Correlation names using insert and outer join - database

Question

I am trying to run a code to insert rows from one table using rows from a different table on a different database.
I had this:
INSERT [testDB].[dbo].[table1]
SELECT * FROM
[sourceDB].[dbo].[table1]
LEFT OUTER JOIN [testDB].[dbo].[table1]
ON [sourceDB].[dbo].[table1].[PKcolumn] = [testDB].[dbo].[table1].[PKcolumn]
WHERE [testDB].[dbo].[table1].[PKcolumn] IS NULL
However I was told to add correlation names so I made this:
INSERT test
SELECT * FROM
[sourceDB].[dbo].[table1] as source
LEFT OUTER JOIN
[testDB].[dbo].[table1] as test
ON
source.[PKcolumn] = test.[PKcolumn]
WHERE test.[PKcolumn] IS NULL
I ended up getting this as an error message:
Msg 208, Level 16, State 1, Line 1
Invalid object name 'test'.
Does anyone know what I'm doing wrong?

Answer 1

In the first line you should use the real table name as in
insert into testDB.dbo.table1
SQLServer does not accept an alias or correlation name in that spot, and I confirmed that by testing.
But you can use the alias later in the query and it can be quite useful to do so to avoid ambiguity about which table a column comes from.
Another potential problem in this query is the use of select *. This tries to insert the combined column set from sourcedb.dbo.table1 and testdb.dbo.table1 into testdb.dbo.table1. That can't work.
Instead of select * you could say...(assuming source and test have exactly the same columns)
select source.*
or you could call out the specific columns as in...
select source.colA, source.col3, etc....
I don't know the names of your columns.

Answer 2

INSERT test
SELECT *
FROM [sourceDB].[dbo].[table1] as source
LEFT OUTER JOIN [testDB].[dbo].[table1] as test
ON source.[PKcolumn] = test.[PKcolumn]
WHERE test.[PKcolumn] IS NULL
Let's talk about what is wrong with this. First select * would have all the columns from source and test in it which is clearly more columns than the table you plan to insert into has. It never acceptable to use select * in an insert statement for several reasons.
First, if anyone changes the order of the columns or the structure of table, the insert breaks. Second, when you have a join like this, then it has the wrong number of columns. Third, even if they have the same columns if they are orginally in a differnt order, you may put the data into the worng column. If they are similar datatypes and the data fits or can be implicity converted, the database won't stop you from doing this.
Next you can't use an alias from the select as the destination in an insert, you must reference the actual tablename.
Finally, it is a very poor practice to not use a column list in every insert. This helps with maintenance and makes sure you can check to see if the columns inteh selct match up to the columns in the insert. Further, if you have an autogenerated field, you must use a column list or it will try to insert into the autogenerated field and thus error.
So your statement should look something like this:
INSERT [testDB].[dbo].[table1] (field1, field2, field3)
SELECT source.field1, source.field2, source.field3
FROM [sourceDB].[dbo].[table1] as source
LEFT OUTER JOIN [testDB].[dbo].[table1] as test
ON source.[PKcolumn] = test.[PKcolumn]
WHERE test.[PKcolumn] IS NULL
Or (possibly more efficient, you will have to test in your particular situation):
INSERT [testDB].[dbo].[table1] (field1, field2, field3)
SELECT source.field1, source.field2, source.field3
FROM [sourceDB].[dbo].[table1] as source
WHERE NOT EXISTS (SELECT * FROM testDB].[dbo].[table1] test
WHERE source.[PKcolumn]=test.[PKcolumn])