How to do GROUP BY with INNER JOIN - sql

Question

I have a small problem with GROUP BY and INNER JOIN. I try to get spent hours in incidents group by user, to see that whitch user are more efficient.
In this moment, I only get the error
'Column 'XXX' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause'
I understand that all of the fields in SELECT statement are necessary in GROUP BY, but if I group for all of this fields, I don't receive that I want.
I think that my problem is because I don't understand GROUP BY and INNER JOIN correctly, so I was trying to learn a lot of websites, but at least in this moment I can't see my error.
Here is my code:
SELECT
id_incident_project AS ID,
i.title AS title,
companyname as social_name,
username as tech_name,
ia.DESCRIPTION_TEXT as description,
CONVERT(varchar, ia.ACTIONDATE, 101) as action_date,
CAST(TIME as INT) as acting_time,
CAST(actions_time as INT) AS total_time
FROM
incident i
INNER JOIN incident_0001 i1 ON i1.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN incident_action ia ON ia.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN agent a ON a.ID_AGENT = ia.ID_AGENT
INNER JOIN username u on u.ID_USERNAME = a.ID_USERNAME
WHERE
ia.ACTIONDATE BETWEEN '## START DATE (YYYYMMDD)##' AND '## END DATE (YYYYMMDD)##'
GROUP BY username

Answer 1

All columns in the SELECT clause that you aren't grouping by must go through an aggregate function. To just get something, you can pass them through MAX:
SELECT
MAX(id_incident_project) AS ID,
MAX(i.title) AS title,
MAX(companyname) as social_name,
username as tech_name,
MAX(ia.DESCRIPTION_TEXT) as description,
MAX(CONVERT(varchar, ia.ACTIONDATE, 101)) as action_date,
MAX(CAST(TIME as INT)) as acting_time,
MAX(CAST(actions_time as INT)) AS total_time
FROM
incident i
INNER JOIN incident_0001 i1 ON i1.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN incident_action ia ON ia.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN agent a ON a.ID_AGENT = ia.ID_AGENT
INNER JOIN username u on u.ID_USERNAME = a.ID_USERNAME
WHERE
ia.ACTIONDATE BETWEEN '## START DATE (YYYYMMDD)##' AND '## END DATE (YYYYMMDD)##'
GROUP BY username
However, when there are multiple rows for the same username after all the joins are completed, this approach will risk values from different source rows being outputted. E.g., if username 'JohnSmith' had (say) two incidents, one on 2019-01-01 with a title of 'Zombie Sighting' and the second on 2019-03-31 with a title of 'Aitch Dropping', then the 'maximum' title pulled out will be 'Zombie Sighting' with a 'maximum' date pulled off of 2019-03-31, so for different incidents.
To avoid this, you can replace the GROUP BY with a partitioning on username instead, order the incidents, and pick out one per username in a consistent fashion:
SELECT
ID, title, social_name, tech_name, description,
action_date, acting_time, total_time
FROM (
SELECT
id_incident_project) AS ID,
i.title AS title,
companyname as social_name,
username as tech_name,
ia.DESCRIPTION_TEXT) as description,
CONVERT(varchar, ia.ACTIONDATE, 101) as action_date,
CAST(TIME as INT) as acting_time,
CAST(actions_time as INT) AS total_time,
ROW_NUMBER () OVER (
PARTITION BY username
ORDER BY
ia.ACTIONDATE DESC, -- pick out latest
i1.ID_INCIDENT DESC -- tie breaker
) AS OrderNum
FROM
incident i
INNER JOIN incident_0001 i1 ON i1.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN incident_action ia ON ia.ID_INCIDENT = i.ID_INCIDENT
INNER JOIN agent a ON a.ID_AGENT = ia.ID_AGENT
INNER JOIN username u on u.ID_USERNAME = a.ID_USERNAME
WHERE
ia.ACTIONDATE BETWEEN '## START DATE (YYYYMMDD)##' AND '## END DATE (YYYYMMDD)##'
) t
WHERE t.OrderNum = 1