ADQL Changes in Events Service 20.9.0

This page describes the ADQL-related changes caused by an update to the Events Service in version 20.9.0.

The following ADQL changes are currently only available for customers with accounts provisioned in the following regions: Bombay, and São Paulo. Contact your system administrator to confirm where your account is provisioned.

ADQL Percentile Query Results

In the Events Service 20.9.0, the ADQL percentile query now returns the larger number instead of the median number in a range of candidates. The ADQL percentile query syntax for the Events Service 20.9.0 has not changed.

Sample ADQL Percentile Query Syntax

ADQL Terms

Term	Description
`SELECT`	A `SELECT` keyword is the first keyword in an ADQL query. It is followed by a list of columns or expressions to query data in a table.
`Percentile`	The `percentile` keyword represents the similarly named ADQL function. It contains a field and the percentile value. In the sample ADQL percentile query syntax, the percentile value is the 50th percentile.
`Field`	A field is the column that is queried in a table. In the sample ADQL percentile query syntax, the field is the `age`.
`FROM`	A `FROM` keyword comes after the `SELECT` keyword and a list of fields or expressions.
`Table`	A table is the collection of fields and their values. In the sample ADQL percentile query syntax, `transactions` is the table that contains the `age` queried field.

ADQL Example

This example displays the 50th percentile age of employees working in the office. Based on the distribution, the 50th percentile age is 30 and 31. This is because out of a total of 130 employees in the table, 65 employees are 30 or younger and 65 employees are 31 or older.

Results

For Events Service < 20.9.0, the 50th percentile age value is 30.5, which is the median of 30 and 31.
For Events Service >= 20.9.0, the 50th percentile age value is 31, which is the larger candidate.

In a production environment, there can be millions of values in each field. As a result, the percentile query returning the larger number in a percentile bucket is a more accurate data distribution.