按X天的范围分组

Question

我有一组记录，希望按照某个范围进行计数和分组，例如，我想要按照X天的时间段来计算被创建的记录数。

e.g. SELECT COUNT(*) FROM `table` GROUP BY /*`created` 3 days/*

- Lizard

Answer 1

以下是一个与日期相关的示例。

create table t1(created date not null);

insert 
  into t1(created) values (date '2011-01-09')
                         ,(date '2011-01-10')
                         ,(date '2011-01-11')
                         ,(date '2011-01-12')
                         ,(date '2011-01-13')
                         ,(date '2011-01-14')
                         ,(date '2011-01-15')
                         ,(date '2011-01-16')
                         ,(date '2011-01-17')
                         ,(date '2011-01-18')
                         ,(date '2011-01-19')
                         ,(date '2011-01-20');

select floor(datediff(now(), created) / 3) * 3 as days_ago
      ,min(created)
      ,max(created)
      ,count(*)
  from t1
 group 
    by floor(datediff(now(), created) / 3);

+----------+--------------+--------------+----------+
| days_ago | min(created) | max(created) | count(*) |
+----------+--------------+--------------+----------+
|        0 | 2011-01-18   | 2011-01-20   |        3 |
|        3 | 2011-01-15   | 2011-01-17   |        3 |
|        6 | 2011-01-12   | 2011-01-14   |        3 |
|        9 | 2011-01-09   | 2011-01-11   |        3 |
+----------+--------------+--------------+----------+
4 rows in set (0.00 sec)

- Ronnis

Answer 2

您可以执行类似以下代码的操作
SELECT COUNT(*) FROM table GROUP BY FLOOR(created / 3)

... 我认为是这样的。

尽管如果created是日期字段，则需要进行一些调整，将其转换为数字值才能使此操作生效。

- Sam Dufel

Answer 3

感谢 @Ronnis，我使用了你的示例，最终解决了我的问题。

还有一个小错误我发现了，在示例中，我添加了一行。

insert into t1(created) values (date '2011-01-21')

现在我明白了：

+----------+--------------+--------------+----------+
| days_ago | min(created) | max(created) | count(*) |
+----------+--------------+--------------+----------+
|     1986 | 2011-01-20   | 2011-01-21   |        2 |
|     1989 | 2011-01-17   | 2011-01-19   |        3 |
|     1992 | 2011-01-14   | 2011-01-16   |        3 |
|     1995 | 2011-01-11   | 2011-01-13   |        3 |
|     1998 | 2011-01-09   | 2011-01-10   |        2 |
+----------+--------------+--------------+----------+

正如您所看到的，这些天被分为2、3、3、3、2，这不是我所期望的。

我将SQL更改为

select floor(datediff(created, '2011-01-09') / 3) * 3 as days_before
      ,min(created)
      ,max(created)
      ,count(*)
from t1
group by floor(datediff(created, '2011-01-09') / 3);

只需从创建日期到开始日期获取日期差，现在我得到了：

+-------------+--------------+--------------+----------+
| days_before | min(created) | max(created) | count(*) |
+-------------+--------------+--------------+----------+
|           0 | 2011-01-09   | 2011-01-11   |        3 |
|           3 | 2011-01-12   | 2011-01-14   |        3 |
|           6 | 2011-01-15   | 2011-01-17   |        3 |
|           9 | 2011-01-18   | 2011-01-20   |        3 |
|          12 | 2011-01-21   | 2011-01-21   |        1 |
+-------------+--------------+--------------+----------+

那可能更合适。

- Peigong