Modular Modelling with Monads.pdfVIP

Modular Modelling with Monads Christoph L¨uth FB 3 — Mathematik und Informatik Universit¨at Bremen rmatik.uni-bremen.de/~cxl November 26, 2005 One useful application of category theory in computer science are computa- tional monads, which are used to model diverse computational effects (such as stateful computations, exceptional behaviour or non-determinism). This goes back to Moggi [6], and is for example used to great effect in the pure functional language Haskell [8, 11]. In this talk, we use the same technique to model com- putational effects in higher-order logic, and in particular the theorem prover Isabelle [7], thus shallowly embedding a small imperative language embedded shallow into Isabelle. The talk consists of three parts: we first recall the basic notions of monads, then show how the imperative language is built from three basic ingredients, and finally discuss the implementation. Monads A monad is the categorical modelling of an algebraic theory (that is, a set of operations and equations on them). One attraction of monads is that they come with a rich categorical model theory, giving a uniform treatment of both algebras and terms. The latter are given by the Kleisi category of a monad; this is where the computations modelled by the monad live. An adjunction between the base category and the Kleisli category gives us a notion of lifting every function from the base to a pure computation. To combine monads, we can (under some mild preconditions on the base category) use colimits [3]. However, as the construction of colimits in general is quite intricate, two special cases of combinations of monads which have been considered: the coproduct, in particular