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
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:
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.
Here is one option (untested):
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 JOIN
on 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.
I see answers with JOINS
and Subquerys
, but I believe a simple HAVING
clause should do the trick:
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.
EDIT 2: Is it even needed to group by id_event when you're already filtering it to 1 event?
EDIT 3: Don't know why the downvote, this sql fiddle shows it works.
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.
The following works, but unfortunately it has a subquery again and won't perform much better than the other solutions posted here.
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;
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.
Could use group_concat_ws to account for possible null fields.
Use a simple group by id_member, but select:
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.
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
select concat(from_unixtime(timestamp),attendance), timestamp, attendance
to get a better idea of how it uses max to get the right attendance.
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.