Validity invariants and effects

  • Authors:

  • Yi Lu;John Potter;Jingling Xue

  • Affiliations:

  • School of Computer Science and Engineering, University of New South Wales, Sydney;School of Computer Science and Engineering, University of New South Wales, Sydney;School of Computer Science and Engineering, University of New South Wales, Sydney

  • Venue:
  • ECOOP'07 Proceedings of the 21st European conference on Object-Oriented Programming
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

Object invariants describe the consistency of object states, and are crucial for reasoning about the correctness of object-oriented programs. However, reasoning about object invariants in the presence of object abstraction and encapsulation, arbitrary object aliasing and re-entrant method calls, is difficult. We present a framework for reasoning about object invariants based on a behavioural contract that specifies two sets: the validity invariant-- objects that must be valid before and after the behaviour; and the validity effect--objects that may be invalidated during the behaviour. The overlap of these two sets is critical because it captures precisely those objects that need to be re-validated at the end of the behaviour. When there is no overlap, no further validity checking is required. We also present a type system based on this framework using ownership types to confine dependencies for object invariants. In order to track the validity invariant, the type system restricts updates to permissible contexts, even in the presence of re-entrant calls. Object referencing and read access are unrestricted, unlike earlier ownership type systems.