Triplicated records on insert with select

Question

The following code aways creates triplicated records, but if I remove the transaction commands, it works as expected, inserting only 1 record for each number on the in clause:

insert into table1(id2, col) 
select col2, 1 
from table2
where col2 in (1, 2, 3) and col2 not in (select id2 from table1 where col = 1)
group by col2;

Is this a MySQL bug? The same statement don't cause duplicated records on MSSQL Server.
Is it possible to run this queries or similar ones with transaction on without duplicating records?

Example data before queries:

table1
id | id2 | col
==============
 -- empty --

table2
id | col2
==============
1  | 1
2  | 1
3  | 1
4  | 2
5  | 2
6  | 2
7  | 3
8  | 3
9  | 3

Example data after queries:

table1 (* duplicated records I don't want)
id | id2 | col
==============
1  | 1   | 1
2  | 1   | 1  *
3  | 1   | 1  *
4  | 2   | 1
5  | 2   | 1  *
6  | 2   | 1  *
7  | 3   | 1
8  | 3   | 1  *
9  | 3   | 1  *

table2 keeps the same

Obs.:

• table2 has duplicated values.
• Running the select query alone returns only the desired results (no duplicates).
• It seems that insert into() select ignores distinct and group by clauses.

Edit:

I found that this problem occurs because MySQL ignores the GROUP BY clause on an INSERT INTO ... SELECT query. Is there a way to group this records on insert?

Answer 1

I found that this problem occurs because MySQL ignores the GROUP BY clause on an INSERT INTO ... SELECT query.

Don't know exactly why, but if I use a query that doesn't use GROUP BY or DISTICT it works.