some thoughts inline:
On Sun, 9 Feb 2020 at 08:34, Max Rydahl Andersen <manderse(a)redhat.com> wrote:
While working on Quarkus a few of us (Georgios in cc and I in
particular) been pondering on doing a guide on best approaches on
how to access data a bit more raw (i.e. raw sql, stateless season, etc.)
as we got quite a lot of feedback stating Hibernate/JPA was considered
complex in comparison to things like Spring Data and JDBC.
When you go look its as if Spring Data and JDBC copied a lot of the
patterns/access patterns we had in Hibernate before neither of those two
Unfortunately, there are though a few gotchas I think would be great if
we could somehow improve.
Sanne suggested me to post to this list to start the conversation -
there might come some more once I get going on writing the full guide;
but wanted to get it started sooner rather than later.
### Document simplest programatic configuration
(This might just be "max is stupid" issue - please tell me :)
Trying to do a simple java main method booting Hibernate I could not
figure out with the "new" metadata API a simple way to just start
What is the equivalent way to do this in new metadata api:
Just setup dialect, give connection info and get a
I found apis to start from a file containing those settings; but I
really just want a programmatic api
to use in batch script etc. without a need for external files.
Agreed we should explore making the "frameworkless bootstrap" simpler;
booting the ORM today requires some expertise, resulting in people to
rely on other frameworks to do the job.
However, the possible configuration options we have today are
significant; there's good reasons to have them so while an helper
would be welcome, I don't expect we'll want to change or simplify the
existing boot code, as it's working and doing a great job. This would
rather provide a simple, limited alternative.
It will be necessary to clarify that the "simple" alternative is also
very limiting: for example I could see myself volunteering to create
such a little toy, but don't expect it to work in modular
environments, containers, pick up transaction managers, integrate with
DI frameworks, allow for Datasource injection, etc. etc.. Ok?
A consequence of such limitations would be that runtime performance
and efficiency would also be limited; let's be upfront about the
intent in the API documentation.
I can try toy with a POC but can't promise a quick turnaround on this
one; not least I'd need Steve's blessing and review:
Please comment on the JIRA or send a PR if you have a more concrete
idea or a draft.
### Non-documented Deprecation of `setResultTransformer()` and
The [approach of using
we've had "forever" and its exactly what spring data and jdbc are using
in examples where they say data access are easier.
We want to focus and use this kind of API in our Quarkus guides to
educate this is available.
Unfortunately in Hibernate 5.2+ that api got marked [as
with no other documentation than a `@todo develop a new approach to
The actual only proper reason I could find was mentioned in
where its pointed
out that for now we couldn't do a related functional style api for this
until Hibernate 6.
The @deprecation makes all code using this in many ides get a strike
through of that code, making people think it is bad to use when it is
very much not the case.
Any chance that deprecation can either be removed or at least documented
to be less scary/more informative ?
(I'll happily make a PR if can be pointed to the new better api?)
I'll leave this question to others: TBH I don't know the plan here.
### doWork missing on stateless session
In similar vein `.connection()` on stateless session is marked as
deprecated with [no alternative nor actual
on Session the alternative is `doWork`. Any reason why that api isn't
available on stateless session ?
Here again, the ask is so we can educate and point people to use
statelessSession rather than dropping Hibernate fully and from that
api if needed go do `.doWork()` rather than refer to deprecated
Makes sense, I believe we should be able to do this quickly:
Andrea volunteered to have a look; assigned to him assuming it's
straight forward; if it gets tricky we'll have to wait for Steve's