Condition in GROUP_CONCAT selection

Question

I have 3 tables and I'm joining them to get some data.

-----------------
Table Name: users
-------------------------------
|user_id  | user_name |
-------------------------------
123      | abc
-------------------------------
223      | bcd
-------------------------------
323      | cde
-------------------------------

-----------------
Table Name: limit
-------------------------------
user_id  | limit_id
-------------------------------
123      | 1
-------------------------------
223      | 2
-------------------------------
323      | 3
-------------------------------
323      | 4
-------------------------------

-------------------------
Table Name: limit_setting
-------------------------------
limit_id | date_limit
-------------------------------
1        | 2016-09-29 12:00:00
-------------------------------
2        | 2016-09-28 12:00:00
-------------------------------
3        | 2016-09-27 12:00:00
-------------------------------
1        | 2016-09-27 12:00:00
-------------------------------
1        | 2016-09-24 12:00:00
-------------------------------
4        | 2016-09-25 12:00:00
-------------------------------
4        | 2016-09-26 12:00:00
-------------------------------

I need to get a result like this. I am stuck with the GROUP_CONCAT for the dates column. The date column should have all entries other than the MAX date. If there is only one entry in the limit_setting table for that limit_id then it shouldn't show anything for that user. count_dates : its the number of entries which are there in the limit_setting table.


Desired output

----------------------------------------------------------------------
user_name | dates                                       | count_dates   
----------------------------------------------------------------------
abc       | 2016-09-27 12:00:00 , 2016-09-24 12:00:00   | 3
----------------------------------------------------------------------
bcd       |                                             | 1 
----------------------------------------------------------------------
cde       |                                             | 1 
-----------------------------------------------------------------------
cde       | 2016-09-26 12:00:00                         | 2 
-----------------------------------------------------------------------

SELECT PP.`user_name`, count(ESL.Limit_id) as count_dates,
  GROUP_CONCAT(ESL.date_limit SEPARATOR ',') as dates 
FROM users as PP INNER JOIN `limit` as PAL ON PP.Id = PAL.PlayerId
LEFT JOIN limit_setting as ESL ON ESL.LimitId = PAL.limitId 
GROUP BY PAL.limitId

Additionally i tried with (which returned nothing)

SELECT ESL.date_limit, MAX(date_limit) as max_date, PP.`user_name`, count(ESL.Limit_id) as count_dates, 
  GROUP_CONCAT(ESL.date_limit SEPARATOR ',') as dates 
FROM users as PP INNER JOIN `limit` as PAL ON PP.Id = PAL.PlayerId 
LEFT JOIN limit_setting as ESL ON ESL.LimitId = PAL.limitId 
GROUP BY PAL.limitId
HAVING ESL.date_limit > max_date

I tried with Find_in_set but not sure how to use it effectively.

Giorgos Betsos · Accepted Answer · 2016-09-29T09:31:52.243

Try this:

SELECT user_name,        
       CASE WHEN COUNT(*) > 1 
            THEN SUBSTRING_INDEX(GROUP_CONCAT(date_limit ORDER BY date_limit), 
                                 ',', COUNT(*) - 1) 
            ELSE ''
       END AS dates,             
       COUNT(*) AS count_dates
FROM users as PP 
INNER JOIN `limit` as PAL ON PP.user_id = PAL.user_id
LEFT JOIN limit_setting as ESL ON ESL.limit_id = PAL.limit_id
GROUP BY user_name

The query uses SUBSTRING_INDEX function in order to get all dates returned by GROUP_CONCAT except from the last date. Using ORDER BY inside GROUP_CONCAT we can place the maximum date at the end, so that SUBSTRING_INDEX truncates exactly this date.

Demo here

Thank you so much for the help and also for the explanation... I never used `SUBSTRING_INDEX` before and had no idea about it... thanks once again :) — user2595861, Sep 29 '16 at 09:40

score 0 · Answer 2 · answered Sep 29 '16 at 09:00

I would use a subquery to return the max date per limit and join this back to the limit setting table to eliminate the maximum dates from the group_concat() with a case expression:

SELECT 
    PP.`user_name`,
    count(ESL.Limit_id) as count_dates,
    GROUP_CONCAT(CASE
                     WHEN ESL.date_limit<>ls.maxdate THEN ESL.date_limit
                     ELSE ''
                 END SEPARATOR ',') as dates 
FROM users as PP
INNER JOIN `limit` as PAL ON PP.Id = PAL.PlayerId
LEFT JOIN limit_setting as ESL ON ESL.LimitId = PAL.limitId
LEFT JOIN (SELECT LimitId, MAX(date_limit) as maxdate
           FROM limit_setting
           GROUP BY LimitId
          ) ls ON ESL.LimitId=ls.LimitId
GROUP BY PP.`user_name`

I got the result with this as well but with a ',' in the beginning of the `dates` field... like this : `,2016-09-23 00:00:00,2016-09-21 00:00:00,2016-09-2..` — user2595861, Sep 29 '16 at 09:51

Condition in GROUP_CONCAT selection

2 Answers2