You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by "Julia Wang (QIUHE)" <Qi...@microsoft.com> on 2015/09/16 02:33:26 UTC
makeRootServiceConfiguration
In Java, when making makeRootServiceConfiguration, we merge serviceConfiguration with all the configurations in evaluatorConfigurationProviders. May I assume this merge is for Java Evaluator only?
Thanks,
Julia
.
RE: makeRootServiceConfiguration
Posted by "Julia Wang (QIUHE)" <Qi...@microsoft.com>.
Great! That is what I observed as well. In fact, we are not supposed to add any more service configuration at Java side for .Net evaluator. If there is any, it should be added at .Net side before crossing the bridge.
I will go to the route to skip deserialization for context and service configuration.
-----Original Message-----
From: Markus Weimer [mailto:markus@weimo.de]
Sent: Tuesday, September 15, 2015 6:15 PM
To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
Subject: Re: makeRootServiceConfiguration
On Tue, Sep 15, 2015 at 5:50 PM, Julia Wang (QIUHE) < Qiuhe.Wang@microsoft.com> wrote:
> In AllocatedEvaluatorImpl class:
>
> private Optional<Configuration> makeRootServiceConfiguration(final
> Optional<Configuration> serviceConfiguration)
>
Yes. In this method we have the following code:
if (EvaluatorType.CLR == evaluatorType) {
LOG.log(Level.FINE, "Not using the ConfigurationProviders as we are configuring a {0} Evaluator.", evaluatorType);
return serviceConfiguration;
}
So, basically nothing is done for CLR Evaluators. At the callsite, we just pass an Optional<ServiceConfiguration>. I assume that is always not set.
Hence, we would be perfectly fine to put the check for the Evaluator type much earlier in the call stack. That would allow us to also pass the Service configuration as a string, which in turn would bypass all class hierarchy (de-)serialization issues.
Markus
-----Original Message-----
> From: Markus Weimer [mailto:markus@weimo.de]
> Sent: Tuesday, September 15, 2015 5:43 PM
> To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
> Subject: Re: makeRootServiceConfiguration
>
> Which code file / lines are you referring to?
>
> Markus
>
> On Tue, Sep 15, 2015 at 5:33 PM, Julia Wang (QIUHE) <
> Qiuhe.Wang@microsoft.com> wrote:
>
> > In Java, when making makeRootServiceConfiguration, we merge
> > serviceConfiguration with all the configurations in
> > evaluatorConfigurationProviders. May I assume this merge is for Java
> > Evaluator only?
> >
> >
> >
> > Thanks,
> >
> > Julia
> >
> > .
> >
>
Re: makeRootServiceConfiguration
Posted by Markus Weimer <ma...@weimo.de>.
On Tue, Sep 15, 2015 at 5:50 PM, Julia Wang (QIUHE) <
Qiuhe.Wang@microsoft.com> wrote:
> In AllocatedEvaluatorImpl class:
>
> private Optional<Configuration> makeRootServiceConfiguration(final
> Optional<Configuration> serviceConfiguration)
>
Yes. In this method we have the following code:
if (EvaluatorType.CLR == evaluatorType) {
LOG.log(Level.FINE, "Not using the ConfigurationProviders as we are
configuring a {0} Evaluator.", evaluatorType);
return serviceConfiguration;
}
So, basically nothing is done for CLR Evaluators. At the callsite, we just
pass an Optional<ServiceConfiguration>. I assume that is always not set.
Hence, we would be perfectly fine to put the check for the Evaluator type
much earlier in the call stack. That would allow us to also pass the
Service configuration as a string, which in turn would bypass all class
hierarchy (de-)serialization issues.
Markus
-----Original Message-----
> From: Markus Weimer [mailto:markus@weimo.de]
> Sent: Tuesday, September 15, 2015 5:43 PM
> To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
> Subject: Re: makeRootServiceConfiguration
>
> Which code file / lines are you referring to?
>
> Markus
>
> On Tue, Sep 15, 2015 at 5:33 PM, Julia Wang (QIUHE) <
> Qiuhe.Wang@microsoft.com> wrote:
>
> > In Java, when making makeRootServiceConfiguration, we merge
> > serviceConfiguration with all the configurations in
> > evaluatorConfigurationProviders. May I assume this merge is for Java
> > Evaluator only?
> >
> >
> >
> > Thanks,
> >
> > Julia
> >
> > .
> >
>
RE: makeRootServiceConfiguration
Posted by "Julia Wang (QIUHE)" <Qi...@microsoft.com>.
In AllocatedEvaluatorImpl class:
private Optional<Configuration> makeRootServiceConfiguration(final Optional<Configuration> serviceConfiguration)
-----Original Message-----
From: Markus Weimer [mailto:markus@weimo.de]
Sent: Tuesday, September 15, 2015 5:43 PM
To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
Subject: Re: makeRootServiceConfiguration
Which code file / lines are you referring to?
Markus
On Tue, Sep 15, 2015 at 5:33 PM, Julia Wang (QIUHE) < Qiuhe.Wang@microsoft.com> wrote:
> In Java, when making makeRootServiceConfiguration, we merge
> serviceConfiguration with all the configurations in
> evaluatorConfigurationProviders. May I assume this merge is for Java
> Evaluator only?
>
>
>
> Thanks,
>
> Julia
>
> .
>
Re: makeRootServiceConfiguration
Posted by Markus Weimer <ma...@weimo.de>.
Which code file / lines are you referring to?
Markus
On Tue, Sep 15, 2015 at 5:33 PM, Julia Wang (QIUHE) <
Qiuhe.Wang@microsoft.com> wrote:
> In Java, when making makeRootServiceConfiguration, we merge
> serviceConfiguration with all the configurations in
> evaluatorConfigurationProviders. May I assume this merge is for Java
> Evaluator only?
>
>
>
> Thanks,
>
> Julia
>
> .
>