SQL - Dynamically selecting information into a variable from different databases - sql

Question

i have 3 databases, db 1 and db 2 are for jobs and db 3 is for workflows.
both db 1 and 2 have a process to create workflows directly into db 3.
i have a trigger in db 3 that activates when there is a workflow creation because i need to add more information into the workflow from the jobs.
so what i did was create a variable that i set with a normal select from db 1 (select job from db1.dbo.job)
i want to know if there is a way to do the select statement dynamically
for example:
if db1 then #db = db1
if db2 then #db = db2
then in the #variable = select job from #db
right now what i have is the following query but i am simple adding all my logic to another IF and then changing the selects statements... i dont want to do that becuase i will be adding more data bases to integrate with db3
Declare
#flag int,
#variable varchar(10)
set #flag = 1
if #flag = 1
begin
set #variable = (select job from db1.dbo.job)
end
else if #flag = 2
begin
set #variable = (select job from db2.dbo.job)
end
update db3 set job = #variable
is it even possible to do what im trying to do????

Answer 1

You need dynamic SQL for this.
Note:
Always use QUOTENAME to correctly escape names
Use sp_executesql to pass data in parameters all the way through. You can use OUTPUT parameters too
Declare
#flag int
set #flag = 1
DECLARE #sql nvarchar(max) = N'
UPDATE db3
SET job = (
SELECT TOP 1 job
FROM ' + QUOTENAME(CASE WHEN #flag = 1 THEN 'db1' ELSE 'db2' END) + '.dbo.job
);
';
EXEC sp_executesql #sql;

Answer 2

This kind of thing is possible using techniques like dynamic sql and sp_executesql, or by creating synonyms on the fly.
HOWEVER, you have a more fundamental issue with the approach.
When a trigger is called, you cannot make the assumption that only a single row will be in the inserted table, because it is possible to insert multiple rows into a table all at once.
To demonstrate the problem, here is a "solution" to your problem which "works" if only one row is inserted into table T. But think about what would happen if more than one row was inserted. What would the "correct" flag value be if multiple rows are inserted with different flag values? How can we update the row in T based on the value of the #id variable if multiple rows are affected?
You could use this pattern and iterate over the rows in the inserted table using a cursor, but it would be a bit ugly.
create table T(id int, flag int, extraData nvarchar(128));
create or alter trigger T_ins on T for insert as
begin
declare #sql nvarchar(max) = 'update T set extraData = (select top 1 name from {otherDB}.sys.tables) where id = #id'
declare #db sysname, #id int;
-- this select doesn't make any sense if more than one row exists in the inserted table!
select #id = id,
#db = iif(inserted.flag = 1, 'master', 'msdb')
from inserted;
set #sql = replace(#sql, '{otherDB}', #db);
exec sp_executesql #sql, N'#id int', #id;
end
A way to do it without a cursor is to left join onto all of the possible sources of additional data in the other databases, including the mutually exclusive flag value in the join condition, and coalesce the columns into the update statement. I think the code is also much cleaner:
create or alter trigger jobs_ins on jobs for insert as
begin
update j
set extraData = coalesce(j1.somedata, j2.somedata)
from jobs j
join inserted ins on ins.id = j.id
left join otherDB1.dbo.jobdata j1 on i.flag = 1 and j1.id = ins.id
left join otherDB2.dbo.jobdata j2 on i.flag = 2 and j2.id = ins.id;
end