如何計算 window function 中的不同值

Question

編輯我添加了另一個 ID 列以表明我不能只做一個簡單的分組依據以及為什么需要某種 window function。

我需要計算每個 tx_id 的貨幣數量。

我可以使用下面的代碼來做到這一點，但感覺太復雜了。
我認為應該可以在一個 window function 中實現，但我無法確定語法

-- test data
WITH cte AS (
    SELECT * FROM (
        VALUES
            (1,123, 'GBP'), -- 2 ccys
            (2,123, 'USD'),
            (3,123, 'USD'),
            (4,124, 'GBP'), -- 1 ccys
            (5,124, 'GBP'),
            (6,125, 'EUR'), -- 3 ccys
            (7,125, 'EUR'),
            (8,125, 'JPY'),
            (9,125, 'USD'),
            (10,125, 'EUR')
) AS a (id, tx_id, ccy)
    )

,ccy_count as (
    select id, tx_id, ccy,
        dense_rank() over  (PARTITION BY group_id ORDER BY ccy ) as dense_rank_ccy
    from cte
)
    
select id, 
        tx_id, 
        ccy, 
        max(dense_rank_ccy) over  (PARTITION BY group_id  ) as ccy_count
    from ccy_count
 order by tx_id, ccy

Answer 1

如果您只想報告每筆交易的不同貨幣數量，那么您應該聚合，而不是使用 window function：

SELECT tx_id, COUNT(DISTINCT ccy) AS cnt
FROM cte
GROUP BY tx_id;

Answer 2

您可以在標量子查詢中使用count(distinct ccy) ，如下所示：

WITH cte AS (
    SELECT * FROM (
        VALUES
            (1,123, 'GBP'), -- 2 ccys
            (2,123, 'USD'),
            (3,123, 'USD'),
            (4,124, 'GBP'), -- 1 ccys
            (5,124, 'GBP'),
            (6,125, 'EUR'), -- 3 ccys
            (7,125, 'EUR'),
            (8,125, 'JPY'),
            (9,125, 'USD'),
            (10,125, 'EUR')
  ) AS a (id, tx_id, ccy)
)
select id, tx_id, ccy,
    (
     select count(distinct D.ccy)
     from cte D 
     where D.tx_id = T.tx_id
    ) distinct_ccy
from cte T

或者您可以使用 dense_rank function，如下所示：

select id, tx_id, ccy,
  dense_rank() over (partition by tx_id order by ccy) + 
  dense_rank() over (partition by tx_id order by ccy desc) - 1 as distinct_ccy
from cte T
order by id

看演示