Update Table with Duplicates - sql-server

Question

My Table, #MetaData looks like this
Table_Name Element Join_prefix
Incident hold_reason h
Incident impact i
Incident incident_state i
Incident notify n
Incident severity s
Incident state s
Change impact i
Change incident_state i
I want to update the join_prefix where it is the same, to the first 2 characters of the element, within the Table_Name. So the table looks like
Table_Name Element Join_prefix
Incident hold_reason h
Incident impact im
Incident incident_state in
Incident notify n
Incident severity se
Incident state st
Change impact im
Change incident_state in
I've been using the following sql but it updates all the rows
update #MetaData
set join_prefix=substring(element,1,2)
where exists(
select [Table_Name],[Join_prefix]
from #MetaData
group by [Table_Name],[Join_prefix]
having count(join_prefix)>1)

Answer 1

One method would be to use an updatable CTE. Within the CTE you can use a windowed COUNT to count how many rows have the same prefix, and then update those rows:
SELECT *
INTO dbo.YourTable
FROM (VALUES('Incident','hold_reason ',CONVERT(varchar(4),'h')),
('Incident','impact ',CONVERT(varchar(4),'i')),
('Incident','incident_state',CONVERT(varchar(4),'i')),
('Incident','notify ',CONVERT(varchar(4),'n')),
('Incident','severity ',CONVERT(varchar(4),'s')),
('Incident','state ',CONVERT(varchar(4),'s')),
('Change ','impact ',CONVERT(varchar(4),'i')),
('Change ','incident_state',CONVERT(varchar(4),'i')))V(Table_Name,Element,Join_prefix)
GO
WITH CTE AS(
SELECT Element,
Join_prefix,
COUNT(Join_prefix) OVER (PARTITION BY Join_prefix) AS C
FROM dbo.YourTable)
UPDATE CTE
SET Join_prefix = LEFT(Element,2)
WHERE C > 1;
GO
SELECT *
FROM dbo.YourTable;
GO
DROP TABLE dbo.YourTable;

Answer 2

Your subquery is not correlated to the outside, so it always returns true. You need a WHERE
Note that exists doesn't need to select anything, you can select 1.
Also count(*) and count(non_null_value) is the same
update m1
set join_prefix = substring(element, 1, 2)
from #MetaData m1
where exists (select 1
from #MetaData m2
where m2.join_prefix = m1.join_prefix
group by m2.Table_Name, m2.Join_prefix
having count(*) > 1
);
A better method would be an updatable CTE
with CTE as (
select *,
cnt = count(*) over (partition by m.Table_Name, m.join_prefix)
from #MetaData m
)
update CTE
set join_prefix = substring(element,1,2)
from #MetaData m1
where t.cnt > 1;