我没有Spark测试，但一种方法应该是使用 HAVING 找出组中是否有“签入”行；

SELECT `Cust No`, SUM(CASE WHEN Action='PAID' THEN val END) paid 
FROM mytable 
GROUP BY `Cust No` 
HAVING MAX(CASE WHEN Action='Checked In' THEN 1 ELSE 0 END) > 0

+---------+------+
| Cust No | paid |
+---------+------+
|      10 |   40 |
|      30 |   60 |
+---------+------+

另一种是使用 WHERE 先求群，然后求和；

SELECT `Cust No`, SUM(val) paid 
FROM mytable 
WHERE action='PAID' AND `Cust No` IN (
  SELECT `Cust No` FROM mytable WHERE action='Checked In'
) 
GROUP BY `Cust No`

使用 CASE 条件以及 SUM 喜欢

sum(case when Action = 'paid' then val end)

如果使用外部查询稍微调整一下查询，比如

select customer_number, PaymentDone
from (
select customer_number, 
sum(case when Action = 'paid' then val end) as PaymentDone,
group_concat(Action ORDER BY Action asc) as Action_List 
from tbl1
group by customer_number) xxx
where Action_List like 'Checked In%';

这应该管用

select customer_number, sum(value) as total_paid 
from table_name where action = 'PAID' group by customer_number;

这应该起作用：

select customer.customer_number, sum(value) as total_paid  from customers
    left join
    (
      select customer_number from users where action = 'Checked In' group by customer_number
    )checkins on checkins.customer_number = customers.customer_number 
    where checkins.customer_number is not null and customers.action = 'PAID'
    group by customers.customer_number