Question

现在，在这个论坛中已经多次提出过类似的问题，但是，我一直在努力计算posrgresql (Version 9.4.2)表中列的每月运行总和：

created_at           |  points
-----------------------------
2015-12-10 12:26:34  |  199
2016-01-10 12:26:34  |  199
2016-02-10 12:26:34  |  199
2016-03-10 12:26:34  |  500
2016-03-10 12:26:39  |  199

预期结果是：

created_at           |  sum
-----------------------------
2015-12-01           |  199
2016-01-01           |  398
2016-02-01           |  597
2016-03-01           |  1296

我到达的工作sql，主要是通过阅读其他主题后的试错，是：

SELECT 
    date_trunc('month',created_at)::date as month_created, 
    sum(sum(points)) 
OVER 
    (
        ORDER BY date_trunc('month',created_at)::date
    ) 
FROM 
    point_histories
GROUP BY 1 
ORDER BY 1 ASC

我的困惑是为什么以下（我显然是误解）不起作用？

SELECT 
    month_created, 
    sum(points), 
    sum(sum(points)) 
OVER 
    (
        PARTITION BY month_created 
        ORDER BY created_at
    ) 
    AS cum_amt
FROM   
    (
        SELECT 
            *, 
            date_trunc('month', created_at)::date as month_created 
        FROM 
            point_histories
    ) 
    as derivedTable
ORDER  BY created_at
GROUP BY month_created

上面给出了以下错误：

********* Error **********

ERROR: syntax error at or near "group"
SQL state: 42601
Character: 314

如果我删除GROUP BY子句，我会得到：

********** Error **********

ERROR: column "derivedtable.month_created" must appear in the GROUP BY clause or be used in an aggregate function
SQL state: 42803
Character: 12

Answer 1

IMO，order by子句应该完成你的查询。

SELECT 
    month_created, 
    sum(points), 
    sum(sum(points)) 
OVER 
    (
        PARTITION BY month_created 
        ORDER BY created_at
    ) 
    AS cum_amt
FROM   
    (
        SELECT 
            *, 
            date_trunc('month', created_at)::date as month_created 
        FROM 
            point_histories
    ) 
    as derivedTable
GROUP BY month_created
ORDER  BY created_at

posrgresql：运行总列的混乱

1 个答案: