ResolvableType should support resolving lambda types [SPR-13698]

[spring-projects-issues]

Oliver Drotbohm opened SPR-13698 and commented

This test case currently fails:

Function<Long, String> toString = value -> value.toString();

ResolvableType type = ResolvableType.forClass(Function.class, toString.getClass());

assertThat(type.getGeneric(0).getRawClass(), is(typeCompatibleWith(Long.class)));
assertThat(type.getGeneric(1).getRawClass(), is(typeCompatibleWith(String.class)));

This helper library (itself inspired by Spring's GenericTypeResolver) seems to use an additional code path to use API introduced in Java 8 to resolve lambdas in a special way.

---

Issue Links:

• GenericTypeResolver should be able to introspect generic arguments from lambdas [SPR-12525] #17130 GenericTypeResolver should be able to introspect generic arguments from lambdas ("is duplicated by")
• Add a functional way to register a bean [SPR-14832] #19398 Add a functional way to register a bean

[spring-projects-issues]

Stéphane Nicoll commented

I thought we had already such report (with a lengthy discussion with thomas.darimont) but I can't find it anymore

[spring-projects-issues]

Juergen Hoeller commented

Indeed, we rejected in back then since the "API introduced in Java 8" is actually internal (sun.reflect.ConstantPool), and the entire approach is rather hacky, even relying on the index of entries in the constant pool... This is very likely to break on JDK 9 eventually, so I'm reluctant to add support for it at this point.

Juergen

[spring-projects-issues]

Oliver Drotbohm commented

Maybe we should bring this up with the JDK team then, they might be able to fix this? I haven't looked deeply into the code but is there a particular reason the standard lookup algorithm doesn't work here? I just think it's kind of weird to see a lambda failing while the very same declaration works if you create the anonymous class. It basically subverts our "use Java 8 and everything will be better" message to some degree.

[spring-projects-issues]

Oliver Drotbohm commented

I just ran into this again with code that was expecting a Converter and I thought I could just hand in a lambda, which I couldn't as the called code was inspecting the given instance for generic types. I think that's the main issue with this one: for a user it's not at all obvious whether you can use a lambda with a SAM type or not.

[spring-projects-issues]

Oliver Drotbohm commented

For reference, here's the discussion I had with the JDK team: http://mail.openjdk.java.net/pipermail/compiler-dev/2016-December/010623.html

[spring-projects-issues]

Bulk closing outdated, unresolved issues. Please, reopen if still relevant.