----------------------------------------------------------------------------- -- | -- Module : Control.Arrow.Operations -- Copyright : (c) Ross Paterson 2003 -- License : BSD-style (see the LICENSE file in the distribution) -- -- Maintainer : ross@soi.city.ac.uk -- Stability : experimental -- Portability : non-portable (multi-parameter type classes) -- -- Subclasses of 'Arrow' providing additional operations. -- -- The signatures are designed to be compatible with the proposed -- notation for arrows, cf. <http://www.haskell.org/arrows/>. module Control.Arrow.Operations ( -- * Conventions -- $conventions -- * State transformers ArrowState(..), -- * State readers ArrowReader(..), -- * Monoid writers ArrowWriter(..), -- * Errors ArrowError(..), tryInUnlessDefault, -- * Synchronous circuits ArrowCircuit(..), ) where import Control.Arrow import Data.Monoid -- $conventions -- The arrow classes defined in this module have names like @Arrow@/Foo/, -- and contain operations specific to such arrows. Some of these include -- a method @new@/Foo/, which maps computations to computations of the -- same arrow type, but exposing some of the internals of the arrow. -- -- Arrow transformers have names like /Bar/@Arrow@, and are -- instances of appropriate arrow classes. For each arrow -- transformer, there is typically an encapsulation operator -- @run@/Bar/ that removes that transformer from the outside of an -- arrow type. The 'Control.Arrow.Transformer.lift' method of the -- 'Control.Arrow.Transformer.ArrowTransformer' class adds an arrow -- transformer to the outside of an arrow type. -- -- Typically a composite arrow type is built by applying a series of arrow -- transformers to a base arrow (usually either a function arrow or a -- 'Kleisli' arrow. The 'Control.Arrow.Transformer.lift' method and the -- @run@/Bar/ function operate only on the arrow transformer at the top -- of this stack. For more sophisticated manipulation of this stack of -- arrow transformers, many arrow transformers provide an @ArrowAdd@/Bar/ -- class, with methods methods @lift@/Bar/ and @elim@/Bar/ to add and remove -- the transformer anywhere in the stack. -- | An arrow type that provides a read-only state (an environment). -- If you also need to modify the state, use 'ArrowState'. class Arrow a => ArrowReader r a | a -> r where -- | Obtain the current value of the state. readState :: a b r -- | Run a subcomputation in the same arrow, but with a different -- environment. The environment of the outer computation is -- unaffected. -- -- Typical usage in arrow notation: -- -- > proc p -> ... -- > (|newReader cmd|) env newReader :: a e b -> a (e,r) b -- | An arrow type that provides a modifiable state, -- based of section 9 of /Generalising Monads to Arrows/, by John Hughes, -- /Science of Computer Programming/ 37:67-111, May 2000. class Arrow a => ArrowState s a | a -> s where -- | Obtain the current value of the state. fetch :: a e s -- | Assign a new value to the state. store :: a s () -- | An arrow type that collects additional output (of some 'Monoid' type). class (Monoid w, Arrow a) => ArrowWriter w a | a -> w where -- | Add a piece of additional output. write :: a w () -- | Run a subcomputation in the same arrow, making its additional -- output accessible. -- -- Typical usage in arrow notation: -- -- > proc p -> do -- > ... -- > (value, output) <- (|newWriter cmd|) newWriter :: a e b -> a e (b,w) -- | An arrow type that includes errors (or exceptions). -- -- Minimal definition: 'raise' and 'tryInUnless'. -- -- /TODO:/ the operations here are inconsistent with other arrow transformers. class Arrow a => ArrowError ex a | a -> ex where -- | Raise an error. raise :: a ex b -- | Traditional exception construct. -- -- Typical usage in arrow notation: -- -- > proc p -> ... -- > body `handle` \ex -> handler handle :: a e b -- ^ computation that may raise errors -> a (e,ex) b -- ^ computation to handle errors -> a e b handle f h = tryInUnless f (arr snd) h -- | Exception construct in the style of /Exceptional Syntax/, -- by Nick Benton and Andrew Kennedy, /JFP/ 11(4):395-410, July 2001. -- -- Typical usage in arrow notation: -- -- > proc p -> ... -- > (|tryInUnless -- > body -- > (\res -> success) -- > (\ex -> handler) -- > |) tryInUnless :: a e b -- ^ computation that may raise errors -> a (e,b) c -- ^ computation to receive successful results -> a (e,ex) c -- ^ computation to handle errors -> a e c -- | Handler that returns the error as a value. newError :: a e b -> a e (Either ex b) newError f = handle (f >>> arr Right) (arr (Left . snd)) -- | A suitable value for 'tryInUnless' when the arrow type belongs to -- 'ArrowChoice'. To use it, you must define either 'handle' or 'newError'. tryInUnlessDefault :: (ArrowError ex a, ArrowChoice a) => a e b -- ^ computation that may raise errors -> a (e,b) c -- ^ computation to receive successful results -> a (e,ex) c -- ^ computation to handle errors -> a e c tryInUnlessDefault f s h = arr id &&& newError f >>> arr dist >>> h ||| s where dist (e, Left ex) = Left (e, ex) dist (e, Right b) = Right (e, b) -- tryInUnless (and thus handle) could be replaced by newError if: -- 1. When ArrowChoice is available, tryInUnless and newError are equivalent. -- 2. When tryInUnless is available, so is ArrowChoice. -- (Counterexample: general CoKleisli) -- | An arrow type that can be used to interpret synchronous circuits. class ArrowLoop a => ArrowCircuit a where -- | A delay component. delay :: b -- ^ the value to return initially. -> a b b -- ^ an arrow that propagates its input -- with a one-tick delay.