Conclusion

In this paper we have argued that in the term ``Attribute Grammar'' the notion of grammar does not necessarily imply the existence of an underlying tree, and that the notion of attribute does not necessarily mean decoration of a tree. We have presented two simple extensions to the Attribute Grammar formalism that allow the full exploitation of the power of this observation. They are consistent with the general ideas underlying Attribute Grammars, hence we retain the benefits of the results and techniques that are already available in that domain.

Our goal in providing these extensions to the Attribute Grammar formalism is to bring this powerful tool into a larger context of usefulness and applicability. The Attribute Grammar\ programming style (declarative and structured) and existing Attribute Grammar techniques (static analysis) become more general under this extended view and reveal themselves as complementary to other formalisms such as functional programming or inference rule programming.

This approach is of practical interest because the mechanisms necessary to support the extensions proposed here were already part of the FNC-2 system, which has proved its usefulness on real applications; this made their implementation easy. It is also promising because it opens the way to the application of good results developped for Attribute Grammars to other programming paradigms.



Web page maintained by Didier Parigot
Fri Feb 27 10:05:31 MET 1998