Clickhouse：计算多个数组列的平均值答案

【问题标题】：Clickhouse: calculating average on multiple array columnsClickhouse：计算多个数组列的平均值
【发布时间】：2021-10-05 09:51:17
【问题描述】：

我有一个表格，其中包含遥测信息，例如冷却风扇速度、温度等。

遥测数据存储在数组中，因为冷却风扇或其他项目的数量是动态的，可能会不时变化，最大大小未知。

我正在尝试编写查询以获取 n 次（例如 5 分钟）的风扇速度、温度和功耗的平均值

表格是这样的：

CREATE TABLE stats
(
    `datetime`      DateTime('UTC') DEFAULT now(),
    `worker_id`     UInt64,
    `project_id`    UInt64,
    `platform_type` UInt8,
    `temp`          Array(Int32),
    `fan`           Array(Int32),
    `units`         Int32,
    `power`         Array(Float32),
    `power_total`   Float32
)
ENGINE = MergeTree()
PARTITION BY toYYYYMM(datetime)
ORDER BY (worker_id, datetime);

数据如下：

────────────datetime─┬─worker_id─┬─temp─────────────────────────────────────┬─power─────────────────────────────────────┬─fan──────────────────────────────────────┐
│ 2021-07-26 16:37:00 │      3081 │ [51,55,53,51,56,62,58,57,55,64,63,55,61] │ [120,91,60,91,60,90,90,60,90,89,90,60,89] │ [70,70,70,70,70,70,70,70,70,80,80,70,70] │
│ 2021-07-26 16:37:10 │      3081 │ [51,56,54,52,56,63,58,58,55,64,63,56,62] │ [120,89,60,90,60,90,89,60,88,88,90,60,90] │ [70,70,70,70,70,70,70,70,70,80,80,70,70] │
│ 2021-07-26 16:37:20 │      3081 │ [51,56,54,52,56,63,58,58,55,64,63,56,62] │ [120,91,60,92,60,90,91,60,88,88,90,60,89] │ [70,70,70,70,70,70,70,70,70,80,80,70,70] │
│ 2021-07-26 16:37:30 │      3081 │ [51,56,54,52,57,62,58,58,56,64,63,57,62] │ [119,90,60,92,60,90,91,60,91,88,90,60,89] │ [70,70,70,69,70,70,70,70,70,80,80,70,70] │
│ 2021-07-26 16:37:40 │      3081 │ [51,56,55,52,57,63,58,58,56,64,63,56,62] │ [119,91,60,88,60,90,90,60,90,90,89,60,89] │ [70,70,70,70,70,70,70,70,68,80,80,70,70] │
└─────────────────────┴───────────┴──────────────────────────────────────────┴───────────────────────────────────────────┴──────────────────────────────────────────┘

我能够编写一个查询来获取单个列的平均值，但很难为其余列编写最佳查询。

select worker_id, round(avg(temp_value), 1) as temp_avg, temp_index
from (
      select datetime, worker_id, temp from stats
      WHERE `worker_id` = 3081
        and `datetime` between
          toDateTime('2021-07-26 19:40:00', 'UTC') AND
          toDateTime('2021-07-26 19:45:00', 'UTC')
      order by `datetime` ASC
         ) array join
     temp as temp_value,
     arrayEnumerate(temp) as temp_index
group by (worker_id, temp_index)
order by temp_index ASC;

您能否建议我一个最佳查询来计算 temp、fan、power 和 power_total 的平均值？

【问题讨论】：

标签： sql clickhouse

【解决方案1】：

考虑使用ForEach聚合函数组合器：

SELECT worker_id, avgForEach(temp) temp_avg, avgForEach(power) power_avg, avgForEach(fan) fan_avg
FROM (
    /* emulate the test dataset */
    SELECT data.1 datetime, data.2 worker_id, data.3 temp, data.4 power, data.5 fan
    FROM (
        SELECT arrayJoin([
            (('2021-07-26 16:37:00'), 3081, [51,55,53,51,56,62,58,57,55,64,63,55,61], [120,91,60,91,60,90,90,60,90,89,90,60,89], [70,70,70,70,70,70,70,70,70,80,80,70,70]),
            (('2021-07-26 16:37:10'), 3081, [51,56,54,52,56,63,58,58,55,64,63,56,62], [120,89,60,90,60,90,89,60,88,88,90,60,90], [70,70,70,70,70,70,70,70,70,80,80,70,70]),
            (('2021-07-26 16:37:20'), 3081, [51,56,54,52,56,63,58,58,55,64,63,56,62], [120,91,60,92,60,90,91,60,88,88,90,60,89], [70,70,70,70,70,70,70,70,70,80,80,70,70]),
            (('2021-07-26 16:37:30'), 3081, [51,56,54,52,57,62,58,58,56,64,63,57,62], [119,90,60,92,60,90,91,60,91,88,90,60,89], [70,70,70,69,70,70,70,70,70,80,80,70,70]),
            (('2021-07-26 16:37:40'), 3081, [51,56,55,52,57,63,58,58,56,64,63,56,62], [119,91,60,88,60,90,90,60,90,90,89,60,89], [70,70,70,70,70,70,70,70,68,80,80,70,70])]) as data)
    )
/*WHERE {timerange condition}*/
GROUP BY worker_id

/*
┌─worker_id─┬─temp_avg───────────────────────────────────────────────┬─power_avg─────────────────────────────────────────────────┬─fan_avg──────────────────────────────────────┐
│      3081 │ [51,55.8,54,51.8,56.4,62.6,58,57.8,55.4,64,63,56,61.8] │ [119.6,90.4,60,90.6,60,90,90.2,60,89.4,88.6,89.8,60,89.2] │ [70,70,70,69.8,70,70,70,70,69.6,80,80,70,70] │
└───────────┴────────────────────────────────────────────────────────┴───────────────────────────────────────────────────────────┴──────────────────────────────────────────────┘
*/

【讨论】：

【解决方案2】：

select worker_id, round(avg(temp_value), 1) as temp_avg,
                  round(avg(fan_value), 1) as fan_avg, 
                  round(avg(power_value), 1) as power_avg, 
                  temp_index
from (
      select datetime, worker_id, temp, fan, power from stats
      WHERE `worker_id` = 3081
        and `datetime` between
          toDateTime('2021-07-26 19:40:00', 'UTC') AND
          toDateTime('2021-07-26 19:45:00', 'UTC')
      order by `datetime` ASC
         ) array join
     temp as temp_value,
     fan as fan_value,
     power as power_value,
     arrayEnumerate(temp) as temp_index
group by (worker_id, temp_index)
order by temp_index ASC;




select worker_id, 
       avgForEach(temp), 
       avgForEach(fan), 
       avgForEach(power)
from stats
WHERE `worker_id` = 3081
  and `datetime` between
    toDateTime('2021-07-26 19:40:00', 'UTC') AND
    toDateTime('2021-07-26 19:45:00', 'UTC')
group by worker_id

【讨论】：