Java Closure Proposals Compared, JCA Position Paper Announced
Howard Lovatt, the author of the C3S proposal for closures in Java, has written a detailed comparison of the four best known proposals (C3S, FCM, CICE, and BGGA). At the same time the authors of the FCM proposal have released a new position paper building on FCM for control abstraction. Ricky Clarkson thinks that CICE is insufficient and wonders if internal politics at Google are affecting it.
Lovatt walks through the four proposals and compares each in terms of eleven features:
- Short syntax for the creation of an instance of an inner class/closure
- Access to both this pointers and methods within an inner class
- More than one method in an inner class/closure instance
- Implementation of methods defined in classes
- Type inference
- Method, constructor, and field literals
- Short syntax (particularly for control structures and short methods)
- Assignment to local variables and no final requirement
- Variable number of exceptions
- Method/function types (with shorter syntax)
- Non-local, return, break, and continue
His goal in writing is to separate out the inner class / closure part of the proposal from any other extras that might come with it. On the heels of this comparison, Stephen Colebourne, Stefan Schulz, and Ricky Clarkson have built upon FCM with a position paper (they clarify that it is not a proposal and thus not complete) for Java Control Abstraction. They motivate the need for JCA because there are places closures are not applicable. Such a place is where an API exists that is used very similarly to a built in keyword.
Lastly, Ricky Clarkson wonders if politics at Google have affected the advancement of BGGA proposal. There are restrictions on who is able to contribute to the JCP, specifically if your employer is a member, you cannot be. Clarkson suggests that Josh Bloch, Google's contact on the JCP, may be uninterested in creating a JSR for as he has already proposed CICE. Bob Lee, one of the co-authors of CICE responds to Clarkson saying this is an unfair suggestion and that there is no rush to get closures into Java.
Josh Bloch recently replied to a blog post somewhere to remind us of his solution to many of the use cases that CICE alone misses out on - docs.google.com/View?docid=dffxznxr_1nmsqkz - I still think with CICE and ARM blocks he's likely to miss some use cases, but I haven't identified any yet.