A more concrete question: did we consider having the method `partition` take `ProduceRecord` as one of its parameters and `Cluster` as the other?
Ismael On Sat, Aug 26, 2023 at 12:50 PM Greg Harris <greg.har...@aiven.io.invalid> wrote: > Hey Ismael, > > > The mention of "runtime" is specific to Connect. When it comes to > clients, > one typically compiles and runs with the same version or runs with a newer > version than the one used for compilation. This is standard practice in > Java and not something specific to Kafka. > > When I said "older runtimes" I was being lazy, and should have said > "older versions of clients at runtime," thank you for figuring out > what I meant. > > I don't know how common it is to compile a partitioner against one > version of clients, and then distribute and run the partitioner with > older versions of clients and expect graceful degradation of features. > If you say that it is very uncommon and not something that we should > optimize for, then I won't suggest otherwise. > > > With regards to the Admin APIs, they have been extended several times > since introduction (naturally). One of them is: > > > https://github.com/apache/kafka/commit/1d22b0d70686aef5689b775ea2ea7610a37f3e8c > > Thanks for the example. I also see that includes a migration from > regular arguments to the DTO style, consistent with your > recommendation here. > > I think the DTO style and the proposed additional argument style are > both reasonable. > > Thanks, > Greg > > On Sat, Aug 26, 2023 at 9:46 AM Ismael Juma <m...@ismaeljuma.com> wrote: > > > > Hi Greg, > > > > The mention of "runtime" is specific to Connect. When it comes to > clients, > > one typically compiles and runs with the same version or runs with a > newer > > version than the one used for compilation. This is standard practice in > > Java and not something specific to Kafka. > > > > With regards to the Admin APIs, they have been extended several times > since > > introduction (naturally). One of them is: > > > > > https://github.com/apache/kafka/commit/1d22b0d70686aef5689b775ea2ea7610a37f3e8c > > > > Ismael > > > > On Sat, Aug 26, 2023 at 8:29 AM Greg Harris <greg.har...@aiven.io.invalid > > > > wrote: > > > > > Hey Ismael, > > > > > > Thank you for clarifying where the DTO pattern is used already, I did > > > not have the admin methods in mind. > > > > > > > With the DTO approach, you don't create a new DTO, you simply add a > new > > > overloaded constructor and accessor to the DTO. > > > > > > With this variant, partitioner implementations would receive a > > > `NoSuchMethodException` when trying to access newer methods in older > > > runtimes. Do we expect the interface implementers will maintain the > > > try-catch to support backwards-compatibility? > > > Fortunately here the Headers type already exists, but in the future if > > > a new subtype is added at the same time as the change to the DTO is > > > made, interface implementers will need to be careful to avoid > > > NoClassDefFoundErrors. > > > We used this "add a new method" style extension in > > > > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-610%3A+Error+Reporting+in+Sink+Connectors > > > and had to be very specific in recommending how users interact with > > > the new extension point, and there ended up being lots of sharp edges > > > in practice. > > > > > > Do you have any examples of a DTO-based API that has been extended > > > since it was initially implemented? I'm not familiar with the > > > evolution of the Admin APIs. > > > > > > Thanks! > > > Greg > > > > > > On Sat, Aug 26, 2023 at 6:45 AM Ismael Juma <m...@ismaeljuma.com> wrote: > > > > > > > > Hi Greg, > > > > > > > > The point is that the approach proposed here introduces complexity > > > forever. > > > > Each new user of this interface that needs access to the parameters > not > > > > exposed originally needs to implement the abstract method with an > empty > > > > implementation and it needs to override whichever additional default > they > > > > care about (this KIP introduces a second method, but future KIPs > would > > > > introduce additional methods for new parameters). One would never > design > > > > the interface like this from the start. > > > > > > > > With the DTO approach, you don't create a new DTO, you simply add a > new > > > > overloaded constructor and accessor to the DTO. The implementers of > the > > > > interface still have a single method (two here since we made a > mistake > > > > originally) and they can decide which of the values from the DTO they > > > would > > > > like to access. This approach has been the recommended approach for > years > > > > and it's how the Admin apis work (they're the most recent client). An > > > > example: > > > > > > > > createTopics(Collection<NewTopic> newTopics, CreateTopicsOptions > > > options); > > > > > > > > This makes it easy to add new fields to `NewTopic` or > > > `CreateTopicsOptions`. > > > > > > > > Ismael > > > > > > > > On Wed, Aug 23, 2023 at 11:48 AM Greg Harris > > > <greg.har...@aiven.io.invalid> > > > > wrote: > > > > > > > > > Hey Jack, > > > > > > > > > > The design of this KIP is also consistent with the way header > support > > > > > was added to Connect: > > > > > > > > > > > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP-440%3A+Extend+Connect+Converter+to+support+headers > > > > > I think making argument for precedent here is reasonable. > > > > > > > > > > Hi Ismael, > > > > > > > > > > Can you expand what you mean by "without breaking compatibility"? I > > > > > think the approach proposed here (a default method) would be > backwards > > > > > compatible. If an implementation wishes to make use of the new > > > > > signature, they can override the new method and the version of > Kafka > > > > > will determine which implementation is used without instance > checking, > > > > > reflection, or exceptions. > > > > > > > > > > I believe that when you pass a DTO, that some sort of instance > > > > > checking, reflection, or exceptions would be required for the > > > > > Partitioner to determine whether additional information is present. > > > > > For example, if we wished to add some information X to the > partitioner > > > > > in the future, the caller could pass either a `PartitionInfo` or > > > > > `PartitionInfoWithX` DTO instance, and the callee could use an > > > > > `instanceof` check and a cast before accessing the X information. > That > > > > > seems to be more machinery for the Partitioner implementation to > > > > > manage as compared to maintaining multiple methods, which may just > be > > > > > one-line calls to other methods. > > > > > > > > > > Please let me know if I've misunderstood your DTO design. > > > > > > > > > > Thanks! > > > > > Greg > > > > > > > > > > On Tue, Aug 22, 2023 at 9:33 PM Jack Tomy <jacktomy...@gmail.com> > > > wrote: > > > > > > > > > > > > Hi Ismael, > > > > > > > > > > > > That would be totally different from the pattern currently being > > > followed > > > > > > in all the interfaces, for example serializer. > > > > > > I personally don't favour that either. Let's see if the community > > > has any > > > > > > opinions on the same. > > > > > > > > > > > > Hey everyone, please share your thoughts on using a DTO instead > of > > > > > separate > > > > > > params for the interface. > > > > > > > > > > > > Thanks. > > > > > > > > > > > > On Mon, Aug 21, 2023 at 8:06 PM Ismael Juma <m...@ismaeljuma.com> > > > wrote: > > > > > > > > > > > > > Hi Jack, > > > > > > > > > > > > > > I mean a DTO. That means you can add additional parameters > later > > > > > without > > > > > > > breaking compatibility. The current proposal would result in > yet > > > > > another > > > > > > > method each time we need to add parameters. > > > > > > > > > > > > > > Ismael > > > > > > > > > > > > > > On Sun, Aug 20, 2023 at 4:53 AM Jack Tomy < > jacktomy...@gmail.com> > > > > > wrote: > > > > > > > > > > > > > > > Hey Ismael, > > > > > > > > > > > > > > > > Are you suggesting to pass a param like a DTO or you are > > > suggesting > > > > > to > > > > > > > pass > > > > > > > > the record object? > > > > > > > > > > > > > > > > I would also like to hear other devs' opinions on this as I > > > > > personally > > > > > > > > favour what is done currently. > > > > > > > > > > > > > > > > On Thu, Aug 17, 2023 at 9:34 AM Ismael Juma < > m...@ismaeljuma.com> > > > > > wrote: > > > > > > > > > > > > > > > > > Hi, > > > > > > > > > > > > > > > > > > Thanks for the KIP. The problem outlined here is a great > > > example > > > > > why we > > > > > > > > > should be using a record-like structure to pass the > parameters > > > to a > > > > > > > > method > > > > > > > > > like this. Then we can add more parameters without having > to > > > > > introduce > > > > > > > > new > > > > > > > > > methods. Have we considered this option? > > > > > > > > > > > > > > > > > > Ismael > > > > > > > > > > > > > > > > > > On Mon, Aug 7, 2023 at 5:26 AM Jack Tomy < > > > jacktomy...@gmail.com> > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > Hey everyone. > > > > > > > > > > > > > > > > > > > > I would like to call for a vote on KIP-953: partition > method > > > to > > > > > be > > > > > > > > > > overloaded to accept headers as well. > > > > > > > > > > > > > > > > > > > > KIP : > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=263424937 > > > > > > > > > > Discussion thread : > > > > > > > > > > > > > https://lists.apache.org/thread/0f20kvfqkmhdqrwcb8vqgqn80szcrcdd > > > > > > > > > > > > > > > > > > > > Thanks > > > > > > > > > > -- > > > > > > > > > > Best Regards > > > > > > > > > > *Jack* > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > Best Regards > > > > > > > > *Jack* > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > Best Regards > > > > > > *Jack* > > > > > > > > >
