You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@bval.apache.org by Carlos Vara <ba...@gmail.com> on 2010/06/17 18:35:35 UTC
Performance testing apache bval
Hi all,
I have committed to the sandbox a simple project to evaluate the performance
of a jsr303 implementation. It's in a very initial status and only
benchmarks the parsing and validation of beans with basic properties and
nested beans. I intend to add groups/group sequences, nested constraints and
possibly inheritance to make it more exhaustive.
Any opinion / review of the code is very welcomed. Also, I have little
experience on the best practices integrating a custom code generator in
maven, so any help with that part would be very appreciated.
Regards,
Carlos
Re: Performance testing apache bval
Posted by Carlos Vara <ba...@gmail.com>.
The results:
http://carinae.net/2010/06/benchmarking-hibernate-validator-and-apache-beanvalidation-the-two-jsr-303-implementations/
I must say I didn't expect validation speed to differ that much, it has been
a pleasant surprise.
<http://carinae.net/2010/06/benchmarking-hibernate-validator-and-apache-beanvalidation-the-two-jsr-303-implementations/>
Regards,
Carlos
On Fri, Jun 18, 2010 at 11:19 PM, Carlos Vara <ba...@gmail.com> wrote:
> Hello again,
>
> current project in the sandbox now can generate a JSR-303 annotated bean
> graph with groups, inheritance and complex constraints, so it can simulate
> realistic scenarios. Only big missing aspect would be arrays/iterables
> support, but IMHO it won't make a difference in test results.
>
> Current tests check the performance of a jsr-303 impl on these operations:
> - Parsing beans.
> - Validating already parsed beans.
> - Combined parsing+validating speed.
> - Multi-threaded validation of parsed beans.
>
> To run the benchmark, simply execute:
> mvn test
>
> The bean generator procedure is configurable, there is a list of all
> the modifiable properties in generator.default.properties. You can modify
> them in the commandline as:
> mvn -Dgenerator.numBeans=5 -Dgenerator.inheritanceratio=0.2
> -Dtester.numthreads=8
>
> Tests should be reproducible, as the bean generation algorithm will return
> the same beans if executed with the same parameters. The results may vary
> between different versions of the generator though, so the
>
> If there is no opposition, during the weekend I will benchmark our 0.1
> release against hibernate validator 4.1.0.CR1 and post the results in my
> blog.
>
> Regards,
> Carlos
>
>
> On Thu, Jun 17, 2010 at 5:35 PM, Carlos Vara <ba...@gmail.com> wrote:
>
>> Hi all,
>>
>> I have committed to the sandbox a simple project to evaluate the
>> performance of a jsr303 implementation. It's in a very initial status and
>> only benchmarks the parsing and validation of beans with basic properties
>> and nested beans. I intend to add groups/group sequences, nested constraints
>> and possibly inheritance to make it more exhaustive.
>>
>> Any opinion / review of the code is very welcomed. Also, I have little
>> experience on the best practices integrating a custom code generator in
>> maven, so any help with that part would be very appreciated.
>>
>> Regards,
>> Carlos
>>
>
>
Re: Performance testing apache bval
Posted by Mark Struberg <st...@yahoo.de>.
Hi Carlos!
Good idea! I think that Hibernate people will also be very interested in this tool.
LieGrue,
strub
--- On Fri, 6/18/10, Carlos Vara <ba...@gmail.com> wrote:
> From: Carlos Vara <ba...@gmail.com>
> Subject: Re: Performance testing apache bval
> To: bval-dev@incubator.apache.org
> Date: Friday, June 18, 2010, 10:22 PM
> One of the sentences got split:
>
> "The results may vary between different versions of the
> generator though, so
> to replicate them one will need to match both the config
> settings and the
> SVN/release version"
>
> Regards,
> Carlos
>
> On Fri, Jun 18, 2010 at 11:19 PM, Carlos Vara <ba...@gmail.com>
> wrote:
>
> > Hello again,
> >
> > current project in the sandbox now can generate a
> JSR-303 annotated bean
> > graph with groups, inheritance and complex
> constraints, so it can simulate
> > realistic scenarios. Only big missing aspect would be
> arrays/iterables
> > support, but IMHO it won't make a difference in test
> results.
> >
> > Current tests check the performance of a jsr-303 impl
> on these operations:
> > - Parsing beans.
> > - Validating already parsed beans.
> > - Combined parsing+validating speed.
> > - Multi-threaded validation of parsed beans.
> >
> > To run the benchmark, simply execute:
> > mvn test
> >
> > The bean generator procedure is configurable, there is
> a list of all
> > the modifiable properties in
> generator.default.properties. You can modify
> > them in the commandline as:
> > mvn -Dgenerator.numBeans=5
> -Dgenerator.inheritanceratio=0.2
> > -Dtester.numthreads=8
> >
> > Tests should be reproducible, as the bean generation
> algorithm will return
> > the same beans if executed with the same parameters.
> The results may vary
> > between different versions of the generator though, so
> the
> >
> > If there is no opposition, during the weekend I will
> benchmark our 0.1
> > release against hibernate validator 4.1.0.CR1 and post
> the results in my
> > blog.
> >
> > Regards,
> > Carlos
> >
> >
> > On Thu, Jun 17, 2010 at 5:35 PM, Carlos Vara <ba...@gmail.com>
> wrote:
> >
> >> Hi all,
> >>
> >> I have committed to the sandbox a simple project
> to evaluate the
> >> performance of a jsr303 implementation. It's in a
> very initial status and
> >> only benchmarks the parsing and validation of
> beans with basic properties
> >> and nested beans. I intend to add groups/group
> sequences, nested constraints
> >> and possibly inheritance to make it more
> exhaustive.
> >>
> >> Any opinion / review of the code is very welcomed.
> Also, I have little
> >> experience on the best practices integrating a
> custom code generator in
> >> maven, so any help with that part would be very
> appreciated.
> >>
> >> Regards,
> >> Carlos
> >>
> >
> >
>
Re: Performance testing apache bval
Posted by Carlos Vara <ba...@gmail.com>.
One of the sentences got split:
"The results may vary between different versions of the generator though, so
to replicate them one will need to match both the config settings and the
SVN/release version"
Regards,
Carlos
On Fri, Jun 18, 2010 at 11:19 PM, Carlos Vara <ba...@gmail.com> wrote:
> Hello again,
>
> current project in the sandbox now can generate a JSR-303 annotated bean
> graph with groups, inheritance and complex constraints, so it can simulate
> realistic scenarios. Only big missing aspect would be arrays/iterables
> support, but IMHO it won't make a difference in test results.
>
> Current tests check the performance of a jsr-303 impl on these operations:
> - Parsing beans.
> - Validating already parsed beans.
> - Combined parsing+validating speed.
> - Multi-threaded validation of parsed beans.
>
> To run the benchmark, simply execute:
> mvn test
>
> The bean generator procedure is configurable, there is a list of all
> the modifiable properties in generator.default.properties. You can modify
> them in the commandline as:
> mvn -Dgenerator.numBeans=5 -Dgenerator.inheritanceratio=0.2
> -Dtester.numthreads=8
>
> Tests should be reproducible, as the bean generation algorithm will return
> the same beans if executed with the same parameters. The results may vary
> between different versions of the generator though, so the
>
> If there is no opposition, during the weekend I will benchmark our 0.1
> release against hibernate validator 4.1.0.CR1 and post the results in my
> blog.
>
> Regards,
> Carlos
>
>
> On Thu, Jun 17, 2010 at 5:35 PM, Carlos Vara <ba...@gmail.com> wrote:
>
>> Hi all,
>>
>> I have committed to the sandbox a simple project to evaluate the
>> performance of a jsr303 implementation. It's in a very initial status and
>> only benchmarks the parsing and validation of beans with basic properties
>> and nested beans. I intend to add groups/group sequences, nested constraints
>> and possibly inheritance to make it more exhaustive.
>>
>> Any opinion / review of the code is very welcomed. Also, I have little
>> experience on the best practices integrating a custom code generator in
>> maven, so any help with that part would be very appreciated.
>>
>> Regards,
>> Carlos
>>
>
>
Re: Performance testing apache bval
Posted by Carlos Vara <ba...@gmail.com>.
Hello again,
current project in the sandbox now can generate a JSR-303 annotated bean
graph with groups, inheritance and complex constraints, so it can simulate
realistic scenarios. Only big missing aspect would be arrays/iterables
support, but IMHO it won't make a difference in test results.
Current tests check the performance of a jsr-303 impl on these operations:
- Parsing beans.
- Validating already parsed beans.
- Combined parsing+validating speed.
- Multi-threaded validation of parsed beans.
To run the benchmark, simply execute:
mvn test
The bean generator procedure is configurable, there is a list of all
the modifiable properties in generator.default.properties. You can modify
them in the commandline as:
mvn -Dgenerator.numBeans=5 -Dgenerator.inheritanceratio=0.2
-Dtester.numthreads=8
Tests should be reproducible, as the bean generation algorithm will return
the same beans if executed with the same parameters. The results may vary
between different versions of the generator though, so the
If there is no opposition, during the weekend I will benchmark our 0.1
release against hibernate validator 4.1.0.CR1 and post the results in my
blog.
Regards,
Carlos
On Thu, Jun 17, 2010 at 5:35 PM, Carlos Vara <ba...@gmail.com> wrote:
> Hi all,
>
> I have committed to the sandbox a simple project to evaluate the
> performance of a jsr303 implementation. It's in a very initial status and
> only benchmarks the parsing and validation of beans with basic properties
> and nested beans. I intend to add groups/group sequences, nested constraints
> and possibly inheritance to make it more exhaustive.
>
> Any opinion / review of the code is very welcomed. Also, I have little
> experience on the best practices integrating a custom code generator in
> maven, so any help with that part would be very appreciated.
>
> Regards,
> Carlos
>