org.apache.kafka.streams.kstream

Class TimeWindows

java.lang.Object

org.apache.kafka.streams.kstream.Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

org.apache.kafka.streams.kstream.TimeWindows

public final class TimeWindows
extends Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

The fixed-size time-based window specifications used for aggregations.

The semantics of time-based aggregation windows are: Every T1 (advance) milliseconds, compute the aggregate total for T2 (size) milliseconds.

• If advance < size a hopping windows is defined:
  it discretize a stream into overlapping windows, which implies that a record maybe contained in one and or more "adjacent" windows.
  
• If advance == size a tumbling window is defined:
  it discretize a stream into non-overlapping windows, which implies that a record is only ever contained in one and only one tumbling window.

Thus, the specified TimeWindows are aligned to the epoch. Aligned to the epoch means, that the first window starts at timestamp zero. For example, hopping windows with size of 5000ms and advance of 3000ms, have window boundaries [0;5000),[3000;8000),... and not [1000;6000),[4000;9000),... or even something "random" like [1452;6452),[4452;9452),...

For time semantics, see TimestampExtractor.

See Also:

SessionWindows, UnlimitedWindows, JoinWindows, KGroupedStream.windowedBy(Windows), TimestampExtractor

Field Summary

Fields

Modifier and Type	Field and Description
long	advanceMs The size of the window's advance interval in milliseconds, i.e., by how much a window moves forward relative to the previous one.
long	sizeMs The size of the windows in milliseconds.

Fields inherited from class org.apache.kafka.streams.kstream.Windows

segments

Method Summary

Methods

Modifier and Type	Method and Description
TimeWindows	advanceBy(long advanceMs) Return a window definition with the original size, but advance ("hop") the window by the given interval, which specifies by how much a window moves forward relative to the previous one.
boolean	equals(java.lang.Object o)
int	hashCode()
long	maintainMs() Return the window maintain duration (retention time) in milliseconds.
static TimeWindows	of(long sizeMs) Return a window definition with the given window size, and with the advance interval being equal to the window size.
long	size() Return the size of the specified windows in milliseconds.
TimeWindows	until(long durationMs) Set the window maintain duration (retention time) in milliseconds.
java.util.Map<java.lang.Long,org.apache.kafka.streams.kstream.internals.TimeWindow>	windowsFor(long timestamp) Create all windows that contain the provided timestamp, indexed by non-negative window start timestamps.

Methods inherited from class org.apache.kafka.streams.kstream.Windows

segments

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, toString, wait, wait, wait

Field Detail

sizeMs

public final long sizeMs

The size of the windows in milliseconds.

advanceMs

public final long advanceMs

The size of the window's advance interval in milliseconds, i.e., by how much a window moves forward relative to the previous one.

Method Detail

of

public static TimeWindows of(long sizeMs)
                      throws java.lang.IllegalArgumentException

Return a window definition with the given window size, and with the advance interval being equal to the window size. The time interval represented by the N-th window is: [N * size, N * size + size).

This provides the semantics of tumbling windows, which are fixed-sized, gap-less, non-overlapping windows. Tumbling windows are a special case of hopping windows with advance == size.

Parameters:

sizeMs - The size of the window in milliseconds

Returns:

a new window definition with default maintain duration of 1 day

Throws:

java.lang.IllegalArgumentException - if the specified window size is zero or negative

advanceBy

public TimeWindows advanceBy(long advanceMs)

Return a window definition with the original size, but advance ("hop") the window by the given interval, which specifies by how much a window moves forward relative to the previous one. The time interval represented by the N-th window is: [N * advance, N * advance + size).

This provides the semantics of hopping windows, which are fixed-sized, overlapping windows.

Parameters:

advanceMs - The advance interval ("hop") in milliseconds of the window, with the requirement that 0 < advanceMs &le; sizeMs.

Returns:

a new window definition with default maintain duration of 1 day

Throws:

java.lang.IllegalArgumentException - if the advance interval is negative, zero, or larger-or-equal the window size

windowsFor

public java.util.Map<java.lang.Long,org.apache.kafka.streams.kstream.internals.TimeWindow> windowsFor(long timestamp)

Description copied from class: Windows

Create all windows that contain the provided timestamp, indexed by non-negative window start timestamps.

Specified by:

windowsFor in class Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

Parameters:

timestamp - the timestamp window should get created for

Returns:

a map of windowStartTimestamp -> Window entries

size

public long size()

Description copied from class: Windows

Return the size of the specified windows in milliseconds.

Specified by:

size in class Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

Returns:

the size of the specified windows

until

public TimeWindows until(long durationMs)
                  throws java.lang.IllegalArgumentException

Description copied from class: Windows

Set the window maintain duration (retention time) in milliseconds. This retention time is a guaranteed lower bound for how long a window will be maintained.

Overrides:

until in class Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

Parameters:

durationMs - the window retention time

Returns:

itself

Throws:

java.lang.IllegalArgumentException - if duration is smaller than the window size

maintainMs

public long maintainMs()

Return the window maintain duration (retention time) in milliseconds.

For TimeWindows the maintain duration is at least as small as the window size.

Overrides:

maintainMs in class Windows<org.apache.kafka.streams.kstream.internals.TimeWindow>

Returns:

the window maintain duration

equals

public boolean equals(java.lang.Object o)

Overrides:

equals in class java.lang.Object

hashCode

public int hashCode()

Overrides:

hashCode in class java.lang.Object