You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ignite.apache.org by Ilya Kasnacheev <il...@gmail.com> on 2020/02/20 12:52:55 UTC
Let's make BinaryObjectImpl and CacheKeyObject Comparable
Hello!
Since we have merged https://issues.apache.org/jira/browse/IGNITE-6804 we
have to face an embarrassing fact that BinaryObject is not Comparable,
i.e., when you do cache.withKeepBinary().putAll(), there are no obvious
ways to not get a deadlock (or at least warning) here.
One can use LinkedHashMap, but they will have to sort BinaryObject's on
their side, which is not trivial.
So my proposal is to make BinaryObjectImpl and KeyCacheObject (?)
Comparable, by their binary representation. We can't add this constraint to
BinaryObject since it is a public interface (can we), but we can do that
for the implementation types. What do you think?
Regards,
--
Ilya Kasnacheev
Re: Let's make BinaryObjectImpl and CacheKeyObject Comparable
Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!
I think they also should do that, but as phase II.
Regards,
--
Ilya Kasnacheev
ср, 26 февр. 2020 г. в 11:26, Ivan Pavlukhin <vo...@gmail.com>:
> And what about non-java platforms?
>
> Best regards,
> Ivan Pavlukhin
>
> пт, 21 февр. 2020 г. в 14:22, Ilya Kasnacheev <il...@gmail.com>:
> >
> > Hello!
> >
> > I don't see how we can force users to implement Comparable for
> BinaryObject
> > keys, since BinaryObject implementation is not provided by users but by
> us.
> >
> > I think that we, however, could force it for composite non-BinaryObject
> > keys.
> >
> > Regards,
> > --
> > Ilya Kasnacheev
> >
> >
> > пт, 21 февр. 2020 г. в 00:21, Denis Magda <dm...@apache.org>:
> >
> > > Hi Ilya,
> > >
> > > We can oblige users to implement Comparable if they use BinaryObject
> keys.
> > > Ignite can print out a warning if BinaryObject keys passed to putAll
> > > methods don't do that.
> > >
> > > I also wonder how a similar task was solved for Ignite INSERTs. Our
> engine
> > > should use BinaryObjects for compound primary keys and insert them at
> > > patches. That implementation can suggest us some hints.
> > >
> > > -
> > > Denis
> > >
> > >
> > > On Thu, Feb 20, 2020 at 6:53 AM Ilya Kasnacheev <
> ilya.kasnacheev@gmail.com
> > > >
> > > wrote:
> > >
> > > > Hello!
> > > >
> > > > Since we have merged
> https://issues.apache.org/jira/browse/IGNITE-6804
> > > we
> > > > have to face an embarrassing fact that BinaryObject is not
> Comparable,
> > > > i.e., when you do cache.withKeepBinary().putAll(), there are no
> obvious
> > > > ways to not get a deadlock (or at least warning) here.
> > > >
> > > > One can use LinkedHashMap, but they will have to sort BinaryObject's
> on
> > > > their side, which is not trivial.
> > > >
> > > > So my proposal is to make BinaryObjectImpl and KeyCacheObject (?)
> > > > Comparable, by their binary representation. We can't add this
> constraint
> > > to
> > > > BinaryObject since it is a public interface (can we), but we can do
> that
> > > > for the implementation types. What do you think?
> > > >
> > > > Regards,
> > > > --
> > > > Ilya Kasnacheev
> > > >
> > >
>
Re: Let's make BinaryObjectImpl and CacheKeyObject Comparable
Posted by Ivan Pavlukhin <vo...@gmail.com>.
And what about non-java platforms?
Best regards,
Ivan Pavlukhin
пт, 21 февр. 2020 г. в 14:22, Ilya Kasnacheev <il...@gmail.com>:
>
> Hello!
>
> I don't see how we can force users to implement Comparable for BinaryObject
> keys, since BinaryObject implementation is not provided by users but by us.
>
> I think that we, however, could force it for composite non-BinaryObject
> keys.
>
> Regards,
> --
> Ilya Kasnacheev
>
>
> пт, 21 февр. 2020 г. в 00:21, Denis Magda <dm...@apache.org>:
>
> > Hi Ilya,
> >
> > We can oblige users to implement Comparable if they use BinaryObject keys.
> > Ignite can print out a warning if BinaryObject keys passed to putAll
> > methods don't do that.
> >
> > I also wonder how a similar task was solved for Ignite INSERTs. Our engine
> > should use BinaryObjects for compound primary keys and insert them at
> > patches. That implementation can suggest us some hints.
> >
> > -
> > Denis
> >
> >
> > On Thu, Feb 20, 2020 at 6:53 AM Ilya Kasnacheev <ilya.kasnacheev@gmail.com
> > >
> > wrote:
> >
> > > Hello!
> > >
> > > Since we have merged https://issues.apache.org/jira/browse/IGNITE-6804
> > we
> > > have to face an embarrassing fact that BinaryObject is not Comparable,
> > > i.e., when you do cache.withKeepBinary().putAll(), there are no obvious
> > > ways to not get a deadlock (or at least warning) here.
> > >
> > > One can use LinkedHashMap, but they will have to sort BinaryObject's on
> > > their side, which is not trivial.
> > >
> > > So my proposal is to make BinaryObjectImpl and KeyCacheObject (?)
> > > Comparable, by their binary representation. We can't add this constraint
> > to
> > > BinaryObject since it is a public interface (can we), but we can do that
> > > for the implementation types. What do you think?
> > >
> > > Regards,
> > > --
> > > Ilya Kasnacheev
> > >
> >
Re: Let's make BinaryObjectImpl and CacheKeyObject Comparable
Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!
I don't see how we can force users to implement Comparable for BinaryObject
keys, since BinaryObject implementation is not provided by users but by us.
I think that we, however, could force it for composite non-BinaryObject
keys.
Regards,
--
Ilya Kasnacheev
пт, 21 февр. 2020 г. в 00:21, Denis Magda <dm...@apache.org>:
> Hi Ilya,
>
> We can oblige users to implement Comparable if they use BinaryObject keys.
> Ignite can print out a warning if BinaryObject keys passed to putAll
> methods don't do that.
>
> I also wonder how a similar task was solved for Ignite INSERTs. Our engine
> should use BinaryObjects for compound primary keys and insert them at
> patches. That implementation can suggest us some hints.
>
> -
> Denis
>
>
> On Thu, Feb 20, 2020 at 6:53 AM Ilya Kasnacheev <ilya.kasnacheev@gmail.com
> >
> wrote:
>
> > Hello!
> >
> > Since we have merged https://issues.apache.org/jira/browse/IGNITE-6804
> we
> > have to face an embarrassing fact that BinaryObject is not Comparable,
> > i.e., when you do cache.withKeepBinary().putAll(), there are no obvious
> > ways to not get a deadlock (or at least warning) here.
> >
> > One can use LinkedHashMap, but they will have to sort BinaryObject's on
> > their side, which is not trivial.
> >
> > So my proposal is to make BinaryObjectImpl and KeyCacheObject (?)
> > Comparable, by their binary representation. We can't add this constraint
> to
> > BinaryObject since it is a public interface (can we), but we can do that
> > for the implementation types. What do you think?
> >
> > Regards,
> > --
> > Ilya Kasnacheev
> >
>
Re: Let's make BinaryObjectImpl and CacheKeyObject Comparable
Posted by Denis Magda <dm...@apache.org>.
Hi Ilya,
We can oblige users to implement Comparable if they use BinaryObject keys.
Ignite can print out a warning if BinaryObject keys passed to putAll
methods don't do that.
I also wonder how a similar task was solved for Ignite INSERTs. Our engine
should use BinaryObjects for compound primary keys and insert them at
patches. That implementation can suggest us some hints.
-
Denis
On Thu, Feb 20, 2020 at 6:53 AM Ilya Kasnacheev <il...@gmail.com>
wrote:
> Hello!
>
> Since we have merged https://issues.apache.org/jira/browse/IGNITE-6804 we
> have to face an embarrassing fact that BinaryObject is not Comparable,
> i.e., when you do cache.withKeepBinary().putAll(), there are no obvious
> ways to not get a deadlock (or at least warning) here.
>
> One can use LinkedHashMap, but they will have to sort BinaryObject's on
> their side, which is not trivial.
>
> So my proposal is to make BinaryObjectImpl and KeyCacheObject (?)
> Comparable, by their binary representation. We can't add this constraint to
> BinaryObject since it is a public interface (can we), but we can do that
> for the implementation types. What do you think?
>
> Regards,
> --
> Ilya Kasnacheev
>