You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by Dhruv Mahajan <dh...@gmail.com> on 2015/10/12 20:11:03 UTC
IIMRUClient in c#
Hi
Why does IIMRUClient in c# need to have a generic type.
Why cannot we simply make the class without generics but rather make the
Submit function like
Submit<TMapInput,TMapOutput,Result>()
This has the advantage where the IIMRUClient needs to be generated or
invoked before we know the generic types. For example we can invoke one
IIMRUClient and use it in multiple algorithms.
This will decouple the invocation and actual use of client.
Dhruv
RE: IIMRUClient in c#
Posted by Dhruv Mahajan <dh...@gmail.com>.
Yes I am talking about submit in IMRU client.
Jobsubmissionbuilder does not depend on the generic types. Submit is the
only function.
On Oct 12, 2015 5:20 PM, "Julia Wang (QIUHE)" <Qi...@microsoft.com>
wrote:
> It make sense to remove generic arguments from IIMRUClient, that also
> makes configuration setting easier from outside if it is necessary.
>
> When you say Submit<TMapInput,TMapOutput,Result>(), I guess you are
> talking about the submit in IMRUClient, right?
>
> As we discussed last week, if in future, we would like to have IMRUClient
> to return JobSubmission object and let tlcreef to call IREEFClient to
> submit the job, how would that work in this case?
>
> Thanks,
> Julia
>
> -----Original Message-----
> From: Markus Weimer [mailto:markus@weimo.de]
> Sent: Monday, October 12, 2015 2:34 PM
> To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
> Subject: Re: IIMRUClient in c#
>
> Sounds good!
>
> Please file JIRAs for them. "Discussed changes" sounds like stuff that
> isn't written down anywhere, which we can't have :-)
>
> Markus
>
> On Mon, Oct 12, 2015 at 11:28 PM, Dhruv Mahajan <dh...@gmail.com>
> wrote:
> > Cool..While I am at it I will also do the IMRUJobDefinition and other
> > discussed changes.
> >
> > Dhruv
> >
> > On Mon, Oct 12, 2015 at 1:30 PM, Markus Weimer <ma...@weimo.de> wrote:
> >
> >> Sounds reasonable. +1 on that change. Also, I'd be OK with this being
> >> a breaking change, given how fresh IMRU still is.
> >>
> >> Markus
> >>
> >> On Mon, Oct 12, 2015 at 8:11 PM, Dhruv Mahajan
> >> <dh...@gmail.com>
> >> wrote:
> >> > Hi
> >> >
> >> > Why does IIMRUClient in c# need to have a generic type.
> >> >
> >> > Why cannot we simply make the class without generics but rather
> >> > make the Submit function like
> >> >
> >> > Submit<TMapInput,TMapOutput,Result>()
> >> >
> >> > This has the advantage where the IIMRUClient needs to be generated
> >> > or invoked before we know the generic types. For example we can
> >> > invoke one IIMRUClient and use it in multiple algorithms.
> >> >
> >> > This will decouple the invocation and actual use of client.
> >> >
> >> > Dhruv
> >>
>
RE: IIMRUClient in c#
Posted by "Julia Wang (QIUHE)" <Qi...@microsoft.com>.
It make sense to remove generic arguments from IIMRUClient, that also makes configuration setting easier from outside if it is necessary.
When you say Submit<TMapInput,TMapOutput,Result>(), I guess you are talking about the submit in IMRUClient, right?
As we discussed last week, if in future, we would like to have IMRUClient to return JobSubmission object and let tlcreef to call IREEFClient to submit the job, how would that work in this case?
Thanks,
Julia
-----Original Message-----
From: Markus Weimer [mailto:markus@weimo.de]
Sent: Monday, October 12, 2015 2:34 PM
To: REEF Developers Mailinglist <de...@reef.incubator.apache.org>
Subject: Re: IIMRUClient in c#
Sounds good!
Please file JIRAs for them. "Discussed changes" sounds like stuff that isn't written down anywhere, which we can't have :-)
Markus
On Mon, Oct 12, 2015 at 11:28 PM, Dhruv Mahajan <dh...@gmail.com> wrote:
> Cool..While I am at it I will also do the IMRUJobDefinition and other
> discussed changes.
>
> Dhruv
>
> On Mon, Oct 12, 2015 at 1:30 PM, Markus Weimer <ma...@weimo.de> wrote:
>
>> Sounds reasonable. +1 on that change. Also, I'd be OK with this being
>> a breaking change, given how fresh IMRU still is.
>>
>> Markus
>>
>> On Mon, Oct 12, 2015 at 8:11 PM, Dhruv Mahajan
>> <dh...@gmail.com>
>> wrote:
>> > Hi
>> >
>> > Why does IIMRUClient in c# need to have a generic type.
>> >
>> > Why cannot we simply make the class without generics but rather
>> > make the Submit function like
>> >
>> > Submit<TMapInput,TMapOutput,Result>()
>> >
>> > This has the advantage where the IIMRUClient needs to be generated
>> > or invoked before we know the generic types. For example we can
>> > invoke one IIMRUClient and use it in multiple algorithms.
>> >
>> > This will decouple the invocation and actual use of client.
>> >
>> > Dhruv
>>
Re: IIMRUClient in c#
Posted by Markus Weimer <ma...@weimo.de>.
Sounds good!
Please file JIRAs for them. "Discussed changes" sounds like stuff that
isn't written down anywhere, which we can't have :-)
Markus
On Mon, Oct 12, 2015 at 11:28 PM, Dhruv Mahajan <dh...@gmail.com> wrote:
> Cool..While I am at it I will also do the IMRUJobDefinition and other
> discussed changes.
>
> Dhruv
>
> On Mon, Oct 12, 2015 at 1:30 PM, Markus Weimer <ma...@weimo.de> wrote:
>
>> Sounds reasonable. +1 on that change. Also, I'd be OK with this being
>> a breaking change, given how fresh IMRU still is.
>>
>> Markus
>>
>> On Mon, Oct 12, 2015 at 8:11 PM, Dhruv Mahajan <dh...@gmail.com>
>> wrote:
>> > Hi
>> >
>> > Why does IIMRUClient in c# need to have a generic type.
>> >
>> > Why cannot we simply make the class without generics but rather make the
>> > Submit function like
>> >
>> > Submit<TMapInput,TMapOutput,Result>()
>> >
>> > This has the advantage where the IIMRUClient needs to be generated or
>> > invoked before we know the generic types. For example we can invoke one
>> > IIMRUClient and use it in multiple algorithms.
>> >
>> > This will decouple the invocation and actual use of client.
>> >
>> > Dhruv
>>
Re: IIMRUClient in c#
Posted by Dhruv Mahajan <dh...@gmail.com>.
Cool..While I am at it I will also do the IMRUJobDefinition and other
discussed changes.
Dhruv
On Mon, Oct 12, 2015 at 1:30 PM, Markus Weimer <ma...@weimo.de> wrote:
> Sounds reasonable. +1 on that change. Also, I'd be OK with this being
> a breaking change, given how fresh IMRU still is.
>
> Markus
>
> On Mon, Oct 12, 2015 at 8:11 PM, Dhruv Mahajan <dh...@gmail.com>
> wrote:
> > Hi
> >
> > Why does IIMRUClient in c# need to have a generic type.
> >
> > Why cannot we simply make the class without generics but rather make the
> > Submit function like
> >
> > Submit<TMapInput,TMapOutput,Result>()
> >
> > This has the advantage where the IIMRUClient needs to be generated or
> > invoked before we know the generic types. For example we can invoke one
> > IIMRUClient and use it in multiple algorithms.
> >
> > This will decouple the invocation and actual use of client.
> >
> > Dhruv
>
Re: IIMRUClient in c#
Posted by Markus Weimer <ma...@weimo.de>.
Sounds reasonable. +1 on that change. Also, I'd be OK with this being
a breaking change, given how fresh IMRU still is.
Markus
On Mon, Oct 12, 2015 at 8:11 PM, Dhruv Mahajan <dh...@gmail.com> wrote:
> Hi
>
> Why does IIMRUClient in c# need to have a generic type.
>
> Why cannot we simply make the class without generics but rather make the
> Submit function like
>
> Submit<TMapInput,TMapOutput,Result>()
>
> This has the advantage where the IIMRUClient needs to be generated or
> invoked before we know the generic types. For example we can invoke one
> IIMRUClient and use it in multiple algorithms.
>
> This will decouple the invocation and actual use of client.
>
> Dhruv