Development Environment for Language Processors Based on Attribute Grammars

To develop language processors efficiently is difficult because they need to deal with large data with complex structures. Using tools based on formal specification is one of the strategies to reduce the cost in developing language processors. However, the area of application which can be developed by these tools is limited. This dissertation aims at realization of the practical development environment for language processors based on the formalism of attribute grammars (AGs). AGs are a formalism that integrates syntax and semantics of languages. AGs have several advantages for development of language processors in that the formalization is intuitive and clear, and that attribute evaluators are automatically obtained from descriptions. However, the following disadvantages make it difficult to develop language processors by AGs: (1) difficulty in direct application of the pure (standard) AG formalism to tools because of the weak computation power, and (2) complication in debugging of AG descriptions, which arises from the difficulty specific to AGs. To overcome the first problem, we proposed circular remote attribute grammars (CRAGs), an extension of AGs that allows both circularity of attribute dependency and remote attribute references. We also showed a way to obtain efficient evaluators for CRAGs. By this research any control structure of the subject language can be treated and circularities of attribute dependency are allowed in AGs. As a result, stronger expressive power is obtained and efficient attribute evaluators are developed. For the second problem, the generalized framework for systematic debugging of AGs was proposed, which allows several forms of query and enables integration of various query based methods. This reduces the complication in debugging of AG descriptions. These accomplishments widen the application area of attribute grammar based software, which improves the development environment for language