[hibernate-dev] Simplify SQL function registration when bootstrapping via JPA

[Christian Beikov]

It looks ok to me although I'm not sure if it wouldn't be more 
appropriate to instantiate the contributor via ManagedBeanRegistry.


Mit freundlichen Grüßen,
------------------------------------------------------------------------
*Christian Beikov*
Am 17.05.2018 um 11:20 schrieb Vlad Mihalcea:
> In the end, I thought of a more generic approach to for JPA bootstrapping
> which, not only allows us to register SqlFunctions, but we can apply other
> changes on the MetadataBuilder object used during bootstrap.
>
> This is the Pull Request:
>
> https://github.com/hibernate/hibernate-orm/pull/2288
>
> Let me know what you think.
>
> Vlad
>
> On Wed, May 16, 2018 at 3:55 PM, Steve Ebersole <steve at hibernate.org> wrote:
>
>> Its not so much hindering 6.0 that I am concerned with.  The problem is
>> updatability for the user.  The more things they use that change = the more
>> work to upgrade.
>>
>> On Wed, May 16, 2018 at 6:51 AM Vlad Mihalcea <mihalcea.vlad at gmail.com>
>> wrote:
>>
>>> I suppose this will be refactored considerably in 6.x.
>>>
>>> However, it's just a small change and I don't think it will hinder any
>>> 6.x changes.
>>>
>>> I'm thinking of defining an SqlFunctionContributor interface
>>> (@FunctionalInterface)
>>> which can be provided via the properties  Map that's supplied when using
>>> the Persistence#createEntityManagerFactory method.
>>>
>>> In EntityManagerFactoryBuilder, I can just inspect that and pass it
>>> further to MetamodelBuilder.
>>>
>>> So, it does not take any effort to implement it and users can benefit
>>> from it. I recently answered a question on the forum on this topic:
>>>
>>> https://discourse.hibernate.org/t/i-cant-use-group-concat-
>>> in-queries/752/14
>>>
>>> And, realized that I was wrong about suggesting doing it via the
>>> Integrator mechanism (since the Metamodel is already frozen by the time we
>>> execute the Integrator).
>>>
>>> Vlad
>>>
>>> On Wed, May 16, 2018 at 2:41 PM, Steve Ebersole <steve at hibernate.org>
>>> wrote:
>>>
>>>> This should maybe wait for 6.0.  We are ditching SQLFunction in favor of
>>>> a more AST-friendly contract.
>>>>
>>>> Beyond that, go for it.
>>>>
>>>>
>>>> On Wed, May 16, 2018, 6:34 AM Vlad Mihalcea <mihalcea.vlad at gmail.com>
>>>> wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I realized that only the native Hibernate bootstrapping mechanism allows
>>>>> for passing custom SQL functions.
>>>>>
>>>>> For JPA, we can extend the Dialect to register, but that's not always
>>>>> desirable as it requires a code change
>>>>> every time the user switches to a new Dialect version.
>>>>>
>>>>> For this reason, I created this Jira issue:
>>>>>
>>>>> https://hibernate.atlassian.net/browse/HHH-12589
>>>>>
>>>>> Let me know if anyone has anything against implementing this feature.
>>>>>
>>>>> Vlad
>>>>> _______________________________________________
>>>>> hibernate-dev mailing list
>>>>> hibernate-dev at lists.jboss.org
>>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>>>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev