org.apache.hadoop.hive.ql.exec
Annotation Type WindowFunctionDescription

@Retention(value=RUNTIME)
@Target(value=TYPE)
@Documented
public @interface WindowFunctionDescription

Required Element Summary

Description

description

Optional Element Summary

boolean	impliesOrder Using in analytical functions to specify that UDF implies an ordering
boolean	pivotResult A WindowFunc is implemented as GenericUDAFResolver2.
boolean	rankingFunction Used in translations process to validate arguments
boolean	supportsWindow controls whether this function can be applied to a Window.

Element Detail

description

public abstract Description description

supportsWindow

public abstract boolean supportsWindow

controls whether this function can be applied to a Window.

Ranking function: Rank, Dense_Rank, Percent_Rank and Cume_Dist don't operate on Windows. Why? a window specification implies a row specific range i.e. every row gets its own set of rows to process the UDAF on. For ranking defining a set of rows for every row makes no sense.

All other UDAFs can be computed for a Window.

Default:

true

pivotResult

public abstract boolean pivotResult

A WindowFunc is implemented as GenericUDAFResolver2. It returns only one value. If this is true then the function must return a List which is taken to be the column for this function in the Output table returned by the WindowingTableFunction. Otherwise the output is assumed to be a single value, the column of the Output will contain the same value for all the rows.

Default:

false

rankingFunction

public abstract boolean rankingFunction

Used in translations process to validate arguments

Returns:

true if ranking function

Default:

false

impliesOrder

public abstract boolean impliesOrder

Using in analytical functions to specify that UDF implies an ordering

Returns:

true if the function implies order

Default:

false