筛选和归约函数 - Amazon Timestream
使用情况信息查询示例

要获得与亚马逊 Timestream 类似的功能 LiveAnalytics，可以考虑适用于 InfluxDB 的亚马逊 Timestream。适用于 InfluxDB 的 Amazon Timestream 提供简化的数据摄取和个位数毫秒级的查询响应时间，以实现实时分析。点击此处了解更多信息。

本文属于机器翻译版本。若本译文内容与英语原文存在差异，则一律以英文原文为准。

筛选和归约函数

Amazon Timestream 支持对时间序列数据执行筛选和归约操作的函数。本节提供 LiveAnalytics 筛选和缩减函数的时间流的用法信息，以及示例查询。

使用情况信息

函数 输出数据类型 说明

filter(timeseries(T), function(T, Boolean))

timeseries(T)

根据输入的时间序列构造时间序列，其中使用的值是传递的 function 返回的 true。

reduce(timeseries(T), initialState S, inputFunction(S, T, S), outputFunction(S, R))

R

返回从时间序列中减去的单个值。inputFunction 将按顺序对时间序列中的每个元素进行调用。除获取当前元素以外，inputFunction 还会获取当前状态（初始为 initialState）并返回新状态。将调用 outputFunction，以将最终状态转换为结果值。outputFunction 可以是恒等函数。

查询示例

构造主机的 CPU 利用率时间序列，并筛选测量值大于 70 的数据点：

WITH time_series_view AS (
    SELECT INTERPOLATE_LINEAR(
        CREATE_TIME_SERIES(time, ROUND(measure_value::double,2)), 
            SEQUENCE(ago(15m), ago(1m), 10s)) AS cpu_user
    FROM sample.DevOps
    WHERE hostname = 'host-Hovjv' and measure_name = 'cpu_utilization'
        AND time > ago(30m)
    GROUP BY hostname
)
SELECT FILTER(cpu_user, x -> x.value > 70.0) AS cpu_above_threshold
from time_series_view

构造主机的 CPU 利用率时间序列，并计算测量值的平方和：

WITH time_series_view AS (
    SELECT INTERPOLATE_LINEAR(
        CREATE_TIME_SERIES(time, ROUND(measure_value::double,2)), 
            SEQUENCE(ago(15m), ago(1m), 10s)) AS cpu_user
    FROM sample.DevOps
    WHERE hostname = 'host-Hovjv' and measure_name = 'cpu_utilization'
        AND time > ago(30m)
    GROUP BY hostname
)
SELECT REDUCE(cpu_user,
    DOUBLE '0.0',
    (s, x) -> x.value * x.value + s,
    s -> s)
from time_series_view

构造主机的 CPU 利用率时间序列，并确定超过 CPU 阈值的样本所占比例：

WITH time_series_view AS (
    SELECT INTERPOLATE_LINEAR(
        CREATE_TIME_SERIES(time, ROUND(measure_value::double,2)), 
            SEQUENCE(ago(15m), ago(1m), 10s)) AS cpu_user
    FROM sample.DevOps
    WHERE hostname = 'host-Hovjv' and measure_name = 'cpu_utilization'
        AND time > ago(30m)
    GROUP BY hostname
)
SELECT ROUND(
    REDUCE(cpu_user, 
      -- initial state 
      CAST(ROW(0, 0) AS ROW(count_high BIGINT, count_total BIGINT)),
      -- function to count the total points and points above a certain threshold
      (s, x) -> CAST(ROW(s.count_high + IF(x.value > 70.0, 1, 0), s.count_total + 1) AS ROW(count_high BIGINT, count_total BIGINT)),
      -- output function converting the counts to fraction above threshold
      s -> IF(s.count_total = 0, NULL, CAST(s.count_high AS DOUBLE) / s.count_total)), 
    4) AS fraction_cpu_above_threshold
from time_series_view