Help Center> Data Lake Insight> Spark SQL Syntax Reference> Built-In Functions> Aggregate Functions> percentile_approx
Updated on 2023-10-25 GMT+08:00
View PDF

percentile_approx

This function is used to approximate the pth percentile (including floating-point numbers) of a numeric column within a group.

Syntax

percentile_approx(DOUBLE col, p [, B])

Parameters

Table 1 Parameters

Parameter

Mandatory

Description

col

Yes

Columns with a data type of numeric. If the values are of any other type, NULL is returned.

p

Yes

The value should be between 0 and 1. Otherwise, NULL is returned.

B

Yes

The parameter B controls the accuracy of the approximation, with a higher value of B resulting in a higher level of approximation. The default value is 10000. If the number of non-repeating values in the column is less than B, an exact percentile is returned.

Return Values

The return value is of the DOUBLE type.

Example Code

  • Calculates the 0.5 percentile of all offering inventories (items), with an accuracy of 100. An example command is as follows:
    select stddev_samp(items,0.5, 100) from warehouse;

    The command output is as follows:

    _c0        
500
  • When used with group by, it groups all offerings by warehouse (warehourseId) and returns the 0.5 percentile of the offering inventory (items) in the same group, with an accuracy of 100. An example command is as follows:
    select warehourseId, stddev_samp(items, 0.5, 100) from warehourse group by warehourseId;

    The command output is as follows:

    warehouseId _c1        
city1     499        
city2     354        
city3     565
Parent topic: Aggregate Functions