Object-oriented type inference

Abstract

We present a new approach to inferring types in untyped object-oriented programs with inheritance, assignments, and late binding. It guarantees that all messages are understood, annotates the program with type information, allows polymorphic methods, and can be used as the basis of an optimizing compiler.Types are finite sets of classes and subtyping is set inclusion. Using a trace graph, our algorithm constructs a set of conditional type constraints and computes the least solution by least fixed-point derivation.

From well-formedness to meaning preservation: model refactoring for almost free

“…1b is also well-formed. Admittedly, the typing rule expressed in (2) is somewhat simplistic in that it does not consider inheritance or subtyping; however, we do not delve into the technicalities necessary for this here because they complicate matters unduly (and have been addressed in great detail elsewhere; see, e.g., [11,24]). …”

“…3.4, the premise must contain only constraints whose properties are fixed for all model elements quantified over (so that they can be evaluated at rule application time). Thus, the first step in transforming a well-formedness rule such as (8) into a constraint rule is rewriting it to the form ∀l, m : l ∈ lifelines ∧ m ∈ l.messages → m.operation ∈ l.classifier.operations ∧ m.name = m.operation.name (11) The remainder of the transformation depends on which properties are fixed and which are non-fixed for a given refactoring, as shown in Table 1 for three sample refactorings. For RENAME, name is the only non-fixed property so that (11) can be transformed to l ∈ lifelines m ∈ l.messages m.name = m.operation.name (12) This is so because l.messages never changes so that the constraint m ∈ l.messages can be evaluated at rule application time, after m and l have been bound to concrete model elements (recall that both m and l are implicitly universally quantified), and because neither of m.operation, l.classifier, and l.classifier.operations can change their values, so that m.operation ∈ l.classifier.operations must remain satisfied and can be dropped (as stated in Sect.…”

The Theory of Classification Part 20: Modular Checking of Classtypes.

“…In a parametric type system, we would expect to be able to recover the exact vehicle-type. This is because, when the Location is constructed with a value of the exact type Car, this type is propagated into the vehicle-type parameter τ of Location's polymorphic variable myVeh, which we imagine might have the type: [15]. They were using a type substitution mechanism, which has only slightly less expressive power than the full parametric mechanism used in our approach 2 .…”

Reifying configuration management for object-oriented software

“…Ole Agesen's recent PhD thesis [2] contains a complete survey of related work. Reviewed systems range from purely theoretical ones [33] to systems in regular use by a large community [22], via partially implemented systems [29,30] and systems implemented on small languages [14,25].…”