Reifying wildcards in Java using the EGO approach

Abstract

Providing runtime information about generic types-that is, reifying generics-is a challenging problem studied in several research papers in the last years. However, the quest for finding effective and efficient solutions specifically targeted to the Java programming language is still open. In particular, the new mechanism of wildcards introduced in Java 5.0 significantly complicates the overall semantics of generics: its reification aspects are currently unexplored and pose serious implementation issues.In this paper we analyse such issues and study how they have been supported in the context of the EGO compiler. EGO is an approach for efficiently supporting runtime generics at compile-time in Java: synthetic code is automatically added to the source code by the compiler, so as to create generic runtime type information on a by-need basis, store it into object instances, and retrieve it when necessary in typedependent operations. We show how the design of EGO has been completely and successfully extended to represent wildcard types, and how we deal with subtle issues concerning subtyping, capture conversion and wildcards capture in method calls.

On the reification of Java wildcards

“…Such a language is currently not deployed, hence it is very difficult to gather large-size source code upon which performing correctness/performance tests. Other than small-size synthetic programs, our reference case was the code for javac itself, which largely relies on generics and performs some legacy-style type conversions to unbounded generics such as C<?> [5]. From that experience, we were able to observe the correctness of our approach with respect to the results of the compilation process, and a zero overhead with respect to the performance of EGO before supporting wildcards.…”

Java Wildcards Meet Definition-Site Variance

“…. OK and (b) ∆ C<T> ≺: N and (c) ftype(f; N ) = T, then ∆ ftype(f; C<T>) <: T. 6 Class and Method Typing: …”

“…Our approach offers a generalization and a high-level way to reason soundly about the variance of a type. Other recent work discusses the complex relationship between type-erasure and wildcards [6], as well as the concept of variance at the level of tuning access to a path type in tree-like class definitions [12].…”