Programming language evolution

Programming languages are the way developers communicate with computers—just like natural languages let us communicate with one another. In both cases multiple languages have evolved. However, the programming language ecosystem changes at a much higher rate compared to natural languages. In fact, programming languages need to constantly evolve in response to user needs, hardware advances and research developments or they are otherwise displaced by newcomers. As a result, existing code written by developers may stop working with a newer language version. Consequently, developers need to search (analyse) and replace (refactor) code in their code bases to support new language versions. Traditionally, tools have focused on the replace aspect (refactoring) to support developers evolving their code bases. This dissertation argues that developers also need machine support focused on the search aspect. This dissertation starts by characterising factors driving programming language evolution based on external versus internal forces. Next, it introduces a classification of changes introduced by language designers that affect developers. It then contributes three techniques to support developers in analysing their code bases. First, we show a source code query system based on graphs that express code queries at a mixture of syntax-tree, type, control-flow graph and data-flow levels. We demonstrate how this system can support developers and language designers in locating various code patterns relevant in evolution. Second, we design an optional run-time type system for Python, that lets developers manually specify contracts to identify semantic incompatibilities between Python 2 and Python 3. Third, recognising that existing codebases do not have such contracts, we describe a dynamic analysis to automatically generate them.