specifying preferred results for GROUP BY values - mysql

Question

I have the following (simplified) query:
SELECT
u.idnumber AS user_idnumber,
sst_status.value AS scorm_status,
sst_starttime.value AS scorm_starttime,
sst_duration.value AS scorm_duration,
sst_score.value AS scorm_score,
u.id as uid,
s.id as scormid,
ss.id as ssid
FROM {user} u
LEFT JOIN {prog_user_assignment} prua ON prua.userid = u.id
LEFT JOIN {prog} pr ON prua.programid = pr.id
LEFT JOIN {prog_courseset} prcs ON prcs.programid = pr.id
LEFT JOIN {prog_courseset_course} prcsc ON prcsc.coursesetid = prcs.id
LEFT JOIN {course} c ON prcsc.courseid = c.id
LEFT JOIN {scorm} s ON s.id = cm.instance
LEFT JOIN {scorm_scoes} ss ON ss.scorm = s.id
LEFT JOIN {scorm_scoes_track} sst_status
ON sst_status.userid = u.id
AND sst_status.scormid = ss.scorm
AND sst_status.element = 'cmi_core.lesson_status'
AND sst_status.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_starttime
ON sst_starttime.userid = u.id
AND sst_starttime.scormid = ss.scorm
AND sst_starttime.element = 'x.start.time'
AND sst_starttime.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_duration
ON sst_duration.userid = u.id
AND sst_duration.scormid = ss.scorm
AND sst_duration.element = 'cmi.core.total_time'
AND sst_duration.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_score
ON sst_score.userid = u.id
AND sst_score.scormid = ss.scorm
AND sst_score.element = 'cmi.core.score.raw'
AND sst_score.scoid = ss.id
WHERE u.id IN([SOME_EXAMPLE_IDS_HERE]) AND u.deleted = 0
group by u.idnumber, c.id
Now the problem is that without the GROUP BY clause, some of the records have values for scorm_status, scorm_starttime etc. However, these contain duplicates. But, with the group by statement, only NULL values are returned for these columns. Does anyone have a clue on how to resolve this?

Answer 1

Using GROUP_CONCAT and SUBSTRING_INDEX:-
SELECT
u.idnumber AS user_idnumber,
SUBSTRING_INDEX(GROUP_CONCAT(sst_status.value), ',', 1) AS scorm_status,
SUBSTRING_INDEX(GROUP_CONCAT(sst_starttime.value), ',', 1) AS scorm_starttime,
SUBSTRING_INDEX(GROUP_CONCAT(sst_duration.value), ',', 1) AS scorm_duration,
SUBSTRING_INDEX(GROUP_CONCAT(sst_score.value), ',', 1) AS scorm_score,
u.id as uid,
s.id as scormid,
ss.id as ssid
FROM {user} u
LEFT JOIN {prog_user_assignment} prua ON prua.userid = u.id
LEFT JOIN {prog} pr ON prua.programid = pr.id
LEFT JOIN {prog_courseset} prcs ON prcs.programid = pr.id
LEFT JOIN {prog_courseset_course} prcsc ON prcsc.coursesetid = prcs.id
LEFT JOIN {course} c ON prcsc.courseid = c.id
LEFT JOIN {scorm} s ON s.id = cm.instance
LEFT JOIN {scorm_scoes} ss ON ss.scorm = s.id
LEFT JOIN {scorm_scoes_track} sst_status
ON sst_status.userid = u.id
AND sst_status.scormid = ss.scorm
AND sst_status.element = 'cmi_core.lesson_status'
AND sst_status.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_starttime
ON sst_starttime.userid = u.id
AND sst_starttime.scormid = ss.scorm
AND sst_starttime.element = 'x.start.time'
AND sst_starttime.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_duration
ON sst_duration.userid = u.id
AND sst_duration.scormid = ss.scorm
AND sst_duration.element = 'cmi.core.total_time'
AND sst_duration.scoid = ss.id
LEFT JOIN {scorm_scoes_track} sst_score
ON sst_score.userid = u.id
AND sst_score.scormid = ss.scorm
AND sst_score.element = 'cmi.core.score.raw'
AND sst_score.scoid = ss.id
WHERE u.id IN([SOME_EXAMPLE_IDS_HERE]) AND u.deleted = 0
group by u.idnumber, c.id
This is getting the first non null values of these fields. Note that this is relying on the fields not containing commas (but easy enough to specify a different separator for the GROUP_CONCAT). You can specify an order for the GROUP_CONCAT to get your preferred value if required.

Answer 2

Here is a very simplified version of your query:
select u.idnumber AS user_idnumber, sst_status.value AS scorm_status,
sst_starttime.value AS scorm_starttime, sst_duration.value AS scorm_duration,
sst_score.value AS scorm_score, u.id as uid, s.id as scormid, ss.id as ssid
. . .
group by u.idnumber, c.id;
You will notice that most of the columns in the select are not in the group by. This is using a (mis)feature of MySQL and would not generally be allowed in other databases.
What MySQL does is it chooses an abritrary value from one of the rows for the result. The idea is that this works nicely if the column is constant for the group. But, it doesn't usually work as expected in other circumstances.
If you want a non-NULL value, then you can use max() or min() to get the biggest or smallest value.
For instance:
min(sst_status.value) AS scorm_status
and so on.