SQL/Impala：将多个查询（具有不同的 where 子句）合并为一个

Posted 2023-03-22

【中文标题】SQL/Impala：将多个查询（具有不同的 where 子句）合并为一个

【英文标题】：SQL/Impala: combined multiple query (with different where clause) into one

【发布时间】：2016-08-16 22:36:21

【问题描述】：

我有以下疑问：

'select team, count(distinct id) as distinct_id_count_w1 from myTable where timestamp > t1  and timestamp < t2 group by team'

'select team, count(distinct id) as distinct_id_count_w2 from myTable where timestamp > t2  and timestamp < t3 group by team'

是否可以将这两个查询合二为一？谢谢！

【参考方案1】：

很容易 :) 这应该适用于大多数常见的数据库引擎：

select team, count(distinct id) as distinct_id_count_w1, null as distinct_id_count_w2 from myTable where timestamp > t1  and timestamp < t2 group by team

UNION ALL

select team, null as distinct_id_count_w1, count(distinct id) as distinct_id_count_w2 from myTable where timestamp > t2  and timestamp < t3 group by team

正如毛豆所说，您可能希望阅读每个团队的两个结果。问题本身并不清楚，但可以这样解决：

SELECT
    COALESCE(interval1.team interval2.team) AS team,
    interval1.distinct_id_count_w1,
    interval2.distinct_id_count_w2
FROM (
    select team, count(distinct id) as distinct_id_count_w1 from myTable where timestamp > t1  and timestamp < t2 group by team
) AS interval1
FULL OUTER JOIN
(
    select team, count(distinct id) as distinct_id_count_w2 from myTable where timestamp > t2  and timestamp < t3 group by team
) AS interval2
ON interval1.team IS NULL OR interval2.team IS NULL OR interval1.team = interval2.team

【讨论】：

但结果没有加入使用团队。一半的行具有 distinct_id_count_w2 = None，另一半具有 distinct_id_count_w1 = None。每行都应该有一个有效值 distinct_id_count_w1 和一个有效值 distinct_id_count_w2

【参考方案2】：

如果你认为返回的结果不同，你应该使用“UNION ALL”，因为你只使用“UNION”，sql会区分结果以影响查询的性能