MySQL:带有JOIN的SUM()返回不正确的值

Question

我试图为表中的每个用户获取一个SUM(),但MySQL返回错误的值.

它应该是这样的(http://sqlfiddle.com/#!2/7b988/4/0)

user    amount
110     20.898319244385
114     43.144836425781
115     20.487638473511
116     26.07483291626
117     93.054000854492

但这就是它的外观(http://sqlfiddle.com/#!2/7b988/2/0)

user    amount
110     167.186554
114     129.434509
115     143.413469
116     208.598663
117     744.432007

这是我试图运行的查询:

SELECT 
    blocks.user_id, 
    SUM(payout_history.amount) as amount
FROM blocks
LEFT JOIN payout_history
ON blocks.user_id = payout_history.user_id
WHERE confirms > 520
GROUP BY blocks.user_id

我究竟做错了什么？

Answer 1

试试这个查询:

SELECT bl.user_id, SUM( ph.amount ) PAIDOUT
FROM (
   SELECT distinct blocks.user_id 
   FROM blocks
   WHERE confirms > 520
) bl
LEFT JOIN  payout_history ph
ON bl.user_id = ph.user_id
GROUP BY ph.user_id
;

SQLFiddle - > http://sqlfiddle.com/#!2/7b988/48

---编辑---解释查询是如何工作的(或者更确切地说,为什么你的查询不起作用)----

查看预期结果,查询似乎应计算amount每个列的列总和user_id,但仅针对那些user_id也在blocks表中的列,并且具有blocks.confirms比520更高的值.
简单连接(也是左外连接)不能在在这种情况下,因为该blocks表可以包含许多相同的记录user_id,例如,返回行的查询仅user_id=110提供以下结果:

SELECT *
FROM blocks
WHERE confirms > 520
      AND user_id = 110;

+ ------- + ------------ + ----------- + ------------- +
| id      | user_id      | reward      | confirms      |
+ ------- + ------------ + ----------- + ------------- +
| 0       | 110          | 20.89832115 | 521           |
| 65174   | 110          | 3.80357075  | 698           |
| 65204   | 110          | 4.41933060  | 668           |
| 65218   | 110          | 4.69059801  | 654           |
| 65219   | 110          | 4.70222521  | 653           |
| 65230   | 110          | 4.82805490  | 642           |
| 65265   | 110          | 5.25058079  | 607           |
| 65316   | 110          | 6.17262650  | 556           |
+ ------- + ------------ + ----------- + ------------- +

straigh连接(和LEFT/RIGHT外连接)以这种方式工作,它从第一个连接表中获取每个记录,并将此记录(组合它)与来自另一个连接表的所有行配对,以满足连接条件.

在我们的例子中,左连接产生以下结果集:

SELECT *
FROM blocks
LEFT JOIN payout_history
ON blocks.user_id = payout_history.user_id
WHERE confirms > 520
    AND blocks.user_id = 110;
+ ------- + ------- + ----------- + -------- + --- + ------- + ----------- +
| id      | user_id | reward      | confirms | id  | user_id | amount      |
+ ------- + ------- + ----------- + -------- + --- + ------- + ----------- +
| 0       | 110     | 20.89832115 | 521      | 1   | 110     | 20.898319   |
| 65174   | 110     | 3.80357075  | 698      | 1   | 110     | 20.898319   |
| 65204   | 110     | 4.41933060  | 668      | 1   | 110     | 20.898319   |
| 65218   | 110     | 4.69059801  | 654      | 1   | 110     | 20.898319   |
| 65219   | 110     | 4.70222521  | 653      | 1   | 110     | 20.898319   |
| 65230   | 110     | 4.82805490  | 642      | 1   | 110     | 20.898319   |
| 65265   | 110     | 5.25058079  | 607      | 1   | 110     | 20.898319   |
| 65316   | 110     | 6.17262650  | 556      | 1   | 110     | 20.898319   |
+ ------- + ------- + ----------- + -------- + --- + ------- + ----------- +

现在,如果我们添加SUM( amount ) .... GROUP BY user_id,MySql将从amount上面的结果集中计算所有值的总和(8行*20.898 = ~167.184)

SELECT blocks.user_id, sum( amount)
FROM blocks
LEFT JOIN payout_history
ON blocks.user_id = payout_history.user_id
WHERE confirms > 520
    AND blocks.user_id = 110
GROUP BY blocks.user_id;
+ ------------ + ----------------- +
| user_id      | sum( amount)      |
+ ------------ + ----------------- +
| 110          | 167.186554        |
+ ------------ + ----------------- +

正如你在这种情况下看到的那样,连接没有给我们想要的结果 - 我们需要一些名字a semi join- 下面是半连接的不同变体,试试看:

SELECT bl.user_id, SUM( ph.amount ) PAIDOUT
FROM (
   SELECT distinct blocks.user_id 
   FROM blocks
   WHERE confirms > 520
) bl
LEFT JOIN  payout_history ph
ON bl.user_id = ph.user_id
GROUP BY ph.user_id
;


SELECT ph.user_id, SUM( ph.amount ) PAIDOUT
FROM payout_history ph
WHERE ph.user_id IN (
     SELECT user_id FROM blocks
     WHERE confirms > 520
  )
GROUP BY ph.user_id
;

SELECT ph.user_id, SUM( ph.amount ) PAIDOUT
FROM payout_history ph
WHERE EXISTS (
     SELECT 1 FROM blocks bl
     WHERE bl.user_id = ph.user_id
        AND bl.confirms > 520
  )
GROUP BY ph.user_id
;