Y combinator using lambda and method references

[Brian Goetz]

The inference rules are still under development so it is not surprising 
that the IDE and the static compiler do not yet agree.  When the spec 
settles, hopefully both will agree :)

On 10/27/2012 5:20 PM, Arul Dhesiaseelan wrote:
> I attempted to port this Y combinator [1] using Java 8 lambdas. The closest
> I can make it work is shown here [2].
>
> It looks like intellij code inspection suggests I can still reduce this
> further by using 2 method references and 1 lambda as shown visually here
> [3], but I run into problems with those conversions.
>
> # 1 Applying the first inspection (line # 19) to a method reference fails
> in a compile error:
> error: cannot find symbol variable f
>
> #2 Applying the second inspection (line # 22) to a method reference
> compiles just fine, but runs into a StackOverflowError. May be this is a
> problem with the code itself.
>
> #3 Applying the third inspection (line # 31) to a lambda fails in a compile
> error:
>
> error: method Y in class YFact cannot be applied to given types;
> required: Func<Func<T>>
> found: (final Fun[...] - 1)
> reason: cyclic inference - cannot infer target type for given lambda/method
> reference expression
> where T is a type-variable:
> T extends Object declared in method <T>Y(Func<Func<T>>)
>
> I am not sure if this is a case where the editor is incorrect in detecting
> compilation errors. I believe the editor should reject these if the
> compiler cannot infer the target type.
>
> Please apologize if this does not belong here.
>
> - Arul
>
> [1] http://www.arcfn.com/2009/03/y-combinator-in-arc-and-java.html
> [2] https://gist.github.com/3965968
> [3] https://dl.dropbox.com/u/3274664/java/yfact-idea-inspections.png
>