All Products
Search

This Product

    Realtime Compute for Apache Flink:HOP

    Document Center

    Realtime Compute for Apache Flink:HOP

    Last Updated:Mar 26, 2026

    Use the HOP function in the GROUP BY clause to define a hopping window (also called a sliding window) in Flink SQL. Each event can belong to multiple overlapping windows simultaneously, making HOP well suited for computing moving aggregates — for example, counting clicks per user over the last minute, recalculated every 30 seconds.

    Syntax

    HOP(<time-attr>, <slide-interval>, <size-interval>)

    Parameters

    Parameter Description Example
    <time-attr> A time attribute column in the stream. Specifies whether to use event time or processing time. For more information, see Time attributes. —
    <slide-interval> How often the window advances. Defines the time difference between the start of consecutive windows. Format: INTERVAL 'num' timeUnit. INTERVAL '10' SECOND — the window advances every 10 seconds
    <size-interval> The total duration covered by each window. Format: INTERVAL 'num' timeUnit. INTERVAL '5' MINUTE — each window spans 5 minutes

    Window overlap behavior

    The relationship between <slide-interval> and <size-interval> controls how windows overlap:

    Condition Behavior
    slide-interval < size-interval Windows overlap. Each event is assigned to multiple windows. This is the standard hopping window pattern.
    slide-interval = size-interval Windows are back-to-back with no gaps — equivalent to tumbling windows.
    slide-interval > size-interval Windows do not overlap and are separated by gaps.

    Window identifier functions

    Use these functions in SELECT to retrieve the start time, end time, or time attribute of a window.

    Function Return type Description
    HOP_START(<time-attr>, <slide-interval>, <size-interval>) TIMESTAMP Start time of the window, inclusive. For example, returns 00:10 for the window [00:10, 00:15].
    HOP_END(<time-attr>, <slide-interval>, <size-interval>) TIMESTAMP End time of the window, inclusive. For example, returns 00:15 for the window [00:00, 00:15].
    HOP_ROWTIME(<time-attr>, <slide-interval>, <size-interval>) TIMESTAMP (rowtime-attr) End time of the window, exclusive. For example, returns 00:14:59.999 for the window (00:00, 00:15). Returns a rowtime attribute that you can use in downstream window operations. Use only in event-time windows. For more information, see Cascading windows.
    HOP_PROCTIME(<time-attr>, <slide-interval>, <size-interval>) TIMESTAMP (rowtime-attr) End time of the window, exclusive. For example, returns 00:14:59.999 for the window (00:00, 00:15). Returns a processing-time attribute for use in downstream window operations. Use only in processing-time windows. For more information, see Cascading windows.

    Example

    This example counts clicks per user over a 1-minute sliding window that advances every 30 seconds. The query uses HOP_START and HOP_END to retrieve the window boundaries.

    Test data (user_clicks)

    username (VARCHAR) click_url (VARCHAR) eventtime (VARCHAR)
    Jark http://taobao.com/xxx 2024-10-10 10:00:00.0
    Jark http://taobao.com/xxx 2024-10-10 10:00:10.0
    Jark http://taobao.com/xxx 2024-10-10 10:00:49.0
    Jark http://taobao.com/xxx 2024-10-10 10:01:05.0
    Jark http://taobao.com/xxx 2024-10-10 10:01:58.0
    Timo http://taobao.com/xxx 2024-10-10 10:02:10.0

    SQL

    CREATE TEMPORARY TABLE user_clicks (
  username VARCHAR,
  click_url VARCHAR,
  eventtime VARCHAR,
  ts AS TO_TIMESTAMP(eventtime),
  WATERMARK FOR ts AS ts - INTERVAL '2' SECOND  -- Define a watermark for the rowtime.
) WITH (
  'connector' = 'kafka',
  'topic' = '<yourTopic>',
  'properties.bootstrap.servers' = '<brokers>',
  'scan.startup.mode' = 'earliest-offset',
  'format' = 'csv'
);

CREATE TEMPORARY TABLE hop_output (
  window_start TIMESTAMP,
  window_end TIMESTAMP,
  username VARCHAR,
  clicks BIGINT
) WITH (
  'connector' = 'print',
  'logger' = 'true'
);

INSERT INTO hop_output
SELECT
    HOP_START(ts, INTERVAL '30' SECOND, INTERVAL '1' MINUTE),
    HOP_END(ts, INTERVAL '30' SECOND, INTERVAL '1' MINUTE),
    username,
    COUNT(click_url)
FROM user_clicks
GROUP BY HOP(ts, INTERVAL '30' SECOND, INTERVAL '1' MINUTE), username;

    Results

    window_start (TIMESTAMP) window_end (TIMESTAMP) username (VARCHAR) clicks (BIGINT)
    2024-10-10 09:59:30.0 ① 2024-10-10 10:00:30.0 Jark 2
    2024-10-10 10:00:00.0 2024-10-10 10:01:00.0 Jark 3
    2024-10-10 10:00:30.0 2024-10-10 10:01:30.0 Jark 2
    2024-10-10 10:01:00.0 2024-10-10 10:02:00.0 Jark 2
    2024-10-10 10:01:30.0 2024-10-10 10:02:30.0 ② Jark 1
    2024-10-10 10:01:30.0 2024-10-10 10:02:30.0 ② Timo 1

    ① First window start time

    When a hopping window cannot determine the exact time the first event entered the stream, it shifts the first window's start time backward by: window duration - sliding step.

    For example:

    Window duration (seconds) Sliding step (seconds) Event time First window start First window end
    120 30 2024-07-31 10:00:00.0 2024-07-31 09:58:30.0 2024-07-31 10:00:30.0
    60 10 2024-07-31 10:00:00.0 2024-07-31 09:59:10.0 2024-07-31 10:00:10.0

    ② Window not yet triggered

    The row with window_end = 2024-10-10 10:02:30.0 does not appear in the results because the window has not been triggered yet. A window triggers when:

    event time >= window_end + watermark offset

    For example: 10:02:30.0 + 2 seconds = 10:02:32.0. An event at or after 10:02:32.0 from any user would trigger this window.