UPDATE query using value as condition - sql-server

Question

I have an update query which takes the form:
update myTable
set aColumn = case when t2.someValue = 3 then 1 else 0 end
from myTable
inner join anotherTable t2 on t2.id = myTable.id
Trouble is, it doesn't work.
The 'case' or t2.someValue seems to be being ignored, and every row in myTable gets the same value (0) in aColumn - including those rows where t2.someValue = 3.
Is this a problem with SQL Server, or have I done something wrong?
EDIT: My updated query, based on a comment below...
declare #myTable table (row_id int, aCol int, bCol int)
insert into #myTable
values
(1,0,0),
(2,0,0),
(3,0,0),
(4,0,0),
(5,0,0)
declare #anotherTable table (row_id int, col_id int, someValue int)
insert into #anotherTable
values
(1,1,9),
(1,2,8),
(1,3,7),
(4,3,6),
(5,5,5)
select * from #myTable
update t
set t.aCol = case when t2.col_id = 1 then t2.someValue else t.aCol end,
t.bCol = case when t2.col_id = 3 then t2.someValue else t.bCol end
from #myTable t
left join #anotherTable t2 on t2.row_id = t.row_id
where t2.col_id IN (1,2,3,5)
select * from #myTable
This doesn't update the bCol of the 1st row with a 7 which it should...
row_id aCol bCol
1 9 0
2 0 0
3 0 0
4 0 6
5 0 0

Answer 1

Try this -
update myTable
set aColumn = (case when t2.someValue = 3 then 1 else 0 end)
from myTable
inner join anotherTable t2 on t2.id = myTable.id

Answer 2

From your edit... you have a 1:Many relationship and this is why it's not happening. You have 3 rows in #anotherTable which can join to tod_id = 1 in #myTable. This is why your update isn't working correctly.
declare #myTable table (row_id int, aCol int, bCol int)
insert into #myTable
values
(1,0,0),
(2,0,0),
(3,0,0),
(4,0,0),
(5,0,0)
declare #anotherTable table (row_id int, col_id int, someValue int)
insert into #anotherTable
values
(1,1,9),
(1,2,8),
(1,3,7),
(4,3,6),
(5,5,5)
select
t.*
,t2.*
from #myTable t
left join #anotherTable t2 on t2.row_id = t.row_id
where t2.col_id IN (1,2,3,5)
+--------+------+------+--------+--------+-----------+
| row_id | aCol | bCol | row_id | col_id | someValue |
+--------+------+------+--------+--------+-----------+
| 1 | 0 | 0 | 1 | 1 | 9 | --notice the implied cross join as
| 1 | 0 | 0 | 1 | 2 | 8 | --row_id 1 is duplicated
| 1 | 0 | 0 | 1 | 3 | 7 | --3 times
| 4 | 0 | 0 | 4 | 3 | 6 |
| 5 | 0 | 0 | 5 | 5 | 5 |
+--------+------+------+--------+--------+-----------+

Answer 3

Ok, Given the advice above - esp. from SCSimon - I have to admit that what I was trying to do was unusual, and too much of an edge-case, for SQL (almost any flavour) to understand.
I have resorted to making the updates to the 12 individual columns as seperate little update statements.
Not as elegant, maybe, but it works without error.
I also doubt that it actually takes any longer to run than a complex query would.