Clickhouse 窗口函数总结——Clickhouse 基础篇（九）

自增行号函数

• 按照值排序时产生一个自增行号，不会重复。针对相同数据，先查出的排名在前，没有重复值

SELECT 
  row_number() OVER (PARTITION BY EventDate ORDER BY cnt ASC) AS rno,
  OS,
  EventDate,
  cnt
FROM 
(
  SELECT
    OS,
    cnt,
    EventDate
  FROM 
  (
    SELECT
      EventDate,
      OS,
      uniqExact(UserID) AS cnt
    FROM tutorial.hits_v1
    GROUP BY
      EventDate, OS
  ) AS t
) AS T
limit 20

跳跃排名函数

• 跳跃排名，值相等时会重复，会产生空位。相同数据排名相同，比如并列第1，则两行数据（这里为rank列）都标为1，下一位将是第3名，中间的2被直接跳过了。排名存在重复值。

SELECT 
  rank() OVER (PARTITION BY EventDate ORDER BY cnt ASC) AS rno,
  OS,
  EventDate,
  cnt
FROM 
(
  SELECT
    OS,
    cnt,
    EventDate
  FROM 
  (
    SELECT
      EventDate,
      OS,
      uniqExact(UserID) AS cnt
    FROM tutorial.hits_v1
    GROUP BY
      EventDate, OS
  ) AS t
) AS T
limit 20

连续排名函数

• 连续排名，值相等时会重复，不会产生空位。比如两条并列第1，则两行数据（这里为rank列）都标为1，下一个排名将是第2名。

SELECT 
  dense_rank() OVER (PARTITION BY EventDate ORDER BY cnt ASC) AS rno,
  OS,
  EventDate,
  cnt
FROM 
(
  SELECT
    OS,
    cnt,
    EventDate
  FROM 
  (
    SELECT
      EventDate,
      OS,
      uniqExact(UserID) AS cnt
    FROM tutorial.hits_v1
    GROUP BY
      EventDate, OS
  ) AS t
) AS T
limit 20

窗口计数函数

SELECT
    number AS x,
    count() OVER (PARTITION BY intDiv(x, 3) ORDER BY x ASC) AS cnt
FROM numbers(10)
SETTINGS max_block_size = 3

Query id: dfeb20ae-1745-4496-877d-74d8aea95414

┌─x─┬─cnt─┐
│ 0 │   1 │
│ 1 │   2 │
│ 2 │   3 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 3 │   1 │
│ 4 │   2 │
│ 5 │   3 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 6 │   1 │
│ 7 │   2 │
│ 8 │   3 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 9 │   1 │
└───┴─────┘

窗口最大值函数

SELECT
    number AS x,
    max(x) OVER (PARTITION BY intDiv(x, 3) ORDER BY x DESC) AS cnt
FROM numbers(10)
SETTINGS max_block_size = 3

Query id: bdb8038f-0d2e-442b-a4fd-258787f5738f

┌─x─┬─cnt─┐
│ 2 │   2 │
│ 1 │   2 │
│ 0 │   2 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 5 │   5 │
│ 4 │   5 │
│ 3 │   5 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 8 │   8 │
│ 7 │   8 │
│ 6 │   8 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 9 │   9 │
└───┴─────┘

窗口最小值函数

SELECT
    number AS x,
    MIN(x) OVER (PARTITION BY intDiv(x, 3) ORDER BY x ASC) AS cnt
FROM numbers(10)
SETTINGS max_block_size = 3

Query id: e763ce9f-3eba-40a6-ad78-b0ae20dcfdda

┌─x─┬─cnt─┐
│ 0 │   0 │
│ 1 │   0 │
│ 2 │   0 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 3 │   3 │
│ 4 │   3 │
│ 5 │   3 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 6 │   6 │
│ 7 │   6 │
│ 8 │   6 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 9 │   9 │
└───┴─────┘

窗口平均值函数

SELECT
    number AS x,
    avg(x) OVER (PARTITION BY intDiv(x, 3) ORDER BY x ASC) AS cnt
FROM numbers(10)
SETTINGS max_block_size = 3

Query id: c9b8ab1b-8a6f-4191-a248-f8a589becb5a

┌─x─┬─cnt─┐
│ 0 │   0 │
│ 1 │ 0.5 │
│ 2 │   1 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 3 │   3 │
│ 4 │ 3.5 │
│ 5 │   4 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 6 │   6 │
│ 7 │ 6.5 │
│ 8 │   7 │
└───┴─────┘
┌─x─┬─cnt─┐
│ 9 │   9 │
└───┴─────┘