Copyright | (c) Ross Paterson 2003 |
---|---|
License | BSD-style (see the LICENSE file in the distribution) |
Maintainer | R.Paterson@city.ac.uk |
Stability | experimental |
Portability | non-portable (multi-parameter type classes) |
Safe Haskell | Safe |
Language | Haskell98 |
An arrow transformer that adds error handling.
TODO: the operations here are inconsistent with other arrow transformers.
- newtype ErrorArrow ex a b c = ErrorArrow (a b (Either ex c))
- runError :: ArrowChoice a => ErrorArrow ex a e b -> a (e, ex) b -> a e b
- class (ArrowError ex a, Arrow a') => ArrowAddError ex a a' | a -> a' where
Documentation
newtype ErrorArrow ex a b c Source #
An arrow that augments an existing arrow with possible errors.
The ArrowError
class contains methods for raising and handling
these errors.
ErrorArrow (a b (Either ex c)) |
:: ArrowChoice a | |
=> ErrorArrow ex a e b | computation that may raise errors |
-> a (e, ex) b | computation to handle errors |
-> a e b |
Encapsulate an error-raising computation, by completely handling any errors.
Typical usage in arrow notation:
proc p -> ... body `runError` \ex -> handler
class (ArrowError ex a, Arrow a') => ArrowAddError ex a a' | a -> a' where Source #
Adding a ErrorArrow
to an
arrow type, but not necessarily as the outer arrow transformer.
Typically a composite arrow type is built by applying a series
of arrow transformer to a base arrow (usually either a function
arrow or a Kleisli
arrow. One can add a transformer to the
top of this stack using the lift
method of the ArrowTransformer
class,
or remove a state transformer from the top of the stack using the
runError
encapsulation operator.
The methods of this class add and remove state transformers anywhere
in the stack. In the instance
instance Arrow a => ArrowAddError ex (ArrowError ex a) a
they are equivalent to lift
and
runError
respectively.
Instances are lifted through other transformers with
instance ArrowAddError ex a a' => ArrowAddError ex (FooArrow a) (FooArrow a')
This could be combined with handle
,
since the resulting arrow is always the arrow of the handler.
Separating them has the advantage of consistency with the other arrows,
and might give more helpful type error messages.
liftError :: a' e b -> a e b Source #
Lift a computation from an arrow to one with error handling.
Typical usage in arrow notation:
proc p -> ... (|liftError cmd|)
elimError :: a e b -> a' (e, ex) b -> a' e b Source #
Elimination of errors from a computation, by completely handling any errors.
Typical usage in arrow notation:
proc p -> ... body `elimError` \ex -> handler
ArrowChoice a => ArrowAddError ex (ErrorArrow ex a) a Source # | |
ArrowAddError ex a a' => ArrowAddError ex (ReaderArrow r a) (ReaderArrow r a') Source # | |
ArrowAddError ex a a' => ArrowAddError ex (StateArrow s a) (StateArrow s a') Source # | |
(ArrowAddError ex a a', Applicative f) => ArrowAddError ex (StaticArrow f a) (StaticArrow f a') Source # | |
(ArrowAddError ex a a', Monoid w) => ArrowAddError ex (WriterArrow w a) (WriterArrow w a') Source # | |