Tobias Wrigstad, Francesco Zappa Nardelli, Sylvain Lebresne, Johan Östlund, Jan Vitek.
Many large software systems originate from untyped scripting language code. While good for initial development, this can impact code-quality and performance in the long run. We study a novel approach for integrating untyped and typed code in the same system to allow an initial script to be smoothly evolved into a more efficient and robust system. Our approach adds a novel intermediate point, called like types, between dynamic and static types. Uses of variables of like type are checked statically, while conformance of a value to a like type is checked at run-time individually at each access. In contrast to previous proposals, this approach captures common programmer mistakes, is compatible with object-orientation and can be implemented efficiently. The parts of a program annotated with concrete types enjoy the usual static safety guarantees and can be optimised, while like types improve local safety and integrate smoothly with untyped parts of a program without hindering performance.
Last update: 3-11-2009|