On language processors and software maintenance

Software maintenance factories are infrastructures that generate tools for software maintenance from language descriptions and additional information. Besides the maintenance of language specifications, software language evolution requires maintenance of software written in that language as well as maintenance of software that transforms software written in the evolving language. We argue that a software maintenance factory needs special maintenance. It should provide support for maintenance of language descriptions, derived tools, and the maintenance factory itself. As language processors can form the core of such technology, we investigate their improvement. This is done in three areas: Maintenance: We investigate the consequences of software language evolution. A grammar change can have an effect on the words of the language and can require an adaptation of documents containing the words. A special case is the format evolution of data and a necessary adaptation of the data existing. We study this problem for format evolution for DTDs. While the words of the language remain the same after the refactoring of a context-free grammar, semantic rules have to be adapted in an attribute grammar. We study the consequences of left-recursion removal to attribute grammars and justify that the semantic rules can be automatically adapted while preserving the semantics. Declarative transformation rules can transform programs that are available as abstract syntax trees. When an underlying grammar is extended, existing transformation rules need to be adapted. To study the problem we introduce layout preservation to an existing language processor and demonstrate that transformation rules can be automatically adapted. By defining an aspect-oriented model for Prolog we support modularity of language descriptions based on logic grammars, and consequently, that of derived language processors, which makes them better maintainable, replaceable, and configurable. Technical improvement: By defining basic roles of program extension we provide a way for more disciplined meta-programming for declarative programs including language descriptions. Using language technology we introduce layout preservation in existing language processors for maintenance. We analyse transformation rules regarding their layout preservation behaviour. Besides using the analyse to for automatic decisions on layout preservation, it can advice the coder of a transformation rule to change the specification style to make layout preservation more likely. Usability for the user: Disciplined meta-programming increases the usability, as it allows to define reusable and better understandable recipes of meta-programs. We discuss that the left recursion removal with automatic migration of semantic rules can help to preserve readability of grammars in use. We show that complexity of argument positions in transformation rules can be hidden by applying the approach of grammar extension and automatically migrating the transformation rules. We provide an object-oriented combination model for transformations. Based on it, a workbench for visual interactive explorative transformational programming has been implemented with grammars and transformations as first class citizens.