RFR: JDK-8304439: Subscription based listeners [v2]

[Kevin Rushforth]

On Fri, 16 Jun 2023 09:45:04 GMT, John Hendrikx <jhendrikx at openjdk.org> wrote:

>> I agree that the chosen names `invalidation`, `changes` and `values` are a bit terse.  The whole signature (without reading docs) should make it clear you are creating a subscription, but perhaps we can do better.  The use of `addListener` can be ruled out as it would conflict with the existing method due to having Lambda's with the same arity (the `values` listener would conflict with `addListener(InvalidationListener)`.  Also, an `add` method would probably have users expecting a corresponding `remove` method.
>> 
>> A few ideas listed here:
>> 
>> | invalidation | values | changes |
>> |---|---|---|
>> |`subscribe(Runnable)`(*)|`subscribe(Consumer)`(*)|`subscribe(BiConsumer)`(*)|
>> |`subscribeInvalidations(Runnable)`|`subscribeValues(Consumer)`|`subscribeChanges(BiConsumer)`|
>> |`invalidationsTo(Runnable)`|`valuesTo(Consumer)`|`changesTo(BiConsumer)`|
>> 
>> (*) May limit future listener types that have same arity, but can still be a good choice
>
> On that same topic of naming methods:
> 
> What do people think of `Subscription#unsubscribe`?  Should it be `cancel`? Something else?  Okay as it is?
> 
> Code example:
> 
>         if (subscription != null) {
>             subscription.unsubscribe();
>             subscription = null;
>         }

> | invalidation | values | changes |
> |---|---|---|
> |`subscribe(Runnable)`(*)|`subscribe(Consumer)`(*)|`subscribe(BiConsumer)`(*)|
> |`subscribeInvalidations(Runnable)`|`subscribeValues(Consumer)`|`subscribeChanges(BiConsumer)`|
> |`invalidationsTo(Runnable)`|`valuesTo(Consumer)`|`changesTo(BiConsumer)`|
> 
> (*) May limit future listener types that have same arity, but can still be a good choice

My preference is in the order you listed them. If we go with `subscribe`, using method name overloading, do want to add a future overload that takes a type with the same arity as one of the existing three, we can always assign a new name to that new method (since adding another overload wouldn't be source compatible, we would likely need a new name at that point).

@nlisker @andy-goryachev-oracle @mstr2 - what do you think?

-------------

PR Review Comment: https://git.openjdk.org/jfx/pull/1069#discussion_r1240440189