On Thu, Nov 26, 2015 at 12:22 PM, Vladimir Ozerov <[email protected]>
wrote:

> Yes, because I cannot imagine a use case when user will need type ID
> (except of ID mapper). This appears to be implementation detail.
>

I actually would prefer a reverse approach. Instead of removing typeId, I
would keep it and also add the fieldId methods as well to the BinaryType
API. I think it makes sense to provide all the info to users, instead of
hiding it.


>
> On Thu, Nov 26, 2015 at 12:17 PM, Dmitriy Setrakyan <[email protected]
> >
> wrote:
>
> > On Thu, Nov 26, 2015 at 12:15 PM, Vladimir Ozerov <[email protected]>
> > wrote:
> >
> > > Sure,
> > >
> > >    1. IgniteBinary.typeId(String typeName) -> no replacement, we simply
> > do
> > >    not need it.
> > >    2. IgniteBinary.metadata(int typeId) -> use
> > IgniteBinary.metadata(String
> > >    typeName) instead
> > >    3. BinaryType.typeId() -> use BinaryType.typeName() instead.
> > >
> >
> > Are you proposing to completely remove typeID from the API?
> >
> >
> > >
> > >
> > > On Thu, Nov 26, 2015 at 12:12 PM, Dmitriy Setrakyan <
> > [email protected]
> > > >
> > > wrote:
> > >
> > > > Vladimir,
> > > >
> > > > Can you describe the new APIs to get this information in Ignite?
> > > >
> > > > Thanks,
> > > > D.
> > > >
> > > > On Thu, Nov 26, 2015 at 12:04 PM, Vladimir Ozerov <
> > [email protected]>
> > > > wrote:
> > > >
> > > > > Folks,
> > > > >
> > > > > We have several public methods which operate on "type ID" concept:
> > > > >
> > > > >    - IgniteBinary.typeId(String typeName)
> > > > >    - IgniteBinary.metadata(int typeId)
> > > > >    - BinaryType.typeId()
> > > > >
> > > > > These methods came from initial GridGain portables implementation
> > where
> > > > it
> > > > > is possbile that there is no type metadata so ID is the only way to
> > > > operate
> > > > > on the type.
> > > > >
> > > > > In Ignite we *always* have type metadata. So I think we can safely
> > > remove
> > > > > mentioned methods.
> > > > >
> > > > > Thoughts?
> > > > >
> > > > > Vladimir.
> > > > >
> > > >
> > >
> >
>

Reply via email to