ave a couple of nit comments. The method name is still
>>>>>>>>> segmentSize
>>>>>>>>>>>> in
>>>>>>>>>>>>> the code block vs segmentInterval and the order of the
>> parameters
>>>>&g
t;>>>>>>>> third persistentWindowStore don't match the order in the
> JavaDoc.
> >>>>>>>>>>>
> >>>>>>>>>>> Thanks,
> >>>>>>>>>>> Bill
> >>>>>>>>&g
;
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On Thu, Jun 21, 2018 at 3:32 PM John Roesler
>>>>>>>>> wrote:
>>>>>>>>>>>
>>>>>>>>>>
n Roesler > >
>> >>>>>>> wrote:
>> >>>>>>>>>
>> >>>>>>>>>> Interesting... I did not initially consider it because I didn't
>> >>>>>> want
>> >>>>>>>
ays
> >>>>>>>>>> be 3!
> >>>>>>>>>>
> >>>>>>>>>> There's one caveat to this, which I think was a mistake. The
> >> field
> >>>>>>>>>> `segments` in Windows is pub
gt;>>>>> set
>>>>>>>>>> it directly on any Window instance like:
>>>>>>>>>>
>>>>>>>>>> TimeWindows tw = TimeWindows.of(100);
>>>>>>>>>> tw.segment
. Sadly there's no way to
> > > >> "deprecate"
> > > >>>>> this
> > > >>>>>> exposure, so I propose just to make it private.
> > > >>>>>>
> > > >>>>>> With this new
"segmentInterval" throughout the interface.
> > >>>>>>
> > >>>>>> On Wed, Jun 20, 2018 at 5:06 PM Guozhang Wang >
> > >>>>> wrote:
> > >>>>>>
> > >>>>>>> Hello John,
> >
zhang Wang
> >>>>> wrote:
> >>>>>>
> >>>>>>> Hello John,
> >>>>>>>
> >>>>>>> Thanks for the KIP.
> >>>>>>>
> >>>>>>> Should we consider making t
es#persistentWindowStore`
>>>>>>> parameters as well?
>>>>>>>
>>>>>>>
>>>>>>> Guozhang
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Jun 20, 2018 at 1:31 PM, John
you made that comment to me before, I thought you meant
> > as
> > > > >> opposed
> > > > >> > to "segments". Now it makes sense that you meant as opposed to
> > > > >> > "segmentSize".
> > > &
gmentSize".
> > > >> >
> > > >> > I named it that way to match the peer method "windowSize", which
> is
> > > >> also a
> > > >> > quantity of milliseconds.
> > > >> >
> > &g
ant as opposed to
> > >> > "segmentSize".
> > >> >
> > >> > I named it that way to match the peer method "windowSize", which is
> > >> also a
> > >> > quantity of milliseconds.
> > &g
gt; >> >
> >> > I agree that "interval" is more intuitive, but I think I favor
> >> consistency
> >> > in this case. Does that seem reasonable?
> >> >
> >> > Thanks,
> >> > -John
> >> >
> >>
>> >
>> > I agree that "interval" is more intuitive, but I think I favor
>> consistency
>> > in this case. Does that seem reasonable?
>> >
>> > Thanks,
>> > -John
>> >
>> > On Wed, Jun 20, 2018 at 1:06 PM Ted Yu
t 1:06 PM Ted Yu wrote:
> > >
> > > > Normally size is not measured in time unit, such as milliseconds.
> > > > How about naming the new method segmentInterval ?
> > > > Thanks
> > > > Original message From: Joh
t measured in time unit, such as milliseconds.
> > > How about naming the new method segmentInterval ?
> > > Thanks
> > > Original message From: John Roesler <
> j...@confluent.io>
> > > Date: 6/20/18 10:45 AM (GMT-08:00) To: dev@kafk
ze is not measured in time unit, such as milliseconds.
> > > How about naming the new method segmentInterval ?
> > > Thanks
> > > Original message From: John Roesler <
> j...@confluent.io>
> > > Date: 6/20/18 10:45 AM (GMT-08:00
ally size is not measured in time unit, such as milliseconds.
> > How about naming the new method segmentInterval ?
> > Thanks
> > Original message --------From: John Roesler
> > Date: 6/20/18 10:45 AM (GMT-08:00) To: dev@kafka.apache.org Subject:
> > [DISCUSS] KIP-3
t; > Thanks
> > Original message --------From: John Roesler
> > Date: 6/20/18 10:45 AM (GMT-08:00) To: dev@kafka.apache.org Subject:
> > [DISCUSS] KIP-319: Replace segments with segmentSize in
> > WindowBytesStoreSupplier
> > Hello All,
> >
segmentInterval ?
> Thanks
> Original message From: John Roesler
> Date: 6/20/18 10:45 AM (GMT-08:00) To: dev@kafka.apache.org Subject:
> [DISCUSS] KIP-319: Replace segments with segmentSize in
> WindowBytesStoreSupplier
> Hello All,
>
> I'd like to prop
Normally size is not measured in time unit, such as milliseconds.
How about naming the new method segmentInterval ?
Thanks
Original message From: John Roesler Date:
6/20/18 10:45 AM (GMT-08:00) To: dev@kafka.apache.org Subject: [DISCUSS]
KIP-319: Replace segments
Hello All,
I'd like to propose KIP-319 to fix an issue I identified in KAFKA-7080.
Specifically, we're creating CachingWindowStore with the *number of
segments* instead of the *segment size*.
Here's the jira: https://issues.apache.org/jira/browse/KAFKA-7080
Here's the KIP:
23 matches
Mail list logo
