Hibernate version: 3

I use assigned ids in my mappings and if I want to use saveOrUpdate I have to add unsaved-value=undefined

The result seems to be that Hibernate generates an SQL select statement to determine if the dataset exists and then to perform either an update or an insert depending on the select statement.

Wouldn't it be better to make only an update (which actually includes a where clause to determine the existence of the dataset) and if that fails to create an insert? That would reduce an extra statement with a probability of 0.5 (depending on your data) with the result, also depending on your data, in a performance gain.

If you want to avoid the extra SELECT, you can define a version or time-stamp property with an unsaved value (in combination with the assigned primary key). Of course this won't work if you are working with a legacy database whose schema you can't change...
HTH
Satish

mm - not sure I phrased that particularly well. Trying again:
If you want to use assigned primary keys and wish to avoid the extra SELECT, you can define a version or time-stamp property with an unsaved value. Of course this won't work if you are working with a legacy database whose schema you can't change...
HTH
Satish

I could do that but it was more a question "why" Hibernate makes an extra select statement which costs extra time instead of trying to make and update which would reduce time in case if a dataset with the given id exists if not then you have lost nothing and it just makes an insert.

Have a look at the Interceptor in section 16.4: http://www.hibernate.org/hib_docs/refer ... child.html