计算不同的和 window 个函数

Question

我有一个 ID、交易、这些交易的日期以及这些交易的类别的列表。我想在每个 ID

中创建每个不同类别的计数

我的起始 table 看起来像这样：

id  trxn_dt     trxn_amt trxn_category
1   10/31/2014  58       apple
1   11/9/2014   34       banana
1   12/10/2014  12       apple
2   7/8/2014    78       banana
2   11/20/2014  99       banana
3   1/5/2014    120      orange
4   2/17/2014   588      apple
4   2/18/2014   8        banana
4   3/9/2014    65       orange
4   4/25/2014   74       apple

我希望最终结果看起来像这样：

id  trxn_dt     trxn_amt trxn_category  number_category
1   10/31/2014  58       apple          2
1   11/9/2014   34       banana         2 
1   12/10/2014  12       apple          2
2   7/8/2014    78       banana         1
2   11/20/2014  99       banana         1
3   1/5/2014    120      orange         1
4   2/17/2014   588      apple          3
4   2/18/2014   8        banana         3
4   3/9/2014    65       orange         3
4   4/25/2014   74       apple          3

我试过使用 count(distinct(trxn_category)) over(partition by id,trxn_category order by id) as number_category 但我在使用 'distinct'

时遇到错误

Answer 1

您可以为此使用相关子查询：

SELECT id, trxn_dt, trxn_amt, trxn_category,
       (SELECT COUNT(DISTINCT trxn_category)
        FROM mytable AS t2
        WHERE t2.id = t1.id) AS cnt
FROM mytable AS t1

Demo here

Answer 2

;WITH cteCounts AS (
    SELECT
       id
       ,COUNT(DISTINCT trxn_category) as CategoryCount
    FROM
       table
    GROUP BY
       id
)


SELECT
    t.id
    ,trxn_dt
    ,trxn_amt
    ,trxn_category
    ,c.CategoryCount
FROM
    table t
    INNER JOIN cteCounts c
    ON t.id = c.id

不能在分区聚合中使用 DISTINCT。

Answer 3

大多数 DBMS 在窗口函数中不支持 DISTINCT，但您可以使用两个 DENSE_RANKs:

模拟 COUNT(DISTINCT)

DENSE_RANK() over (partition by id,trxn_category order by id ASC)-
DENSE_RANK() over (partition by id,trxn_category order by id DESC)

或嵌套的 MAX(DENSE_RANK):

select
   MAX(dr) over (partition by id,trxn_category)
from
 (
   select 
      DENSE_RANK() over (partition by id,trxn_category order by id DESC) as dr
 )

计算不同的和 window 个函数

count distinct and window functions

sql

distinct

window-functions