为什么我的窗口函数似乎使用了整个数据集？

Question

我有这个查询，但最大值和平均值似乎没有按预期工作：

SELECT 
    UPPER(SUBSTR(first_query.periodname, 1,3)) AS month,
    ROUND(first_query.amount) AS amount,
    ROUND(MAX(first_query.amount) OVER (ORDER BY first_query.postingPeriod ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING)) AS maxamount,
    ROUND(AVG(first_query.amount) OVER (ORDER BY first_query.postingPeriod ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING)) AS avgamount,
    ROUND(LAG(first_query.amount) OVER (ORDER BY first_query.postingPeriod)) AS lagamount,
    ROUND(LEAD(first_query.amount) OVER (ORDER BY first_query.postingPeriod)) AS leadamount
FROM (
    SELECT
        SUM(tl.foreignamount * t.ExchangeRate * -1) AS amount,
        BUILTIN.DF(t.postingPeriod) AS periodname,
        t.postingPeriod
    FROM
        Transaction t
    JOIN transactionline tl ON t.id = tl.transaction
    JOIN account a ON a.id = tl.expenseaccount
    WHERE
        EXTRACT(YEAR FROM t.trandate) = 2023  
    GROUP BY
        BUILTIN.DF(t.postingPeriod),
        t.postingPeriod
    ORDER BY
        t.postingPeriod
) AS first_query
ORDER BY
    first_query.postingPeriod;

其结果是：

月	金额	最大金额	平均金额	拉加蒙	铅量
一月	1833916	1833916	1833916	空	1857884
二月	1857884	1857884	1845900	1833916	2048416
三月	2048416	2048416	1913405	1857884	1731385
四月	1731385	2048416	1867900	2048416	1543348
五月	1543348	2048416	1802990	1731385	1642432
六月	1642432	2048416	1776230	1543348	1541351
七月	1541351	2048416	1742676	1642432	1862392
八月	1862392	2048416	1757640	1541351	1377795
九月	1377795	2048416	1715435	1862392	1324765
十月	1324765	2048416	1676368	1377795	1042649
十一月	1042649	2048416	1618758	1324765	1365764
十二月	1365764	2048416	1597675	1042649	空

因此您可以看到滞后和超前效果很好，但是很容易看到最大值不起作用。据我所知，它只是不断地将新行添加到窗口中。

对我缺少的东西有什么想法吗？谢谢！

Answer 1

不太确定窗口函数出了什么问题，但找到了使用 rownum 和 WITH 子查询的解决方法。相当混乱，但至少它可以工作。

WITH first_query AS (
    SELECT
        t.postingPeriod AS postingperiod,
        BUILTIN.DF(t.postingPeriod) AS periodname,
        SUM(tl.foreignamount * t.ExchangeRate * -1) AS amount
    FROM
        Transaction t
    JOIN
        TransactionLine tl ON t.id = tl.transaction
    WHERE
        EXTRACT(YEAR FROM t.trandate) = 2023
    GROUP BY
        t.postingPeriod,
        BUILTIN.DF(t.postingPeriod)
),
second_query AS (
    SELECT
        rownum,
        postingperiod,
        periodname,
        ROUND(amount) AS amount
    FROM
        first_query
),
final_query AS (
    SELECT 
        rownum,
        postingperiod,
        periodname,
        amount,
        (SELECT MAX(sq.amount) 
         FROM second_query sq
         WHERE sq.rownum BETWEEN second_query.rownum - 1 AND second_query.rownum + 1) AS max_amount,
        (SELECT AVG(sq.amount) 
         FROM second_query sq
         WHERE sq.rownum BETWEEN second_query.rownum - 1 AND second_query.rownum + 1) AS avg_amount
    FROM 
        second_query
)
SELECT 
    UPPER(SUBSTR(periodname, 1, 3)) AS month,
    ROUND(amount) AS amount,
    ROUND(max_amount) AS maxamount,
    ROUND(avg_amount) AS avgamount
FROM 
    final_query
ORDER BY 
    postingperiod;