TP

# Embedding effect systems in Haskell

Dominic Orchard, Tomas Petricek

Monads are now an everyday tool in functional programming for abstracting and delimiting effects. The link between monads and effect systems is well-known, but in their typical use, monads provide a much more coarse-grained view of effects. Effect systems capture fine-grained information about the effects, but monads provide only a binary view: effectful or pure.

Recent theoretical work has unified fine-grained effect systems with monads using a monad-like structure indexed by a monoid of effect annotations (called parametric effect monads). This aligns the power of monads with the power of effect systems.

This paper leverages recent advances in Haskell's type system (as provided by GHC) to embed this approach in Haskell, providing user-programmable effect systems. We explore a number of practical examples that make Haskell even better and safer for effectful programming. Along the way, we relate the examples to other concepts, such as Haskell's implicit parameters and coeffects.

 1: 2: 3: 4: 5: 6: 7: 8:  @incollection{haskell-effects, title={Embedding effect systems in Haskell}, author={Orchard, Dominic and Petricek, Tomas}, booktitle = {Proceedings of Haskell Symposium}, series = {Haskell 2014}, location = {Gothenburg, Sweden}, year = {2014} }