ArcAngel: a tactic language for refinement

Morgan's refinement calculus is a successful technique for developing software in a precise and consistent way. This technique, however, can be hard to use, as developments may be long and repetitive. Many authors have pointed out that a lot can be gained by identifying commonly used development strategies, documenting them as tactics, and using them as single transformation rules. Also, it is useful to have a notation for describing derivations, so that they can be analysed and modified. In this paper, we present ArcAngel, a language for defining such refinement tactics; we present the language, its semantics, and some of its algebraic laws. Apart from Angel, a general-purpose tactic language that we are extending, no other tactic language has a denotational semantics and proof theory of its own.