Problems with using STUFF - sql

Question

Why the heck isn't this working??? Seems to follow everything I've found around here. I'm getting the error:
Column '#TempTable.clientId' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.
If I add the tt.clientId to the group by, then it doesn't do the stuff, and combine them all into 1 line, they come up as separate rows. Did I make a typo or something?
SELECT tt.Station, STUFF((SELECT ', ' + c.client_code
FROM client c
WHERE tt.clientId = c.ID
FOR XML PATH('')),1,1,'') [Values]
FROM #TempTable tt
GROUP BY tt.Station

Answer 1

SELECT ... FOR XML PATH should be a function of GROUP BY column[s]
tt.station in your case.
Something like that
SELECT tt.Station, STUFF((SELECT ', ' + c.client_code
FROM client c
JOIN #TempTable tt2
ON tt2.clientId = c.ID
AND tt2.Station = tt.Station
FOR XML PATH('')),1,1,'') [Values]
FROM
GROUP BY tt.Station

Answer 2

You have to add tt.ClientId to the GROUP BY because you are using it to correlate the subquery here: WHERE tt.clientId = c.ID
Otherwise, how does SQL Server know which ClientId to use for each Station?
If you don't want to group by ClientId, then you have to correlate by tt.Station.

Answer 3

There are a couple alternatives here, but here are some things to consider:
Determine what you want your left-most item to be -- in this case, it looks like it's supposed to be station. If that is the case, one way to approach it is to first get the distinct set of stations. This can be done using a group by or a distinct.
Determine the level of the items for which you wish to generate a list -- in this case it's client_code. Therefore, you want to get you inner select to be at that level. One way to do that is to resolve the distinct set of client codes prior to attempting to use for xml.
One critique -- it's always good to provide a simple set of data. Makes providing an answer much easier and faster.
Again, there are alternatives here, but here's a possible solution.
The test data
select top (100)
client_id = abs(checksum(newid())) % 100,
client_code = char(abs(checksum(newid())) % 10 + 65)
into #client
from sys.all_columns a
cross join sys.all_columns b;
select top (100)
station = abs(checksum(newid())) % 10,
client_id = abs(checksum(newid())) % 50 -- just a subset
into #temp
from sys.all_columns a
cross join sys.all_columns b;
The query
select station, client_codes = stuff((
select ', ' + cc.client_code
from (
select distinct c.client_code -- distinct client codes
from #temp t
join #client c
on t.client_id = c.client_id
where t.station = s.station) cc
order by client_code
for xml path('')), 1, 2, '')
from (
select distinct station
from #temp) s; -- distinct stations
The results
station client_codes
----------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
0 B, C, D, E, G, J
1 A, B, D, G, H, J
2 A, C, E, F, G, H, J
3 B, C, H, J
4 A, B, C, D, F, H, J
5 H, J
6 D, E, F, G, I
7 A, C, D, F, G, H, J
8 A, E, G
9 C, E, F, G, I
Hope this helps.