CollectTop 返回的行数超出了我在 Azure 流分析中的预期

Posted 2023-05-08

【中文标题】CollectTop 返回的行数超出了我在 Azure 流分析中的预期

【英文标题】：CollectTop is returning more rows than I would expect in Azure Stream Analytics

【发布时间】：2020-08-05 14:12:44

【问题描述】：

我上传了以下输入（在 Azure 门户中进行测试）：

[
  "engineid":"engine001","eventtime":1,"tmp":19.3,"hum":0.22,
  "engineid":"engine001","eventtime":2,"tmp":19.7,"hum":0.21,
  "engineid":"engine002","eventtime":3,"tmp":20.4,"hum":0.25,
  "engineid":"engine001","eventtime":4,"tmp":19.6,"hum":0.24
]

然后我尝试获取记录组，以便每个引擎都有最后 2 行。正如您在示例中看到的那样，我只有 2 个不同的引擎，所以我希望输出有两条记录，每条记录都包含排名记录，但我得到了 4 条输出记录。

这是我的查询：

-- Taking relevant fields from the input stream
WITH RelevantTelemetry AS
(
    SELECT  engineid, tmp, hum, eventtime
    FROM    [engine-telemetry] 
    WHERE   engineid IS NOT NULL
),
-- Grouping by engineid in TimeWindows
TimeWindows AS
(
    SELECT engineid, 
        CollectTop(2) OVER (ORDER BY eventtime DESC) as TimeWindow
    FROM
        [RelevantTelemetry]
    WHERE engineid IS NOT NULL
    GROUP BY SlidingWindow(hour, 24), engineid
)
--Output timewindows for verification purposes
SELECT TimeWindow
INTO debug
FROM TimeWindows

我使用了 TIMESTAMP BY 属性，更改了 GROUP BY 的顺序等，但我仍然有以下 4 条记录，而不是我期望的 2 条：

有什么想法吗？

[
"TimeWindow":
  [
    "rank":1,"value": "engineid":"engine001","tmp":0.0003,"hum":-0.0002,"eventtime":1
  ],
"TimeWindow":
  [
    "rank":1,"value":"engineid":"engine001","tmp":-0.0019,"hum":-0.0002,"eventtime":4,
    "rank":2,"value":"engineid":"engine001","tmp":-0.0026,"hum":-0.0002,"eventtime":2,
    "rank":3,"value":"engineid":"engine001","tmp":0.0003,"hum":-0.0002,"eventtime":1
  ],
"TimeWindow":
  [
    "rank":1,"value":"engineid":"engine002","tmp":0.0017,"hum":0.0003,"eventtime":3
  ],
"TimeWindow":
  [
    "rank":1,"value":"engineid":"engine001","tmp":-0.0019,"hum":-0.0002,"eventtime":4,
    "rank":2,"value":"engineid":"engine001","tmp":-0.0026,"hum":-0.0002,"eventtime":2
  ]
]

【问题讨论】：

你试过用GROUP BY TumblingWindow(hour, 24), engineid吗？

谢谢，史蒂夫 - 这似乎确实有效。

【参考方案1】：

根据@SteveZhao 的建议，您需要使用GROUP BY TumblingWindow(hour, 24), engineid 而不是GROUP BY SlidingWindow(hour, 24), engineid

滑动窗口可以根据时间间隔重叠条目

欲了解更多信息，请参阅： https://docs.microsoft.com/en-us/azure/stream-analytics/stream-analytics-window-functions

【讨论】：

谢谢，普拉纳夫和史蒂夫。这似乎会带来更好的结果。我认为滑动窗口将始终保持最新的 N 条记录可用，并且担心 TumblingWindow 可能没有 N 条记录，如果它们越过边界并且新窗口已经开始......