change to semantics of query with no 'select' list

[gavinking]

I just learned from the migration guide that the semantics of a queries like:

from Person p join p.address

has changed.

I did not know this, and I don't know what the reasoning was.

Perhaps there's a great reason, but at first sight this seems like a pretty bad idea to me:

1. A missing select clause has always been interpreted as approximately something like a SQL select *, i.e. return me everything in the from clause. This is something different. It returns me, what, only root entities in the from clause? Only the first root entity? I'm not sure.
2. It has the potential to break a huge number of existing HQL queries that have been working for decades, and the compiler won't help find them. Of all the things that are breaking in 6, this seems like pretty much the worst breakage of anything in terms of having a potentially wide impact on user code. Given that, the justification has got to be something a bit stronger than: "oh, well, we like this interpretation a bit better now".

So, what was behind this?

[sebersole]

The legacy behavior makes no sense.

• HQL != SQL, and
• no select != explicit select (*)

[gavinking]

The legacy behavior makes no sense.

OK, that may be, but it makes sense to me, and so I would like you to explain why it doesn't make sense.

I mean, this is the sort of change we can only make if it's clearly, objectively more correct.

I'm not seeing that. So help me see it.

[gavinking]

I mean, maybe just give me an example of a query where it doesn't make sense...

[gavinking]

So in trying to figure out exactly what the semantics of this are, I tried the following two very similar-looking HQL queries:

from X x join Y y on x.id=y.id

And

from X x, Y y where x.id=y.id

Clearly these are the "same" query, written using a slightly different syntax.

Now, in Hibernate 5, these two queries would have returned the same-shaped result list: a list of tuples of type [X,Y].

But in Hibernate 6, the first query returns a List of Xs, and the second query returns a list of tuples of shape [X,Y].

I'm really struggling to see how the new behavior is better. In particular I don't see how it's the case that the legacy interpretation of the first query makes "no sense" whereas the legacy interpretation of the second query does make sense.

[sebersole]

Clearly these are the "same" query, written using a slightly different syntax.

This is where we will disagree. They are not the same

[sebersole]

Look, we simply disagree on this. I'll let others chime in. But I have no plans on reverting this

[gavinking]

This issue was fixed before H6 went out.

The resolution (suggested by @sebersole) was to use the result type to imply the select list.