A language independent framework for context-sensitive formatting

Abstract

Automated formatting is an important technique for the software maintainer. It is either applied separately to improve the readability of source code, or as part of a source code transformation tool chain. In this paper we report on the application of generic tools for constructing formatters. In an industrial setting, automated formatters need to be tailored to the requirements of the customer. The (legacy) programming language or dialect and the corporate formatting conventions are specific and non-negotiable. Can generic formatting tools deal with such unexpected requirements? Driven by an industrial case of nearly 80 thousand lines of COBOL code, several limitations in existing formatting technology have been addressed. We improved its flexibility by replacing a generative phase by a generic tool, and we added a little expressiveness to the formatting back end. Most importantly, we employed a multi-stage formatting framework that can cope with any kind of formatting convention using more computational power