Question

我有一个关于操作符优先级的SQL查询，

SELECT
foo,
count(*)
FROM
A
JOIN (SELECT
  SUM(IF(bar = 2,1,0)) as bar_sum,
  SUM(IF(foo >= 1,1,0)) as foo,
  SUM(1) as sum_1
  FROM
  B
  ) as sums
GROUP BY
id,
bar_sum,
foo,
sum_1
ON A.id = B.id

外部括号中的GROUP BY真的适用于内部括号吗？注意，我需要将此SQL从Hive移植到Spark scala Dataframe API，所以我确实需要正确设置操作符优先级。从What is the execution sequence of Group By, Having and Where clause in SQL Server?看来，这通常是正确的，但我没有找到有关()的任何文档。

Answer 1

此查询看起来不正确（缺少ON子句）

SELECT foo,count(*)
FROM A
JOIN (SELECT SUM(IF(bar = 2,1,0)) as bar_sum,
             SUM(IF(foo >= 1,1,0)) as foo,
             SUM(1) as sum_1
      FROM   B) as sums  -- should be `ON`
GROUP BY id, bar_sum, foo, sum_1;
        -- looks like grouping by sum_1, bar_sum is superflous

GROUP BY仅适用于外部查询。请注意，内部查询将返回单行。

SELECT SUM(IF(bar = 2,1,0)) as bar_sum,
       SUM(IF(foo >= 1,1,0)) as foo,
       SUM(1) as sum_1
FROM   B
-- single row

然后将单行连接到表A，并获得与不同的（id，foo）值一样多的行。

带括号的SQL运算符优先级

1 个答案: