8:33 p.m.
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
8:41 p.m.
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(a)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(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
8:51 p.m.
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(a)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(a)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(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
9:55 p.m.
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(a)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(a)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(a)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(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>
>
6:20 p.m.
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(a)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(a)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(a)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(a)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(a)lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>
>>
>
11:22 p.m.
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(a)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(a)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(a)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(a)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(a)lists.jboss.org
>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>>
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
11:49 p.m.
Hi,
Looking at the SessionFactoryImpl class, the only way to provide an
SqlFunction is either via the Dialect or via the SessionFactoryOptions:
this.sqlFunctionRegistry = new SQLFunctionRegistry(
jdbcServices.getJdbcEnvironment().getDialect(),
options.getCustomSqlFunctionMap() );
I'm not sure if the ManagedBeanRegistry would have helped in this case
without requiring more code changes.
Vlad
On Thu, May 17, 2018 at 5:22 PM, Christian Beikov <
christian.beikov(a)gmail.com> wrote:
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(a)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(a)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(a)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(a)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(a)lists.jboss.org
>>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>>>>>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
1:25 a.m.
The functions could be contributed via a CDI/Spring bean which might not
be such a bad idea I think. In a test environment there could be a
different contributor active than in production. Of course, this could
also be handled by passing in different configuration property values,
but that's why I was saying I'm not sure about it. Maybe someone else
has an idea if using ManagedBeanRegistry might make sense?
Mit freundlichen Grüßen,
------------------------------------------------------------------------
*Christian Beikov*
Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
Hi,
Looking at the SessionFactoryImpl class, the only way to provide an
SqlFunction is either via the Dialect or via the SessionFactoryOptions:
this.sqlFunctionRegistry =new SQLFunctionRegistry(
jdbcServices.getJdbcEnvironment().getDialect(), options.getCustomSqlFunctionMap() );
I'm not sure if the ManagedBeanRegistry would have helped in this case
without requiring more code changes.
Vlad
On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
<christian.beikov(a)gmail.com <mailto:christian.beikov@gmail.com>> wrote:
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
<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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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-
<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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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
<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(a)lists.jboss.org
<mailto:hibernate-dev@lists.jboss.org>
>>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
<https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>>>>>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
<https://lists.jboss.org/mailman/listinfo/hibernate-dev>
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/hibernate-dev
<https://lists.jboss.org/mailman/listinfo/hibernate-dev>
2:07 a.m.
I personally agree with Christian, I don't see the use of the
ManagedBeanRegistry being a problem. The new interface certainly opens
the door for a variety of builder settings to be contributed easily and
using the registry allows for a versatile way to provide that bean,
whether it be through some CDI/Spring environment or the fallback
solution where we instantiate it ourselves. I believe the code as you
have it can easily be adapted to use the registry by replacing the
"newInstance()" call with a call into getting the bean from the
registry. The registry itself internally should handle getting the bean
from the container or returning you a new instance we instantiate ourselves.
On 05/17/2018 12:25 PM, Christian Beikov wrote:
The functions could be contributed via a CDI/Spring bean which might
not
be such a bad idea I think. In a test environment there could be a
different contributor active than in production. Of course, this could
also be handled by passing in different configuration property values,
but that's why I was saying I'm not sure about it. Maybe someone else
has an idea if using ManagedBeanRegistry might make sense?
Mit freundlichen Grüßen,
------------------------------------------------------------------------
*Christian Beikov*
Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
> Hi,
>
> Looking at the SessionFactoryImpl class, the only way to provide an
> SqlFunction is either via the Dialect or via the SessionFactoryOptions:
> this.sqlFunctionRegistry =new SQLFunctionRegistry(
jdbcServices.getJdbcEnvironment().getDialect(), options.getCustomSqlFunctionMap() );
> I'm not sure if the ManagedBeanRegistry would have helped in this case
> without requiring more code changes.
>
> Vlad
>
> On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
> <christian.beikov(a)gmail.com <mailto:christian.beikov@gmail.com>> wrote:
>
> 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
> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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-
> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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
> <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(a)lists.jboss.org
> <mailto:hibernate-dev@lists.jboss.org>
> >>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> >>>>>
> > _______________________________________________
> > hibernate-dev mailing list
> > hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.jboss.org>
> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.jboss.org>
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>
>
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
2:50 a.m.
Sure thing. I'll try to adapt it to using the Bean registry.
Vlad
On Thu, 17 May 2018, 20:07 Chris Cranford, <crancran(a)gmail.com> wrote:
I personally agree with Christian, I don't see the use of the
ManagedBeanRegistry being a problem. The new interface certainly opens
the door for a variety of builder settings to be contributed easily and
using the registry allows for a versatile way to provide that bean,
whether it be through some CDI/Spring environment or the fallback
solution where we instantiate it ourselves. I believe the code as you
have it can easily be adapted to use the registry by replacing the
"newInstance()" call with a call into getting the bean from the
registry. The registry itself internally should handle getting the bean
from the container or returning you a new instance we instantiate
ourselves.
On 05/17/2018 12:25 PM, Christian Beikov wrote:
> The functions could be contributed via a CDI/Spring bean which might not
> be such a bad idea I think. In a test environment there could be a
> different contributor active than in production. Of course, this could
> also be handled by passing in different configuration property values,
> but that's why I was saying I'm not sure about it. Maybe someone else
> has an idea if using ManagedBeanRegistry might make sense?
>
>
> Mit freundlichen Grüßen,
> ------------------------------------------------------------------------
> *Christian Beikov*
> Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
>> Hi,
>>
>> Looking at the SessionFactoryImpl class, the only way to provide an
>> SqlFunction is either via the Dialect or via the SessionFactoryOptions:
>> this.sqlFunctionRegistry =new SQLFunctionRegistry(
jdbcServices.getJdbcEnvironment().getDialect(),
options.getCustomSqlFunctionMap() );
>> I'm not sure if the ManagedBeanRegistry would have helped in this case
>> without requiring more code changes.
>>
>> Vlad
>>
>> On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
>> <christian.beikov(a)gmail.com <mailto:christian.beikov@gmail.com>>
wrote:
>>
>> 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
>> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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-
>> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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
>> <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(a)lists.jboss.org
>> <mailto:hibernate-dev@lists.jboss.org>
>> >>>>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>> >>>>>
>> > _______________________________________________
>> > hibernate-dev mailing list
>> > hibernate-dev(a)lists.jboss.org <mailto:
hibernate-dev(a)lists.jboss.org>
>> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>>
>> _______________________________________________
>> hibernate-dev mailing list
>> hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.jboss.org
>
>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>>
>>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
3:07 a.m.
Personally I think both a contributor and CDI integration
(ManagedBeanRegitry) make sense here. Btw, the name for the contributor
should not be SqlFunctionContributor - it should reflect the ultimate goal
here which is SqmFunctionTemplate; I can see SqmFunctionContributor or just
FunctionContributor. This is an example of what I meant about waiting for
6...
On Thu, May 17, 2018 at 12:50 PM Vlad Mihalcea <mihalcea.vlad(a)gmail.com>
wrote:
Sure thing. I'll try to adapt it to using the Bean registry.
Vlad
On Thu, 17 May 2018, 20:07 Chris Cranford, <crancran(a)gmail.com> wrote:
> I personally agree with Christian, I don't see the use of the
> ManagedBeanRegistry being a problem. The new interface certainly opens
> the door for a variety of builder settings to be contributed easily and
> using the registry allows for a versatile way to provide that bean,
> whether it be through some CDI/Spring environment or the fallback
> solution where we instantiate it ourselves. I believe the code as you
> have it can easily be adapted to use the registry by replacing the
> "newInstance()" call with a call into getting the bean from the
> registry. The registry itself internally should handle getting the bean
> from the container or returning you a new instance we instantiate
> ourselves.
>
> On 05/17/2018 12:25 PM, Christian Beikov wrote:
> > The functions could be contributed via a CDI/Spring bean which might
not
> > be such a bad idea I think. In a test environment there could be a
> > different contributor active than in production. Of course, this could
> > also be handled by passing in different configuration property values,
> > but that's why I was saying I'm not sure about it. Maybe someone else
> > has an idea if using ManagedBeanRegistry might make sense?
> >
> >
> > Mit freundlichen Grüßen,
> >
------------------------------------------------------------------------
> > *Christian Beikov*
> > Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
> >> Hi,
> >>
> >> Looking at the SessionFactoryImpl class, the only way to provide an
> >> SqlFunction is either via the Dialect or via the
SessionFactoryOptions:
> >> this.sqlFunctionRegistry =new SQLFunctionRegistry(
> jdbcServices.getJdbcEnvironment().getDialect(),
> options.getCustomSqlFunctionMap() );
> >> I'm not sure if the ManagedBeanRegistry would have helped in this case
> >> without requiring more code changes.
> >>
> >> Vlad
> >>
> >> On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
> >> <christian.beikov(a)gmail.com
<mailto:christian.beikov@gmail.com>>
wrote:
> >>
> >> 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
> >> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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-
> >> <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(a)hibernate.org <mailto:steve@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(a)gmail.com <mailto:mihalcea.vlad@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
> >> <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(a)lists.jboss.org
> >> <mailto:hibernate-dev@lists.jboss.org>
> >> >>>>>
https://lists.jboss.org/mailman/listinfo/hibernate-dev
> >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> >> >>>>>
> >> > _______________________________________________
> >> > hibernate-dev mailing list
> >> > hibernate-dev(a)lists.jboss.org <mailto:
> hibernate-dev(a)lists.jboss.org>
> >> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
> >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> >>
> >> _______________________________________________
> >> hibernate-dev mailing list
> >> hibernate-dev(a)lists.jboss.org <mailto:
hibernate-dev(a)lists.jboss.org
> >
> >> https://lists.jboss.org/mailman/listinfo/hibernate-dev
> >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> >>
> >>
> > _______________________________________________
> > hibernate-dev mailing list
> > hibernate-dev(a)lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
3:34 a.m.
I named it MetadataBuilderContributor because you can do more than just
registering SqlFunctions. While implementing it, I realized that, this way,
it's going to be very flexible to customize the Hibernate-native Metadata
while doing the JPA bootstrap.
Related to ManagedBeanRegitry, I'm trying to convert to using it, but
stumbled on this issue.
If I try to get a Bean like this:
managedBeanRegistry.getBean( MetadataBuilderContributor.class );
And there is no BeanContainer set via "hibernate.resource.beans.container",
then the FallbackContainedBean is used, which tries to instantiate the
provided interface:
beanType.newInstance();
But since I provide an interface, the call will fail.
The only workaround I found is this:
ManagedBeanRegistry managedBeanRegistry = standardServiceRegistry
.getService( ManagedBeanRegistry.class );
BeanContainer beanContainer = managedBeanRegistry.getBeanContainer();
ManagedBean<MetadataBuilderContributor>
metadataBuilderContributorManagedBean = null;
if ( beanContainer != null ) {
metadataBuilderContributorManagedBean = managedBeanRegistry
.getBean( MetadataBuilderContributor.class );
}
MetadataBuilderContributor metadataBuilderContributor =
metadataBuilderContributorManagedBean
.getBeanInstance();
Am I using it the wrong way, or do we need to check the BeanContainer first
to make sure it's not null before getting a Bean from the
ManagedBeanRegistry?
Vlad
On Thu, May 17, 2018 at 9:07 PM, Steve Ebersole <steve(a)hibernate.org> wrote:
Personally I think both a contributor and CDI integration
(ManagedBeanRegitry) make sense here. Btw, the name for the contributor
should not be SqlFunctionContributor - it should reflect the ultimate goal
here which is SqmFunctionTemplate; I can see SqmFunctionContributor or just
FunctionContributor. This is an example of what I meant about waiting for
6...
On Thu, May 17, 2018 at 12:50 PM Vlad Mihalcea <mihalcea.vlad(a)gmail.com>
wrote:
> Sure thing. I'll try to adapt it to using the Bean registry.
>
> Vlad
>
> On Thu, 17 May 2018, 20:07 Chris Cranford, <crancran(a)gmail.com> wrote:
>
> > I personally agree with Christian, I don't see the use of the
> > ManagedBeanRegistry being a problem. The new interface certainly opens
> > the door for a variety of builder settings to be contributed easily and
> > using the registry allows for a versatile way to provide that bean,
> > whether it be through some CDI/Spring environment or the fallback
> > solution where we instantiate it ourselves. I believe the code as you
> > have it can easily be adapted to use the registry by replacing the
> > "newInstance()" call with a call into getting the bean from the
> > registry. The registry itself internally should handle getting the bean
> > from the container or returning you a new instance we instantiate
> > ourselves.
> >
> > On 05/17/2018 12:25 PM, Christian Beikov wrote:
> > > The functions could be contributed via a CDI/Spring bean which might
> not
> > > be such a bad idea I think. In a test environment there could be a
> > > different contributor active than in production. Of course, this could
> > > also be handled by passing in different configuration property values,
> > > but that's why I was saying I'm not sure about it. Maybe someone
else
> > > has an idea if using ManagedBeanRegistry might make sense?
> > >
> > >
> > > Mit freundlichen Grüßen,
> > > ------------------------------------------------------------
> ------------
> > > *Christian Beikov*
> > > Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
> > >> Hi,
> > >>
> > >> Looking at the SessionFactoryImpl class, the only way to provide an
> > >> SqlFunction is either via the Dialect or via the
> SessionFactoryOptions:
> > >> this.sqlFunctionRegistry =new SQLFunctionRegistry(
> > jdbcServices.getJdbcEnvironment().getDialect(),
> > options.getCustomSqlFunctionMap() );
> > >> I'm not sure if the ManagedBeanRegistry would have helped in this
> case
> > >> without requiring more code changes.
> > >>
> > >> Vlad
> > >>
> > >> On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
> > >> <christian.beikov(a)gmail.com
<mailto:christian.beikov@gmail.com>>
> wrote:
> > >>
> > >> 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
> > >> <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(a)hibernate.org <mailto:steve@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(a)gmail.com
<mailto:mihalcea.vlad@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-
> > >> <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(a)hibernate.org <mailto:steve@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(a)gmail.com
<mailto:mihalcea.vlad@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
> > >> <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(a)lists.jboss.org
> > >> <mailto:hibernate-dev@lists.jboss.org>
> > >> >>>>>
https://lists.jboss.org/mailman/listinfo/hibernate-dev
> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> > >> >>>>>
> > >> > _______________________________________________
> > >> > hibernate-dev mailing list
> > >> > hibernate-dev(a)lists.jboss.org <mailto:
> > hibernate-dev(a)lists.jboss.org>
> > >> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> > >>
> > >> _______________________________________________
> > >> hibernate-dev mailing list
> > >> hibernate-dev(a)lists.jboss.org <mailto:hibernate-dev@lists.
> jboss.org
> > >
> > >> https://lists.jboss.org/mailman/listinfo/hibernate-dev
> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
> > >>
> > >>
> > > _______________________________________________
> > > hibernate-dev mailing list
> > > hibernate-dev(a)lists.jboss.org
> > > https://lists.jboss.org/mailman/listinfo/hibernate-dev
> >
> > _______________________________________________
> > hibernate-dev mailing list
> > hibernate-dev(a)lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
9:05 p.m.
There could conceivably be multiple MetadataBuilderContributor instances
needing to be integrated. CDI (and therefore ManagedBeanRegistry) do not
work well in this situation. MetadataBuilderContributor should be done via
ServiceLoader (via ClassLoaderService)
When I mentioned ManagedBeanRegistry, I was talking specifically about just
contributing functions - but MetadataBuilderContributor is more
generalized, which is fine.. it just won't fit into the CDI paradigm
On Thu, May 17, 2018 at 1:35 PM Vlad Mihalcea <mihalcea.vlad(a)gmail.com>
wrote:
I named it MetadataBuilderContributor because you can do more than
just
registering SqlFunctions. While implementing it, I realized that, this way,
it's going to be very flexible to customize the Hibernate-native Metadata
while doing the JPA bootstrap.
Related to ManagedBeanRegitry, I'm trying to convert to using it, but
stumbled on this issue.
If I try to get a Bean like this:
managedBeanRegistry.getBean( MetadataBuilderContributor.class );
And there is no BeanContainer set via
"hibernate.resource.beans.container", then the FallbackContainedBean is
used, which tries to instantiate the provided interface:
beanType.newInstance();
But since I provide an interface, the call will fail.
The only workaround I found is this:
ManagedBeanRegistry managedBeanRegistry = standardServiceRegistry
.getService( ManagedBeanRegistry.class );
BeanContainer beanContainer = managedBeanRegistry.getBeanContainer();
ManagedBean<MetadataBuilderContributor>
metadataBuilderContributorManagedBean = null;
if ( beanContainer != null ) {
metadataBuilderContributorManagedBean = managedBeanRegistry
.getBean( MetadataBuilderContributor.class );
}
MetadataBuilderContributor metadataBuilderContributor =
metadataBuilderContributorManagedBean
.getBeanInstance();
Am I using it the wrong way, or do we need to check the BeanContainer
first to make sure it's not null before getting a Bean from the
ManagedBeanRegistry?
Vlad
On Thu, May 17, 2018 at 9:07 PM, Steve Ebersole <steve(a)hibernate.org>
wrote:
> Personally I think both a contributor and CDI integration
> (ManagedBeanRegitry) make sense here. Btw, the name for the contributor
> should not be SqlFunctionContributor - it should reflect the ultimate goal
> here which is SqmFunctionTemplate; I can see SqmFunctionContributor or just
> FunctionContributor. This is an example of what I meant about waiting for
> 6...
>
> On Thu, May 17, 2018 at 12:50 PM Vlad Mihalcea <mihalcea.vlad(a)gmail.com>
> wrote:
>
>> Sure thing. I'll try to adapt it to using the Bean registry.
>>
>> Vlad
>>
>> On Thu, 17 May 2018, 20:07 Chris Cranford, <crancran(a)gmail.com> wrote:
>>
>> > I personally agree with Christian, I don't see the use of the
>> > ManagedBeanRegistry being a problem. The new interface certainly opens
>> > the door for a variety of builder settings to be contributed easily and
>> > using the registry allows for a versatile way to provide that bean,
>> > whether it be through some CDI/Spring environment or the fallback
>> > solution where we instantiate it ourselves. I believe the code as you
>> > have it can easily be adapted to use the registry by replacing the
>> > "newInstance()" call with a call into getting the bean from the
>> > registry. The registry itself internally should handle getting the
>> bean
>> > from the container or returning you a new instance we instantiate
>> > ourselves.
>> >
>> > On 05/17/2018 12:25 PM, Christian Beikov wrote:
>> > > The functions could be contributed via a CDI/Spring bean which might
>> not
>> > > be such a bad idea I think. In a test environment there could be a
>> > > different contributor active than in production. Of course, this
>> could
>> > > also be handled by passing in different configuration property
>> values,
>> > > but that's why I was saying I'm not sure about it. Maybe
someone else
>> > > has an idea if using ManagedBeanRegistry might make sense?
>> > >
>> > >
>> > > Mit freundlichen Grüßen,
>> > >
>> ------------------------------------------------------------------------
>> > > *Christian Beikov*
>> > > Am 17.05.2018 um 16:49 schrieb Vlad Mihalcea:
>> > >> Hi,
>> > >>
>> > >> Looking at the SessionFactoryImpl class, the only way to provide
an
>> > >> SqlFunction is either via the Dialect or via the
>> SessionFactoryOptions:
>> > >> this.sqlFunctionRegistry =new SQLFunctionRegistry(
>> > jdbcServices.getJdbcEnvironment().getDialect(),
>> > options.getCustomSqlFunctionMap() );
>> > >> I'm not sure if the ManagedBeanRegistry would have helped in
this
>> case
>> > >> without requiring more code changes.
>> > >>
>> > >> Vlad
>> > >>
>> > >> On Thu, May 17, 2018 at 5:22 PM, Christian Beikov
>> > >> <christian.beikov(a)gmail.com
<mailto:christian.beikov@gmail.com>>
>> wrote:
>> > >>
>> > >> 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
>> > >> <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(a)hibernate.org <mailto:steve@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(a)gmail.com
<mailto:mihalcea.vlad@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-
>> > >>
<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(a)hibernate.org <mailto:steve@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(a)gmail.com
<mailto:mihalcea.vlad@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
>> > >> <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(a)lists.jboss.org
>> > >> <mailto:hibernate-dev@lists.jboss.org>
>> > >> >>>>>
https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>> > >> >>>>>
>> > >> > _______________________________________________
>> > >> > hibernate-dev mailing list
>> > >> > hibernate-dev(a)lists.jboss.org <mailto:
>> > hibernate-dev(a)lists.jboss.org>
>> > >> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>> > >>
>> > >> _______________________________________________
>> > >> hibernate-dev mailing list
>> > >> hibernate-dev(a)lists.jboss.org <mailto:
>> hibernate-dev(a)lists.jboss.org
>> > >
>> > >> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> > >> <https://lists.jboss.org/mailman/listinfo/hibernate-dev>
>> > >>
>> > >>
>> > > _______________________________________________
>> > > hibernate-dev mailing list
>> > > hibernate-dev(a)lists.jboss.org
>> > > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> >
>> > _______________________________________________
>> > hibernate-dev mailing list
>> > hibernate-dev(a)lists.jboss.org
>> > https://lists.jboss.org/mailman/listinfo/hibernate-dev
>> _______________________________________________
>> hibernate-dev mailing list
>> hibernate-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
>
2181
days inactive
2183
days old
12 comments
4 participants
participants (4)
-
Chris Cranford -
Christian Beikov -
Steve Ebersole -
Vlad Mihalcea