Bug in auto-increment for MSSQL

Question

I am using kodo 3.1.0 and have a scenario where I am persisting a new object
that uses an auto-increment PK and the database has an on-insert trigger
that inserts a log into an audit table.
At the database level, what is happing is:
1) kodo calls "insert into t_timeentry (........)
2) the database calls "insert into t_timeentrylog (......)
3) kodo calls @@IDENTITY to get the ID to associate with the new timeentry.
The problem is, @@IDENTITY is getting back the last identity used,
regardless of table, so it is returning the id inserted into t_timeentrylog
and kodo thinks that's the id of the timeentry. In the end, I get a
optimistic lock exception when I try to update the new timeentry because it
is trying to update based on the id of the timeentrylog.
What you need to do is call SCOPE_IDENTITY or IDENT_CURRENT rather than
@@IDENTITY when you get auto-incremented columns. Unfortunately, those
functions only exist in SQL 2000 (maybe 7), so people with SQL6.5 (and maybe
7) are just out luck if they have triggers on inserts.
Is there a work-around for this? It's causing problems throughout my
system.
Nathan

Answer 1

If the query string you want executed to get the last auto-inc value
does not rely on context, you can set it easily with:
kodo.jdbc.DBDictionary: LastGeneratedKeyQuery="<sql>"
If it is more complex, then you'll need to extend
kodo.jdbc.sql.SQLServerDictionary and override the following method:
public long getGeneratedKey (kodo.jdbc.schema.Column col,
Connection conn)
throws SQLException
// if you need the col or table name:
String colName = col.getName ();
String tableName = col.getTable ().getName ();
// execute the necessary SQL to return the last inserted value...
Then you can plug your dictionary into Kodo with:
kodo.jdbc.DBDictionary: your.custom.Dictionary