Hi Michael,
Sure. Say we have 2 input topics t1 & t2 below:
t1{
int key;
string t2_id;
...
}
t2 {
string key;
..
}
If we create global tables out of these we'd get:
GlobalKTable<Integer, ?> t1;
GlobalKTable<String, ?> t2;
So the join can only go in 1 direction, i.e, from t1 -> t2 as in order to
perform the join we need to use a KeyValueMapper to extract the t2 key from
the t1 value.
Does that make sense?
Thanks,
Damian
On Wed, 7 Dec 2016 at 10:44 Michael Noll <mich...@confluent.io> wrote:
> > There is no outer-join for GlobalKTables as the tables may be keyed
> > differently. So you need to use the key from the left side of the join
> > along with the KeyValueMapper to resolve the right side of the join. This
> > wont work the other way around.
>
> Care to elaborate why it won't work the other way around? If, for example,
> we swapped the call from leftTable.join(rightTable) to
> rightTable.join(leftTable), that join would work, too. Perhaps I am
> missing something though. :-)
>
>
>
>
> On Wed, Dec 7, 2016 at 10:39 AM, Damian Guy <damian....@gmail.com> wrote:
>
> > Hi Matthias,
> >
> > Thanks for the feedback.
> >
> > There is no outer-join for GlobalKTables as the tables may be keyed
> > differently. So you need to use the key from the left side of the join
> > along with the KeyValueMapper to resolve the right side of the join. This
> > wont work the other way around.
> >
> > On the bootstrapping concern. If the application is failing before
> > bootstrapping finishes, the problem is likely to be related to a terminal
> > exception, i.e., running out of disk space, corrupt state stores etc. In
> > these cases, we wouldn't want the application to continue. So i think
> this
> > is ok.
> >
> > Thanks,
> > Damian
> >
> > On Tue, 6 Dec 2016 at 21:56 Matthias J. Sax <matth...@confluent.io>
> wrote:
> >
> > > Thanks for the KIP Damian. Very nice motivating example!
> > >
> > > A few comments:
> > >
> > > - why is there no outer-join for GlobalKTables
> > > - on bootstrapping GlobalKTable, could it happen that this never
> > > finishes if the application fails before bootstrapping finishes and new
> > > data gets written at the same time? Do we need to guard against this
> > > (seems to be a very rare corner case, so maybe not required)?
> > >
> > >
> > > -Matthias
> > >
> > >
> > > On 12/6/16 2:09 AM, Damian Guy wrote:
> > > > Hi all,
> > > >
> > > > I would like to start the discussion on KIP-99:
> > > >
> > > https://cwiki.apache.org/confluence/pages/viewpage.
> > action?pageId=67633649
> > > >
> > > > Looking forward to your feedback.
> > > >
> > > > Thanks,
> > > > Damian
> > > >
> > >
> > >
> >
>
