Combining domain-specific and general-purpose languages to improve program analysis

点击翻译按钮获取中文摘要

Programming, a key activity in software development, requires a programmer to laboriously encode a high-level understanding of a problem domain into code---a set of precise instructions, suitable for a computer to execute. Domain-specific languages (DSLs) help bridge the gap between the problem domain and the code by mitigating the conceptual mismatch between the software's high-level design and the low-level details required in its implementation. However, because DSLs are very specific, they do not generally suffice for implementing complete software systems. In many cases, they must be combined with a general-purpose language. Recently, program analysis tools have been gaining popularity. For instance, Microsoft requires that certified Windows device drivers pass an automatic program analysis. However, programs written in a combination of a DSL and a general-purpose language are harder to analyze than single-language programs. The proposed research will develop techniques for: 1) developing domain-specific languages that cleanly integrate with general-purpose languages; 2) inferring the domain-specific abstractions that underlie code, thus improving the code's maintainability and potentially its performance; and 3) analyzing and automatically improving programs written in a combination of domain-specific languages and general purpose languages. The impact of this research is tightly linked to software's increasing ubiquity in today's society. In particular, the programming tools produced as a result of this research will help Canadian industry develop better software more quickly.