在 mysql 中使用 group by 仅选择最后一个值

Select only last value using group by at mysql

提问by Cito

I have one table with data about attendance into some events. I have in the table the data of the attendance everytime the user sends new attendance, the information is like this:

我有一张表格，其中包含有关参加某些活动的数据。我在表中有用户每次发送新考勤的考勤数据，信息如下：

mysql> SELECT id_branch_channel, id_member, attendance, timestamp, id_member FROM view_event_attendance WHERE id_event = 782;
+-------------------+-----------+------------+------------+-----------+
| id_branch_channel | id_member | attendance | timestamp  | id_member |
+-------------------+-----------+------------+------------+-----------+
|              1326 |    131327 |        459 | 1363208604 |    131327 |
|              1326 |    131327 |        123 | 1363208504 |    131327 |
|              1326 |    131327 |          1 | 1363208459 |    131327 |
|              1326 |     93086 |          0 |       NULL |     93086 |
|              1326 |     93087 |          0 |       NULL |     93087 |
|              1326 |     93088 |          0 |       NULL |     93088 |
|              1326 |     93093 |          0 |       NULL |     93093 |
|              1326 |     99113 |          0 |       NULL |     99113 |
|              1326 |     99135 |          0 |       NULL |     99135 |
|              1326 |     99199 |          0 |       NULL |     99199 |
|              1326 |     99200 |          0 |       NULL |     99200 |
|              1326 |    131324 |          0 |       NULL |    131324 |
|              1326 |     85850 |          0 |       NULL |     85850 |
|              1326 |     93085 |          0 |       NULL |     93085 |
+-------------------+-----------+------------+------------+-----------+
14 rows in set (0.00 sec)

(This is actually a view, for that reason some of the fields are null).

（这实际上是一个视图，因此某些字段为空）。

I can groupby id_member so I get only one row for every member (that is, only the last attendance the user sent). However, when I do it, I received the first attendance the user sent, not the last one.

我可以通过 id_member 分组，因此每个成员只能获得一行（即，只有用户发送的最后一次出席）。但是，当我这样做时，我收到了用户发送的第一个出席，而不是最后一个。

mysql> SELECT id_branch_channel, id_member, attendance, timestamp, id_member FROM view_event_attendance WHERE id_event = 782 GROUP BY id_event,id_member;
+-------------------+-----------+------------+------------+-----------+
| id_branch_channel | id_member | attendance | timestamp  | id_member |
+-------------------+-----------+------------+------------+-----------+
|              1326 |    131327 |          1 | 1363208459 |    131327 |
|              1326 |     93086 |          0 |       NULL |     93086 |
|              1326 |    131324 |          0 |       NULL |    131324 |
|              1326 |     93087 |          0 |       NULL |     93087 |
|              1326 |     93088 |          0 |       NULL |     93088 |
|              1326 |     93093 |          0 |       NULL |     93093 |
|              1326 |     99113 |          0 |       NULL |     99113 |
|              1326 |     99135 |          0 |       NULL |     99135 |
|              1326 |     85850 |          0 |       NULL |     85850 |
|              1326 |     99199 |          0 |       NULL |     99199 |
|              1326 |     93085 |          0 |       NULL |     93085 |
|              1326 |     99200 |          0 |       NULL |     99200 |
+-------------------+-----------+------------+------------+-----------+
12 rows in set (0.00 sec)

I already tried to add ORDER BY clausules, but they are not working at all... any ideas?

我已经尝试添加 ORDER BY clausules，但它们根本不起作用......有什么想法吗？

Thanks in advance!

提前致谢！

Edit: this is the script that creates the table

编辑：这是创建表的脚本

CREATE OR REPLACE VIEW view_event_attendance 
    AS
        SELECT 
            tbl_event.id_event,
            tbl_member_event.id_member,
            tbl_event.id_branch_channel,
            tbl_member_event_attendance.id_member_event_attendance,
            IF(ISNULL(tbl_member_event_attendance.attendance), 0, tbl_member_event_attendance.attendance) AS attendance,
            tbl_member_event_attendance.timestamp
        FROM 
            tbl_event
            INNER JOIN 
                tbl_member_event ON tbl_member_event.id_event = tbl_event.id_event
                LEFT OUTER JOIN
                    tbl_member_event_attendance ON tbl_member_event_attendance.id_member_event = tbl_member_event.id_member_event
        ORDER BY 
            tbl_member_event_attendance.timestamp DESC;

EDIT 2:

编辑2：

Thanks a lot MichaelBenjamin, but the problem when using subqueries is the size of the view:

非常感谢 MichaelBenjamin，但是使用子查询时的问题是视图的大小：

mysql> DESCRIBE SELECT id_branch_channel, id_member, attendance, timestamp, id_member 
    -> FROM (select * from view_event_attendance order by timestamp desc) as whatever
    -> WHERE id_event = 782 
    -> GROUP BY id_event,id_member;
+----+-------------+-----------------------------+--------+-----------------+-----------------+---------+------------------------------------------------+-------+----------------------------------------------+
| id | select_type | table                       | type   | possible_keys   | key             | key_len | ref                                            | rows  | Extra                                        |
+----+-------------+-----------------------------+--------+-----------------+-----------------+---------+------------------------------------------------+-------+----------------------------------------------+
|  1 | PRIMARY     | <derived2>                  | ALL    | NULL            | NULL            | NULL    | NULL                                           | 16755 | Using where; Using temporary; Using filesort |
|  2 | DERIVED     | tbl_member_event            | index  | id_event        | id_event        | 8       | NULL                                           | 16346 | Using index; Using temporary; Using filesort |
|  2 | DERIVED     | tbl_event                   | eq_ref | PRIMARY         | PRIMARY         | 4       | video_staging.tbl_member_event.id_event        |     1 |                                              |
|  2 | DERIVED     | tbl_member_event_attendance | ref    | id_event_member | id_event_member | 4       | video_staging.tbl_member_event.id_member_event |     1 | Using index                                  |
+----+-------------+-----------------------------+--------+-----------------+-----------------+---------+------------------------------------------------+-------+----------------------------------------------+
4 rows in set (0.08 sec)

As you can see there are a lot of rows in my table, so for that reason I don't want to use subqueries...

如您所见，我的表中有很多行，因此我不想使用子查询...

EDIT 3:

编辑 3：

But adding WHERE to the subquery it looks better...

但是将 WHERE 添加到子查询中看起来更好......

mysql> DESCRIBE SELECT id_branch_channel, id_member, attendance, timestamp, id_member 
    -> FROM (select * from view_event_attendance where id_event = 782 order by timestamp desc) as whatever
    -> WHERE id_event = 782 
    -> GROUP BY id_event,id_member;
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+----------------------------------------------+
| id | select_type | table                       | type  | possible_keys   | key             | key_len | ref                                            | rows | Extra                                        |
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+----------------------------------------------+
|  1 | PRIMARY     | <derived2>                  | ALL   | NULL            | NULL            | NULL    | NULL                                           |   14 | Using where; Using temporary; Using filesort |
|  2 | DERIVED     | tbl_event                   | const | PRIMARY         | PRIMARY         | 4       |                                                |    1 | Using temporary; Using filesort              |
|  2 | DERIVED     | tbl_member_event            | ref   | id_event        | id_event        | 4       |                                                |   12 | Using index                                  |
|  2 | DERIVED     | tbl_member_event_attendance | ref   | id_event_member | id_event_member | 4       | video_staging.tbl_member_event.id_member_event |    1 | Using index                                  |
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+----------------------------------------------+
4 rows in set (0.01 sec)

If I can not find anything else not using subqueries, I think I'll choose this as the answer...

如果我找不到任何不使用子查询的东西，我想我会选择这个作为答案......

Edit 4

编辑 4

After seeing the comments in the answer, I've decided to select another as the answer. Here is the DESCRIBE for both queries, and I think it is obvious what is the best solution:

在看到答案中的评论后，我决定选择另一个作为答案。这是两个查询的 DESCRIBE，我认为最好的解决方案是显而易见的：

mysql> DESCRIBE SELECT 
    ->   id_branch_channel,
    ->   id_member, 
    ->   attendance, 
    ->   timestamp,
    ->   id_member
    -> FROM view_event_attendance AS t1
    -> WHERE id_event = 782
    -> AND timestamp = (SELECT MAX(timestamp)
    ->                  FROM view_event_attendance AS t2 
    ->                  WHERE t1.id_member = t2.id_member 
    ->                    AND t1.id_event = t2.id_event 
    ->                  GROUP BY id_event, id_member)
    -> OR timestamp IS NULL
    -> GROUP BY id_event, id_member;
+----+--------------------+-----------------------------+--------+--------------------+--------------------------+---------+------------------------------------------------+------+-----------------------------------------------------------+
| id | select_type        | table                       | type   | possible_keys      | key                      | key_len | ref                                            | rows | Extra                                                     |
+----+--------------------+-----------------------------+--------+--------------------+--------------------------+---------+------------------------------------------------+------+-----------------------------------------------------------+
|  1 | PRIMARY            | tbl_event                   | index  | PRIMARY            | id_member_branch_channel | 4       | NULL                                           |  208 | Using index; Using temporary; Using filesort              |
|  1 | PRIMARY            | tbl_member_event            | ref    | id_event           | id_event                 | 4       | video_staging.tbl_event.id_event               |   64 | Using index                                               |
|  1 | PRIMARY            | tbl_member_event_attendance | ref    | id_event_member    | id_event_member          | 4       | video_staging.tbl_member_event.id_member_event |    1 | Using where; Using index                                  |
|  2 | DEPENDENT SUBQUERY | tbl_event                   | eq_ref | PRIMARY            | PRIMARY                  | 4       | func                                           |    1 | Using where; Using index; Using temporary; Using filesort |
|  2 | DEPENDENT SUBQUERY | tbl_member_event            | eq_ref | id_event,id_member | id_event                 | 8       | video_staging.tbl_event.id_event,func          |    1 | Using where; Using index                                  |
|  2 | DEPENDENT SUBQUERY | tbl_member_event_attendance | ref    | id_event_member    | id_event_member          | 4       | video_staging.tbl_member_event.id_member_event |    1 | Using where; Using index                                  |
+----+--------------------+-----------------------------+--------+--------------------+--------------------------+---------+------------------------------------------------+------+-----------------------------------------------------------+
6 rows in set (0.00 sec)


mysql> DESCRIBE SELECT *
    -> FROM (SELECT id_branch_channel, id_member, attendance, timestamp, id_event 
    ->       FROM view_event_attendance 
    ->       WHERE id_event = 782 
    ->       ORDER BY timestamp desc
    ->      ) as whatever
    -> GROUP BY id_event,id_member;
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
| id | select_type | table                       | type  | possible_keys   | key             | key_len | ref                                            | rows | Extra                           |
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
|  1 | PRIMARY     | <derived2>                  | ALL   | NULL            | NULL            | NULL    | NULL                                           |   14 | Using temporary; Using filesort |
|  2 | DERIVED     | tbl_event                   | const | PRIMARY         | PRIMARY         | 4       |                                                |    1 | Using temporary; Using filesort |
|  2 | DERIVED     | tbl_member_event            | ref   | id_event        | id_event        | 4       |                                                |   12 | Using index                     |
|  2 | DERIVED     | tbl_member_event_attendance | ref   | id_event_member | id_event_member | 4       | video_staging.tbl_member_event.id_member_event |    1 | Using index                     |
+----+-------------+-----------------------------+-------+-----------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
4 rows in set (0.00 sec)

采纳答案by ysth

Use a simple group by id_member, but select:

使用 id_member 的简单组，但选择：

substring(max(concat(from_unixtime(timestamp),attendance)) from 20) as attendance

This attaches attendance to the timestamp for each row in a group, in order to be able to select the desired timestamp/attendance with max() and then extract just the attendance.

这将出勤附加到组中每一行的时间戳，以便能够使用 max() 选择所需的时间戳/出勤，然后仅提取出勤。

What concat()returns is 19 characters of formatted timestamp (YYYY-mm-dd HH:MM:SS) with the attendance appended starting at character 20; the substring(... from 20)gets just the attendance from the (stringwise) maximum one for the group. You can remove the group by and just

什么concat()回报是格式化的时间戳记的19个字符（YYYY-MM-DD HH：MM：SS）与所述出席所附起始于字符20; 在substring(... from 20)获得距离（stringwise）的最大一个该组的上座率。您可以通过以下方式删除组

select concat(from_unixtime(timestamp),attendance), timestamp, attendance

to get a better idea of how it uses max to get the right attendance.

更好地了解它如何使用 max 来获得正确的出勤率。

回答by Sarang

SUBSTRING_INDEX(SUBSTRING_INDEX(group_concat(%requiredfield%), ',', count(*)),',',-1)

This will get the last value of the 'required field' from any group_concat, if unsorted it will be the last value in the table by default.

这将从任何 group_concat 获取“必填字段”的最后一个值，如果未排序，默认情况下它将是表中的最后一个值。

Could use group_concat_ws to account for possible null fields.

可以使用 group_concat_ws 来说明可能的空字段。

回答by Michael Benjamin

SELECT id_branch_channel, id_member, attendance, timestamp, id_member 
FROM (select * from view_event_attendance order by timestamp desc) as whatever
WHERE id_event = 782 
GROUP BY id_event,id_member;

EDIT:This may yield better performance:

编辑：这可能会产生更好的性能：

SELECT *
FROM (SELECT id_branch_channel, id_member, attendance, timestamp, id_member 
      FROM view_event_attendance 
      WHERE id_event = 782 
      ORDER BY timestamp desc
     ) as whatever
GROUP BY id_event,id_member;

As long as the result-set can fit into the Innodb_buffer_pool, you will not see a significant performance drop.

只要结果集可以放入 Innodb_buffer_pool，您就不会看到显着的性能下降。

回答by JodyT

I see answers with JOINSand Subquerys, but I believe a simple HAVINGclause should do the trick:

我用JOINSand看到答案Subquerys，但我相信一个简单的HAVING子句应该可以解决问题：

SELECT 
  id_branch_channel,
  id_member, 
  attendance, 
  timestamp,
  id_member
FROM view_event_attendance
WHERE id_event = 782 
GROUP BY id_event, id_member
HAVING MAX(timestamp) OR timestamp IS NULL;

EDIT:Added a check for IS NULL if you also want to include those rows.

编辑：如果您还想包含这些行，则添加了对 IS NULL 的检查。

EDIT 2:Is it even needed to group by id_event when you're already filtering it to 1 event?

编辑 2：当您已经将其过滤为 1 个事件时，是否还需要按 id_event 进行分组？

EDIT 3:Don't know why the downvote, this sql fiddleshows it works.

编辑3：不知道为什么downvote，这个sql fiddle显示它有效。

EDIT 4:I have to apologise, @ysth is correct, the SQL Fiddle does not work correctly. I deserved the -1, but when you down vote at least explain why so I can learn something myself as well.

编辑 4：我必须道歉，@ysth 是正确的，SQL Fiddle 无法正常工作。我应得的 -1，但是当你投反对票时，至少解释一下原因，这样我自己也可以学到一些东西。

The following works, but unfortunately it has a subquery again and won't perform much better than the other solutions posted here.

以下工作，但不幸的是它再次有一个子查询，并且不会比这里发布的其他解决方案好得多。

SELECT 
  id_branch_channel,
  id_member, 
  attendance, 
  timestamp,
  id_member
FROM view_event_attendance AS t1
WHERE id_event = 782
AND timestamp = (SELECT MAX(timestamp)
                 FROM view_event_attendance AS t2 
                 WHERE t1.id_member = t2.id_member 
                   AND t1.id_event = t2.id_event 
                 GROUP BY id_event, id_member)
OR timestamp IS NULL
GROUP BY id_event, id_member;

回答by sgeddes

Here is one option (untested):

这是一种选择（未经测试）：

SELECT v.id_branch_channel, v.id_member, v.attendance, v.timestamp, v.id_member 
FROM view_event_attendance v
    JOIN (
        SELECT id_event, id_member, MAX(attendance) maxattendance
        FROM view_event_attendance 
        GROUP BY id_event, id_member ) m ON 
            v.id_event = m.id_event AND
            v.id_member = m.id_member AND
            v.attendance = m.maxattendance
WHERE v.id_event = 782 
GROUP BY v.id_member;

The concept is to get the MAX()of timestamp and use that field to JOINon your view. You might not need all the fields -- really depends on your table structure. But this should get you going in the correct direction.

这个概念是获取MAX()时间戳并将该字段用于JOIN您的视图。您可能不需要所有字段——实际上取决于您的表结构。但这应该让你朝着正确的方向前进。

回答by Steve

One way to do this is to use a window function and a subquery, if you add an entry to your select list as row_number() over (partition by id_member order by timestamp desc)this will resolve to a number ordering the rows by timestamp (with 1 being the oldest) grouped in each id_member group (run it if this doesn't make sense, it will be clear). You can then select from this as a subquery where the extra column = 1 which will only select the rows with the highest timestamp within each group.

执行此操作的一种方法是使用窗口函数和子查询，如果您将条目添加到您的选择列表，因为row_number() over (partition by id_member order by timestamp desc)这将解析为按时间戳对行进行排序的数字（1 是最旧的）分组在每个 id_member 组中（运行如果这没有意义，那就很清楚了）。然后，您可以从中选择作为子查询，其中额外的列 = 1，这将只选择每个组中时间戳最高的行。