Introduce a common superinterface for `Session` and `StatelessSession`

[yrodiere]

Hibernate ORM has SharedSessionContract, but Munity.Session/Munity.StatelessSession don't have an equivalent.

See quarkusio/quarkus#51063 (comment) : we'd need at least isOpen() in that contract.

I feel like unwrap, close, and all the methods from QueryProducer are also good candidates.

SharedSessionContract could probably provide other ideas as well.

[yrodiere]

I thought there was an issue about this, but I couldn't find it, so... here you go.

Feel free to close if my searching skills are just garbage and the issue already exists.

[gavinking]

We had a discussion about this where I argued that QueryProducer was good enough.

[gavinking]

But perhaps this helps: #3176

[yrodiere]

We had a discussion about this where I argued that QueryProducer was good enough.

For users it might be, but for integrators it's not good enough. In particular, sometimes we need to manipulate Session and StatelessSession the same way, e.g. to check they are open. At the moment we need to go out of our way and duplicate code and/or move the isOpen() check to a method with two implementations, one for Session and one for StatelessSession.

FWIW if you want us to use an equivalent to SharedSessionContractImplementor instead, that'd be perfectly fine for our use case. But we don't even have that right now.

But perhaps this helps: #3176

Not really. We can already do casts, if that's the solution you're suggesting. I don't think anyone is worried that a QueryProducer would be something else than a Session or StatelessSession. EDIT: At least not in the integration code I'm interested in.

[gavinking]

I don't understand. A switch over a sum type is not a "cast" in the sense of being something risky.

[gavinking]

I mean, you can write your own function closeSession(QueryProducer) in three lines.

[DavideD]

I thought there was an issue about this, but I couldn't find it, so... here you go.

This is the old issue: #2104

[yrodiere]

I mean, you can write your own function closeSession(QueryProducer) in three lines.

Sure. That's what we're doing.

And we could list common operations for Session and StatelessSession in a single interface with a very similar amount of lines.

I don't understand why you'd think we're in a better world writing various closeSession(QueryProducer) and isOpen(QueryProducer) methods in integrators instead of just... having a common superinterface that actually suits what integrators need.

[gavinking]

I mean, we can introduce SessionImplementor and friends, but that still involves a cast.

[yrodiere]

I mean, we can introduce SessionImplementor and friends, but that still involves a cast.

Right. The main advantage is that we can cast right when we create the session, store the result in a variable/field, and the rest of the code is free from instanceofs and casts.

[gavinking]

Look, in Hibernate the SessionImplementor interface makes sense, because the object which does the implementing is the object which implements Session. But in reactive that's not the case. It's would be super-weird to have MutinySessionImplementor and StageSessionImplementor objects because those things are just trivial delegators.

But nor do I want to complexity the application view of these objects by introducing things with nausea-inducing names like SharedSessionContract. I would much rather keep the user view clean.

But what I can do, I guess is pull isOpen() up onto Closeable and have QueryProducer be the thing that extends Closeable, if that helps.