Difference between revisions of "New monads"

From HaskellWiki
Jump to navigation Jump to search
(add link to MonadLib)
 
(31 intermediate revisions by 11 users not shown)
Line 1: Line 1:
 
__TOC__
 
__TOC__
  +
  +
Remember to add a [ [ Category:Code ] ] tag to any new sub-pages.
   
 
== MonadBase ==
 
== MonadBase ==
Line 5: Line 7:
 
It seems that the liftIO function from MonadIO can be generalized to access whatever the base of a transformer stack happens to be. So there is no need for a liftSTM, liftST, etc.
 
It seems that the liftIO function from MonadIO can be generalized to access whatever the base of a transformer stack happens to be. So there is no need for a liftSTM, liftST, etc.
   
  +
View [[New monads/MonadBase]].
<haskell>
 
-- By Chris Kuklewicz
 
module MonadBase(MonadBase,liftBase,MonadIO',liftIO') where
 
   
  +
== MonadLib ==
import Data.Monoid(Monoid)
 
import Control.Monad.Trans(lift,MonadTrans)
 
-- All the base monads with GHC
 
import Control.Monad.ST.Strict as S(ST)
 
import Control.Monad.ST.Lazy as L(ST)
 
import Control.Concurrent.STM(STM)
 
import Control.Monad.Identity(Identity)
 
import Text.ParserCombinators.Parsec(GenParser)
 
-- And all the MonadIO instances:
 
import Control.Monad.List(ListT)
 
import Control.Monad.Cont(ContT)
 
import Control.Monad.Error(ErrorT,Error)
 
import Control.Monad.Reader(ReaderT)
 
import Control.Monad.State(StateT)
 
import Control.Monad.Writer(WriterT)
 
import Control.Monad.RWS(RWST)
 
   
  +
[[MonadLib]] is written by Iavor S. Diatchki.
class (Monad m,Monad b) => MonadBase b m where
 
liftBase :: b a -> m a
 
   
  +
It is a new version of the mtl package with base monads: Id, and Lift, and transformers ReaderT, WriterT, StateT, ExceptionT, ChoiceT, and ContT.
-- One can recover MonadIO and liftIO from MonadBase
 
class (MonadBase IO m) => MonadIO' m where
 
liftIO' :: IO a -> m a
 
liftIO' = liftBase
 
   
  +
It also defines BaseM which is like MonadBase above.
-- Base monads
 
instance MonadBase IO IO where liftBase = id
 
instance MonadBase STM STM where liftBase = id
 
instance MonadBase (GenParser tok st) (GenParser tok st) where liftBase = id
 
instance MonadBase (S.ST s) (S.ST s) where liftBase = id
 
instance MonadBase (L.ST s) (L.ST s) where liftBase = id
 
instance MonadBase Maybe Maybe where liftBase = id
 
instance MonadBase [] [] where liftBase = id
 
instance (Error e) => MonadBase (Either e) (Either e) where liftBase = id
 
instance MonadBase ((->) a) ((->) a) where liftBase = id
 
instance MonadBase Identity Identity where liftBase = id
 
   
  +
== MonadRandom ==
-- Trans monads
 
instance MonadBase b m => MonadBase b (ListT m) where
 
liftBase = lift . liftBase
 
instance MonadBase b m => MonadBase b (ContT r m) where
 
liftBase = lift . liftBase
 
instance (Error e, MonadBase b m) => MonadBase b (ErrorT e m) where
 
liftBase = lift . liftBase
 
instance MonadBase b m => MonadBase b (ReaderT r m) where
 
liftBase = lift . liftBase
 
instance MonadBase b m => MonadBase b (StateT s m) where
 
liftBase = lift . liftBase
 
instance (Monoid w, MonadBase b m) => MonadBase b (WriterT w m) where
 
liftBase = lift . liftBase
 
instance (Monoid w, MonadBase b m) => MonadBase b (RWST r w s m) where
 
liftBase = lift . liftBase
 
</haskell>
 
   
  +
A simple monad transformer to allow computations in the transformed monad to generate random values.
And an artificial example:
 
   
  +
View [[New monads/MonadRandom]].
<haskell>
 
module Main where
 
   
  +
===MonadRandomSplittable===
import MonadBase
 
  +
A refinement of MonadRandom to integrate RandomGen's split function.
import Control.Monad.Reader
 
import Control.Monad.Writer
 
type Foo a = (WriterT [Int] (ReaderT String [])) a
 
   
  +
View at [[New monads/MonadRandomSplittable]]
foo :: Foo String
 
foo = do
 
x <- liftBase [1,2,3]
 
s <- ask
 
tell [succ x]
 
return (s ++ show x)
 
   
  +
== MaybeT ==
test = runReaderT (runWriterT foo) "hello"
 
</haskell>
 
   
  +
The Maybe monad deserves a transformer, just like the other classic monads.
<pre>
 
*Main> test
 
[("hello1",[2]),("hello2",[3]),("hello3",[4])]
 
</pre>
 
   
  +
View [[New monads/MaybeT]].
   
== MonadLib ==
+
== MonadSupply ==
  +
  +
Here is a simple monad/monad transformer for computations which consume values from a (finite or infinite) supply. Note that due to pattern matching, running out of supply in a non-MonadZero monad will cause an error.
  +
  +
View [[New monads/MonadSupply]].
  +
  +
== MonadUndo ==
  +
  +
Here is a modified state monad transformer for keeping track of undo/redo states automatically.
  +
  +
View [[New monads/MonadUndo]].
  +
  +
== MonadUnique ==
  +
  +
This is a simple (trivial) monad transformer for supplying unique integer values to an algorithm.
  +
  +
View [[New monads/MonadUnique]].
  +
  +
== MonadSTO ==
  +
  +
Here's an extension of the ST monad in which the references are ordered and showable (they list their creation index).
  +
  +
View [[New monads/MonadSTO]].
  +
  +
== MonadNondet ==
  +
  +
There is a [[Sudoku#Monadic_Non-Deterministic_Solver | MonadNondet]] that when compiled with optimizations outperforms List.
  +
  +
== Stateful nondeterminism ==
  +
  +
There is a [[Stateful nondeterminism]] monad for if you want to do nondeterministic computation with local states for each of your threads and a global state shared by all your threads.
  +
  +
== MonadAdvSTM ==
  +
  +
Here is an extension of STM to easy interaction with IO after committing or retrying. Inspired by Simon P-J.
  +
  +
View [[New monads/MonadAdvSTM]].
  +
  +
== TimedStateT ==
  +
  +
A monad transformer which combines State, Reader, and Error functionality to give the effect of a StateT monad which checks clock-time and stops the current computation if a period is exceeded.
  +
  +
darcs get http://www.mapcar.org/haskell/TimedStateT/
  +
  +
Haddocks: http://www.mapcar.org/haskell/TimedStateT/dist/doc/html/
  +
  +
== MonadExit ==
  +
  +
The Exit monad provides [[short-circuiting]] for complex program flow logic.
  +
  +
If you are using CPS (either explicitly, or in a CPS-based monad such as MonadCont or LogicT) only for this purpose, the Exit monad will likely simplify your program considerably.
  +
  +
'''Note:''' Now that a restriction on the Left type has been removed, the standard <hask>Either</hask> type can be used for this purpose. No separate Exit monad is needed anymore. For a monad transformer, use the version of EitherT defined in the [http://hackage.haskell.org/package/either either] package.
  +
  +
View [[New monads/MonadExit|MonadExit]].
  +
  +
== MonadSplit ==
  +
  +
Represents the class of monads such that
  +
  +
<haskell>l == (msplit l >>= \(x,xs) -> return x `mplus` xs)</haskell>
  +
  +
In English, msplit is a counterpart to "mplus".
  +
  +
Using this, you can redefine many of the functions which previously depended on lists: foldM, scanM, inits, tails, and some derived functions.
  +
  +
Note: A more general form of this monad,
  +
[http://haskell.org/ghc/docs/latest/html/libraries/base/Data-Foldable.html Data.Foldable], is now part of the
  +
[http://haskell.org/ghc/docs/latest/html/libraries/ standard libraries].
  +
  +
View [[New monads/MonadSplit]].
  +
  +
== Lazy and Strict variants ==
  +
  +
This section contains monads that have interesting Strict or Lazy properties.
  +
  +
=== LazyWriterT ===
  +
  +
This came up on the mailing list: Why is WriterT never lazy? The answer is it does not use lazy patterns with "~". So here is a more useful [[New monads/LazyWriterT]] that add two "~" to the definition of (>>=) and renames WriterT to LazyWriterT.
   
  +
=== Strict RWS ===
This is by Iavor S. Diatchki and can be found at http://www.cse.ogi.edu/~diatchki/monadLib/
 
   
  +
This was contribute by John Meacham on on the haskell-cafe mailing list. [[New monads/UnboxedRWS]] is an strict variant of RWS.
It a new version of the mtl package with transformers: ReaderT WriterT StateT ExceptT SearchT ContT
 
   
  +
[[Category:Idioms]]
It also defined BaseM which is like MonadBase above.
 
  +
[[Category:Monad]]
  +
[[Category:Proposals]]

Latest revision as of 09:43, 9 December 2014

Remember to add a [ [ Category:Code ] ] tag to any new sub-pages.

MonadBase

It seems that the liftIO function from MonadIO can be generalized to access whatever the base of a transformer stack happens to be. So there is no need for a liftSTM, liftST, etc.

View New monads/MonadBase.

MonadLib

MonadLib is written by Iavor S. Diatchki.

It is a new version of the mtl package with base monads: Id, and Lift, and transformers ReaderT, WriterT, StateT, ExceptionT, ChoiceT, and ContT.

It also defines BaseM which is like MonadBase above.

MonadRandom

A simple monad transformer to allow computations in the transformed monad to generate random values.

View New monads/MonadRandom.

MonadRandomSplittable

A refinement of MonadRandom to integrate RandomGen's split function.

View at New monads/MonadRandomSplittable

MaybeT

The Maybe monad deserves a transformer, just like the other classic monads.

View New monads/MaybeT.

MonadSupply

Here is a simple monad/monad transformer for computations which consume values from a (finite or infinite) supply. Note that due to pattern matching, running out of supply in a non-MonadZero monad will cause an error.

View New monads/MonadSupply.

MonadUndo

Here is a modified state monad transformer for keeping track of undo/redo states automatically.

View New monads/MonadUndo.

MonadUnique

This is a simple (trivial) monad transformer for supplying unique integer values to an algorithm.

View New monads/MonadUnique.

MonadSTO

Here's an extension of the ST monad in which the references are ordered and showable (they list their creation index).

View New monads/MonadSTO.

MonadNondet

There is a MonadNondet that when compiled with optimizations outperforms List.

Stateful nondeterminism

There is a Stateful nondeterminism monad for if you want to do nondeterministic computation with local states for each of your threads and a global state shared by all your threads.

MonadAdvSTM

Here is an extension of STM to easy interaction with IO after committing or retrying. Inspired by Simon P-J.

View New monads/MonadAdvSTM.

TimedStateT

A monad transformer which combines State, Reader, and Error functionality to give the effect of a StateT monad which checks clock-time and stops the current computation if a period is exceeded.

darcs get http://www.mapcar.org/haskell/TimedStateT/

Haddocks: http://www.mapcar.org/haskell/TimedStateT/dist/doc/html/

MonadExit

The Exit monad provides short-circuiting for complex program flow logic.

If you are using CPS (either explicitly, or in a CPS-based monad such as MonadCont or LogicT) only for this purpose, the Exit monad will likely simplify your program considerably.

Note: Now that a restriction on the Left type has been removed, the standard Either type can be used for this purpose. No separate Exit monad is needed anymore. For a monad transformer, use the version of EitherT defined in the either package.

View MonadExit.

MonadSplit

Represents the class of monads such that

l == (msplit l >>= \(x,xs) -> return x `mplus` xs)

In English, msplit is a counterpart to "mplus".

Using this, you can redefine many of the functions which previously depended on lists: foldM, scanM, inits, tails, and some derived functions.

Note: A more general form of this monad, Data.Foldable, is now part of the standard libraries.

View New monads/MonadSplit.

Lazy and Strict variants

This section contains monads that have interesting Strict or Lazy properties.

LazyWriterT

This came up on the mailing list: Why is WriterT never lazy? The answer is it does not use lazy patterns with "~". So here is a more useful New monads/LazyWriterT that add two "~" to the definition of (>>=) and renames WriterT to LazyWriterT.

Strict RWS

This was contribute by John Meacham on on the haskell-cafe mailing list. New monads/UnboxedRWS is an strict variant of RWS.