Hi Graham,
I understand Torque generates, if a column has autoIncrement="true" for target
database Postgresql the sequence, but not the required entry in the generated
SQL in the column.
1. The method getAutoIncrement in
org.apache.torque.templates.platform.PlatformPostgresqlImpl returns an empty
String instead of nextval('" + name + "') (Postgrsql8)
2. in PostgresAdapter the method getIDMethodSQL is not called, as for the
table idMethod="native", you might try to change to idMethod="sequence".
Then it returns ("select nextval('" + name + "')"); which is not right, what we
want, but read on 😊
3. This is, where the "hack" is documented, and which explains why
getAutoIncrement returns nothing:
In org.apache.torque.util.BasePeerImpl, line 597ff:
doInsert(Column[], Criteria, String, Connection)
if (keyGen instanceof SequenceIdGenerator)
{
SequenceIdGenerator sequenceIdGenerator
= (SequenceIdGenerator) keyGen;
String idSql = sequenceIdGenerator.getIdSql(
getIdMethodInfo());
// This is a bit of a hack.
// The idSql is usually a stand-alone statement, but we
// need the part to be inserted as a column.
// Therefore we extract the complete word containing
// the term nextval
Fast checking svn this is already in the code since 2013 (TORQUE-305 fix )..
Having said that, it might be an idea to use the much simpler keyword serial
which is supported since Postgresql9 or even Postgrsql10 syntax ?
Best regards, Georg
-----Ursprüngliche Nachricht-----
Von: Graham Leggett via torque-dev <[email protected]>
Gesendet: Dienstag, 16. Juni 2026 00:17
An: Apache Torque Developers List <[email protected]>
Cc: Graham Leggett <[email protected]>
Betreff: Re: Torque7 and postgresql autoincrementing primary keys not working
On 15 Jun 2026, at 22:54, Graham Leggett <[email protected]> wrote:
> After upgrading from torque3.3 to torque7 (big jump), I have everything
> compiling, but what has suddenly vanished are autoincrementing database
> primary keys for postgres.
Stepping through more code, I get here:
public ObjectKey<?> doInsert(
final ColumnValues insertValues,
final Connection connection)
throws TorqueException
{
if (insertValues == null)
{
throw new TorqueException("insertValues is null");
}
if (connection == null)
{
throw new TorqueException("connection is null");
}
String databaseNameFromInsertValues = insertValues.getDbName();
if (databaseNameFromInsertValues == null)
{
databaseNameFromInsertValues = getDatabaseName();
}
Database database = Torque.getDatabase(databaseNameFromInsertValues);
Object keyInfo = getIdMethodInfo();
IdGenerator keyGen = database.getIdGenerator(
getTableMap().getPrimaryKeyMethod());
SimpleKey<?> id = null;
// can currently generate only single column pks, therefore a single
// columnMap is ok
ColumnMap primaryKey = null;
if (keyGen != null)
{
// fail on multiple pks
primaryKey = getTableMap().getPrimaryKey();
// primaryKey will be null if there is no primary key
// defined for the table we're inserting into.
if (keyGen.isPriorToInsert() && primaryKey != null
&& !insertValues.containsKey(primaryKey))
{
id = getId(primaryKey, keyGen, connection, keyInfo);
insertValues.put(
primaryKey,
new JdbcTypedValue(id.getValue(), id.getJdbcType()));
}
}
The above code appears to have picked up the
org.apache.torque.adapter.HsqldbAdapter based on the JDBC driver in the unit
tests, instead of postgres as defined in the maven torque plugin:
<torque.targetDatabase>postgresql</torque.targetDatabase>
hsqldb was previously in postgres-mode using sql.syntax_pgs=true; - does that
no longer work?
Regards,
Graham
--
