Window

Function Name	Description
COUNT	Returns the total number of records for the specified expression.
COVAR_POP	Returns the population covariance for non-NULL pairs across all input values.
COVAR_SAMP	Returns the sample covariance for non-NULL pairs across all input values.
CUME_DIST	Returns the cumulative distribution of the current row with regard to other values within the same window partition.
DENSE_RANK	Returns the rank of the current row within its partition and ordering. Rows that are equal will have the same rank.
FIRST_VALUE	Returns the first value within an ordered group of a result set.
HLL	Uses HyperLogLog to return an approximation of the distinct cardinality of the input.
LAG	Returns the row before the current one in a partition based on the ORDER BY clause without the need for a self-join. If there are no rows, this function returns NULL.
LEAD	Returns the row after the current one in the same result set without the need for a self-join. If there are no rows, this function returns NULL.
MAX	Returns the maximum value among the non-NULL input expressions.
MIN	Returns the minimum value among the non-NULL input expressions.
NDV	Returns an approximate distinct value number, similar to COUNT(DISTINCT col). NDV can return results faster than using the combination of COUNT and DISTINCT while using a constant amount of memory, resulting in less memory usage for columns with high cardinality.
NTILE	Equally splits the rows in each partition into ranked parts specified by the integer value and starting from 1. This function requires the ORDER BY clause.
PERCENT_RANK	Returns the relative rank of the current row in the partition based on the ORDER BY clause. The displayed percentage ranges from 0.0 to 1.0.
RANK	Returns the rank of the current row within its partition and placement order. Rows that are equal have the same rank. However, the count of tied rows is added to the next rank, instead of being incremented by one. The rank value starts at 1 and increases sequentially.
ROW_NUMBER	Returns the row number for the current row based on the ORDER BY clause within each partition. Rows containing identical values receive different row numbers.
SUM	Returns the sum of non-NULL input expressions.
VAR_POP	Returns the population variance of non-NULL records.
VAR_SAMP	Returns the sample variance of non-NULL records.

Window Function Syntax

A window function performs a calculation across a set of table rows that has some relationship to the current row. This is comparable to how an aggregate function can run a calculation. The difference is that a window function does not group rows into a single output row. With a window function, the rows retain their separate identities.

A window function call uses the OVER() clause directly following the window function’s name and argument(s). The OVER() clause may use the following optional arguments:

• PARTITION BY: Defines multiple window partitions.

• ORDER BY: Orders rows within each partition.

Syntax

window_function (expression) OVER (
   [ PARTITION BY expressionlist ]
   [ ORDER BY fieldlist ] ) 

Aggregate Window Functions

The OVER() clause can be used with regular aggregate functions such as:

• AVG

• COUNT

• MAX

• MIN

• SUM

Example

The following example uses the sample table provided below to show the OVER() clause used with the SUM aggregate function.

select 
   product_id, 
   branch, 
   amount, 
   SUM(amount) OVER (partition by branch order by amount DESC) as total_branch_amount
from transactions

product_id	branch	amount	total_branch_amount
Product1	A	30.0	30.0
Product2	A	24.0	54.0
Product3	A	2.0	56.0
Product3	B	45.0	45.0
Product2	B	10.0	55.0
Product1	B	3.0	58.0

General-Purpose Window Functions

The OVER() clause can be used with the following functions:

Function	Return Type	Description
CUME_DIST()	Double	Calculates the cumulative distribution of the current row within the window partition.
DENSE_RANK()	BIGINT	Returns the rank of the current row within its partition and ordering. Rows that are equal have the same rank.
LAG()	Same as input	Returns the row before the current one in a partition. If there are no rows, returns null.
LEAD()	Same as input	Returns the row after the current one in a partition. If there are no rows, returns null.
NTILE([integer] ntile)	Integer	NTILE function equally splits the rows in each partition into N ranked parts. Has to be used with an ORDER BY clause.
PERCENT_RANK()	Double	Returns the percent rank of the current row in the partition based on the order by clause.
RANK()	BIGINT	Returns the rank of the current row within its partition and ordering. Rows that are equal have the same rank. However, the count of tied rows is added to the next rank, instead of being incremented by just one.
ROW_NUMBER()	BIGINT	Returns the row number for the current row based on the order by clause within each partition.

For more information about Window Functions, see SQL Window Functions.