    response_time DOUBLE,
    region STRING
) DUPLICATE KEY(request_id)
DISTRIBUTED BY HASH(request_id) BUCKETS AUTO
PROPERTIES (
    "replication_allocation" = "tag.location.default: 1"
);
INSERT INTO response_times VALUES
(1, 10.5, 'east'),
(2, 15.2, 'east'),
(3, 20.1, 'west'),
(4, 25.8, 'east'),
(5, 30.3, 'west'),
(6, 35.7, 'east'),
(7, 40.2, 'west'),
(8, 45.9, 'east'),
(9, 50.4, 'west'),
(10, 100.6, 'east');

SELECT 
    region,
    QUANTILE_PERCENT(
        QUANTILE_UNION(
            TO_QUANTILE_STATE(response_time, 2048)
        ),
        0.5
    ) AS median_response_time
FROM response_times
GROUP BY region;

Calculate the 50th percentile of response times by region.

+--------+----------------------+
| region | median_response_time |
+--------+----------------------+
| west   |                35.25 |
| east   |                30.75 |
+--------+----------------------+

SELECT QUANTILE_UNION(TO_QUANTILE_STATE(response_time, 2048))
FROM response_times where response_time is null;

Returns NULL if there is no valid data in the group.

+--------------------------------------------------------+
| QUANTILE_UNION(TO_QUANTILE_STATE(response_time, 2048)) |
+--------------------------------------------------------+
| NULL                                                   |
+--------------------------------------------------------+