Question

我有下表：

id   | timestamp  |  team  
----------------------------
1    | 2016-05-06 |  A
2    | 2016-03-02 |  A
3    | 2015-12-01 |  A
4    | 2016-07-05 |  B
5    | 2016-06-30 |  B
6    | 2016-06-28 |  B
7    | 2016-04-05 |  C
8    | 2016-04-02 |  C
9    | 2016-01-02 |  C

我想按团队分组并找到每个团队的最后时间戳，所以我做了：

select team, max(timestamp) from my_table group by team

到目前为止一切正常。但是，现在我想知道每个团队上个月有多少不同的id。例如，对于A队，它将从2016-04-07到2016-05-06，所以这样的计数是1.对于B队，上个月是2016-06-06到2016-07-05，所以计数是3.对于C队，上个月是2016-03-06到2016-04-05，计数是2.我的预期输出应该是这样的：

team  | max(timestamp)  |  count_in_last_month
------------------------------------------------
A     | 2016-05-06      |  1
B     | 2016-07-05      |  3
C     | 2016-04-05      |  2

可以使用Impala查询导出吗？谢谢！

Answer 1

使用获取最大时间戳的子查询加入原始表。

SELECT t1.team, t2.month_end, COUNT(DISTINCT t1.id) AS count_in_last_month
FROM my_table AS t1
JOIN (SELECT team, MAX(timestamp) AS month_end
      FROM my_table
      GROUP BY team) AS t2
ON t1.team = t2.team 
AND t1.timestamp BETWEEN DATE_SUB(month_end, INTERVAL 1 MONTH) AND month_end
GROUP BY t1.team, t2.month_end

DEMO

Impala / SQL：我可以为每个组设置不同的时间段吗？

1 个答案: