Personal tools

Do notation considered harmful

From HaskellWiki

(Difference between revisions)
Jump to: navigation, search
(monads vs. applicative functors)
(Criticism: Added a link)
(12 intermediate revisions by 4 users not shown)
Line 1: Line 1:
 
== Criticism ==
 
== Criticism ==
   
Haskell's [[do notation]] is popular and ubiquitous.
+
Haskell's [[Keywords#do | do notation]] is popular and ubiquitous.
 
However we shall not ignore that there are several problems.
 
However we shall not ignore that there are several problems.
 
Here we like to shed some light on aspects you may not have thought about, so far.
 
Here we like to shed some light on aspects you may not have thought about, so far.
Line 11: Line 11:
 
The downsides are
 
The downsides are
 
* that, since <hask>do</hask> notation is used almost everywhere, where <hask>IO</hask> takes place, newcomers quickly believe that the <hask>do</hask> notation is necessary for doing <hask>IO</hask>,
 
* that, since <hask>do</hask> notation is used almost everywhere, where <hask>IO</hask> takes place, newcomers quickly believe that the <hask>do</hask> notation is necessary for doing <hask>IO</hask>,
* and that newcomers think, that <hask>IO</hask> is somehow special and non-functional, in contrast to the advertisement for Haskell being purely functional.
+
* that newcomers think, that <hask>IO</hask> is somehow special and non-functional, in contrast to the advertisement for Haskell being purely functional,
  +
* and that newcomers think, that the order of statements determines the order of execution.
   
 
These misunderstandings let people write clumsy code like
 
These misunderstandings let people write clumsy code like
Line 41: Line 41:
 
</haskell>
 
</haskell>
 
.
 
.
  +
  +
The order of statements is also not the criterion for the evaluation order.
  +
Also here only the data dependencies count.
  +
See for instance
  +
<haskell>
  +
do x <- Just (3+5)
  +
y <- Just (5*7)
  +
return (x-y)
  +
</haskell>
  +
where <hask>3+5</hask> and <hask>5*7</hask> can be evaluated in any order, also in parallel.
  +
Or consider
  +
<haskell>
  +
do x <- Just (3+5)
  +
y <- Nothing
  +
return (x-y)
  +
</haskell>
  +
where <hask>3+5</hask> is probably not evaluated at all, because it's result is not necessary to find out,
  +
that the entire <hask>do</hask> describes a <hask>Nothing</hask>.
  +
   
 
=== Library design ===
 
=== Library design ===
   
 
Unfortunately, the <hask>do</hask> notation is so popular that people write more things with monads than necessary.
 
Unfortunately, the <hask>do</hask> notation is so popular that people write more things with monads than necessary.
Even more unfortunate, the [[applicative functor]]s were introduced to Haskell's standard libraries only after [[monad]]s and [[arrow]]s.
+
See for instance the [http://hackage.haskell.org/cgi-bin/hackage-scripts/package/binary-0.4.1 Binary] package.
<!--
+
It contains the <hask>Put</hask> monad, which has in principle [http://www.haskell.org/pipermail/haskell-cafe/2009-January/053317.html nothing to do with a monad].
They are more restrictive than [[monad]]s but it turned out,
+
All "put" operations have the monadic result <hask>()</hask>.
that many applications do not require monad functionality but only those of applicative functors.
+
In fact it is a <hask>Writer</hask> monad using the <hask>Builder</hask> type, and all you need is just the <hask>Builder</hask> monoid.
-->
+
Even more unfortunate,
  +
the [[applicative functor]]s were introduced to Haskell's standard libraries only after [[monad]]s and [[arrow]]s,
  +
thus many types are instances of <hask>Monad</hask> and <hask>Arrow</hask> classes,
  +
but not as much are instances of <hask>Applicative</hask>.
  +
There is no special syntax for applicative functors because it is hardly necessary.
  +
You just write
  +
<haskell>
  +
data Header = Header Char Int Bool
   
  +
readHeader :: Get Header
  +
readHeader = liftA3 Header get get get
  +
</haskell>
  +
or
  +
<haskell>
  +
readHeader = Header <$> get <*> get <*> get
  +
</haskell>
  +
  +
Not using monads and thus <hask>do</hask> notation can have advantages.
  +
Consider a generator of unique identifiers.
  +
First you might think of a <hask>State</hask> monad which increments a counter each time an identifier is requested.
  +
<haskell>
  +
run :: State Int a -> a
  +
run m = evalState m 0
  +
  +
newId :: State Int Int
  +
newId =
  +
do n <- get
  +
modify succ
  +
return n
  +
  +
example :: (Int -> Int -> a) -> a
  +
example f =
  +
run $
  +
do x <- newId
  +
y <- newId
  +
return (f x y)
  +
</haskell>
  +
  +
  +
If you are confident, that you will not need the counter state at the end and
  +
that you will not combine blocks of code using the counter
  +
(where the second block needs the state at the end of the first block),
  +
you can enforce a more strict scheme of usage.
  +
The following is like a <hask>Reader</hask> monad,
  +
where we call <hask>local</hask> on an incremented counter for each generated identifier.
  +
Alternatively you can view it as [[Continuation]] monad.
  +
  +
<haskell>
  +
newtype T a = T (Int -> a)
  +
  +
run :: T a -> a
  +
run (T f) = f 0
  +
  +
newId :: (Int -> T a) -> T a
  +
newId f = T $ \i -> case f i of T g -> g (succ i)
  +
  +
example :: (Int -> Int -> T a) -> a
  +
example f =
  +
run $
  +
newId $ \a ->
  +
newId $ \b ->
  +
f a b
  +
</haskell>
  +
  +
This way users cannot accidentally place a <hask>return</hask>
  +
somewhere in a <hask>do</hask> block where it has no effect.
   
   
 
=== Safety ===
 
=== Safety ===
  +
{{essay}}
   
 
With <hask>do</hask> notation we have kept alive a dark side of the C programming language:
 
With <hask>do</hask> notation we have kept alive a dark side of the C programming language:
Line 71: Line 148:
 
The same applies to
 
The same applies to
 
<haskell>
 
<haskell>
do System.cmd.system "echo foo >bar"
+
do System.Cmd.system "echo foo >bar"
 
</haskell>
 
</haskell>
 
where you ignore the <hask>ExitCode</hask>.
 
where you ignore the <hask>ExitCode</hask>.
Line 77: Line 154:
   
 
In safety oriented languages there are possibilities to explicitly ignore return values
 
In safety oriented languages there are possibilities to explicitly ignore return values
(e.g. <code>EVAL</code> in [http://www.m3.org/ Modula-3]).
+
(e.g. <code>EVAL</code> in [http://www.modula3.org/ Modula-3]).
 
Haskell does not need this, because you can already write
 
Haskell does not need this, because you can already write
 
<haskell>
 
<haskell>
do _ <- System.cmd.system "echo foo >bar"
+
do _ <- System.Cmd.system "echo foo >bar"
 
return ()
 
return ()
 
</haskell>
 
</haskell>
Line 90: Line 167:
 
</haskell>
 
</haskell>
 
This way, you can omit <hask> _ <- </hask> only if the monadic return value has type <hask>()</hask>.
 
This way, you can omit <hask> _ <- </hask> only if the monadic return value has type <hask>()</hask>.
  +
  +
New developments:
  +
* GHC since version 6.12 emits a warning when you silently ignore a return value
  +
* There is a new function called <hask>void</hask> that makes ignoring of return values explicit: [http://hackage.haskell.org/trac/ghc/ticket/3292 GHC ticket 3292]
  +
<!-- related is the problem on inefficient void (mapM f xs) vs. (mapM_ f xs) -->
  +
  +
== Happy with less sugar ==
  +
  +
=== Additional combinators ===
  +
  +
Using the infix combinators for writing functions simplifies the addition of new combinators.
  +
Consider for instance a monad for [http://darcs.haskell.org/probability/src/Numeric/Probability.hs random distributions].
  +
This monad cannot be an instance of <hask>MonadPlus</hask>,
  +
because there is no <hask>mzero</hask> (it would be an empty list of events, but their probabilities do not sum up to 1)
  +
and <hask>mplus</hask> is not associative because we have to normalize the sum of probabilities to 1.
  +
Thus we cannot use standard <hask>guard</hask> for this monad.
  +
However we would like to write the following:
  +
<haskell>
  +
do f <- family
  +
guard (existsBoy f)
  +
return f
  +
</haskell>
  +
  +
Given a custom combinator which performs a filtering with subsequent normalization called <hask>(>>=?) :: Distribution a -> (a -> Bool) -> Distribution a</hask>
  +
we can rewrite this easily:
  +
<haskell>
  +
family >>=? existsBoy
  +
</haskell>
  +
Note that the <hask>(>>=?)</hask> combinator introduces the risk of returning an invalid distribution (empty list of events),
  +
but it seems that we have to live with that problem.
  +
  +
=== Alternative combinators ===
  +
  +
If you are used to write monadic function using infix combinators <hask>(>>)</hask> and <hask>(>>=)</hask>
  +
you can easily switch to a different set of combinators.
  +
This is useful when there is a monadic structure that does not fit into the current <hask>Monad</hask> type constructor class,
  +
where the monadic result type cannot be constrained.
  +
This is e.g. useful for the [http://www.randomhacks.net/articles/2007/03/15/data-set-monad-haskell-macros Set data type],
  +
where the element type must have a total order.
  +
  +
== Useful applications ==
  +
  +
It shall be mentioned that the <hask>do</hask> sometimes takes the burden from you to write boring things.
  +
E.g. in
  +
<haskell>
  +
getRight :: Either a b -> Maybe b
  +
getRight y =
  +
do Right x <- y
  +
return x
  +
</haskell>
  +
a <hask>case</hask> on <hask>y</hask> is included,
  +
which calls <hask>fail</hask> if <hask>y</hask> is not a <hask>Right</hask> (i.e. <hask>Left</hask>),
  +
and thus returns <hask>Nothing</hask> in this case.
  +
  +
Also the <hask>mdo</hask> notation proves useful, since it maintains a set of variables for you in a safe manner.
  +
Compare
  +
<haskell>
  +
mdo x <- f x y z
  +
y <- g x y z
  +
z <- h x y z
  +
return (x+y+z)
  +
</haskell>
  +
and
  +
<haskell>
  +
mfix
  +
(\ ~( ~(x,y,z), _) ->
  +
do x <- f x y z
  +
y <- g x y z
  +
z <- h x y z
  +
return ((x,y,z),x+y+z))
  +
</haskell>
   
   

Revision as of 20:49, 21 March 2013

Contents

1 Criticism

Haskell's do notation is popular and ubiquitous. However we shall not ignore that there are several problems. Here we like to shed some light on aspects you may not have thought about, so far.

1.1 Didactics

The
do
notation hides functional details.

This is wanted in order to simplify writing imperative style code fragments. The downsides are

  • that, since
    do
    notation is used almost everywhere, where
    IO
    takes place, newcomers quickly believe that the
    do
    notation is necessary for doing
    IO
    ,
  • that newcomers think, that
    IO
    is somehow special and non-functional, in contrast to the advertisement for Haskell being purely functional,
  • and that newcomers think, that the order of statements determines the order of execution.

These misunderstandings let people write clumsy code like

do putStrLn "text"

instead of

putStrLn "text"

or

do text <- getLine
   return text

instead of

getLine

or

do
  text <- readFile "foo"
  writeFile "bar" text

instead of

readFile "foo" >>= writeFile "bar"

.

The order of statements is also not the criterion for the evaluation order. Also here only the data dependencies count. See for instance

do x <- Just (3+5)
   y <- Just (5*7)
   return (x-y)
where
3+5
and
5*7
can be evaluated in any order, also in parallel.

Or consider

do x <- Just (3+5)
   y <- Nothing
   return (x-y)
where
3+5
is probably not evaluated at all, because it's result is not necessary to find out, that the entire
do
describes a
Nothing
.


1.2 Library design

Unfortunately, the
do
notation is so popular that people write more things with monads than necessary.

See for instance the Binary package.

It contains the
Put
monad, which has in principle nothing to do with a monad. All "put" operations have the monadic result
()
. In fact it is a
Writer
monad using the
Builder
type, and all you need is just the
Builder
monoid.

Even more unfortunate, the applicative functors were introduced to Haskell's standard libraries only after monads and arrows,

thus many types are instances of
Monad
and
Arrow
classes, but not as much are instances of
Applicative
.

There is no special syntax for applicative functors because it is hardly necessary. You just write

data Header = Header Char Int Bool
 
readHeader :: Get Header
readHeader = liftA3 Header get get get

or

readHeader = Header <$> get <*> get <*> get
Not using monads and thus
do
notation can have advantages.

Consider a generator of unique identifiers.

First you might think of a
State
monad which increments a counter each time an identifier is requested.
run :: State Int a -> a
run m = evalState m 0
 
newId :: State Int Int
newId =
   do n <- get
      modify succ
      return n
 
example :: (Int -> Int -> a) -> a
example f =
   run $
      do x <- newId
         y <- newId
         return (f x y)


If you are confident, that you will not need the counter state at the end and that you will not combine blocks of code using the counter (where the second block needs the state at the end of the first block), you can enforce a more strict scheme of usage.

The following is like a
Reader
monad, where we call
local
on an incremented counter for each generated identifier.

Alternatively you can view it as Continuation monad.

newtype T a = T (Int -> a)
 
run :: T a -> a
run (T f) = f 0
 
newId :: (Int -> T a) -> T a
newId f = T $ \i -> case f i of T g -> g (succ i)
 
example :: (Int -> Int -> T a) -> a
example f =
   run $
   newId $ \a ->
   newId $ \b ->
   f a b
This way users cannot accidentally place a
return
somewhere in a
do
block where it has no effect.


1.3 Safety

This page addresses an aspect of Haskell style, which is to some extent a matter of taste. Just pick what you find appropriate for you and ignore the rest.
With
do
notation we have kept alive a dark side of the C programming language:

The silent neglect of return values of functions. In an imperative language it is common to return an error code and provide the real work by side effects. In Haskell this cannot happen, because functions have no side effects. If you ignore the result of a Haskell function the function will even not be evaluated.

The situation is different for
IO
: While processing the
IO
you might still ignore the contained return value.

You can write

do getLine
   putStrLn "text"
and thus silently ignore the result of
getLine
.

The same applies to

do System.Cmd.system "echo foo >bar"
where you ignore the
ExitCode
.

Is this behaviour wanted?

In safety oriented languages there are possibilities to explicitly ignore return values (e.g. EVAL in Modula-3). Haskell does not need this, because you can already write

do _ <- System.Cmd.system "echo foo >bar"
   return ()
Writing
 _ <-
should always make you cautious whether ignoring the result is the right thing to do. The possibility for silently ignoring monadic return values is not entirely the fault of the
do
notation. It would suffice to restrict the type of the
(>>)
combinator to
(>>) :: m () -> m a -> m a
This way, you can omit
 _ <-
only if the monadic return value has type
()
.

New developments:

  • GHC since version 6.12 emits a warning when you silently ignore a return value
  • There is a new function called
    void
    that makes ignoring of return values explicit: GHC ticket 3292

2 Happy with less sugar

2.1 Additional combinators

Using the infix combinators for writing functions simplifies the addition of new combinators. Consider for instance a monad for random distributions.

This monad cannot be an instance of
MonadPlus
, because there is no
mzero
(it would be an empty list of events, but their probabilities do not sum up to 1) and
mplus
is not associative because we have to normalize the sum of probabilities to 1. Thus we cannot use standard
guard
for this monad.

However we would like to write the following:

do f <- family
   guard (existsBoy f)
   return f
Given a custom combinator which performs a filtering with subsequent normalization called
(>>=?) :: Distribution a -> (a -> Bool) -> Distribution a

we can rewrite this easily:

family >>=? existsBoy
Note that the
(>>=?)
combinator introduces the risk of returning an invalid distribution (empty list of events),

but it seems that we have to live with that problem.

2.2 Alternative combinators

If you are used to write monadic function using infix combinators
(>>)
and
(>>=)

you can easily switch to a different set of combinators.

This is useful when there is a monadic structure that does not fit into the current
Monad
type constructor class,

where the monadic result type cannot be constrained. This is e.g. useful for the Set data type, where the element type must have a total order.

3 Useful applications

It shall be mentioned that the
do
sometimes takes the burden from you to write boring things.

E.g. in

getRight :: Either a b -> Maybe b
getRight y =
   do Right x <- y
      return x
a
case
on
y
is included, which calls
fail
if
y
is not a
Right
(i.e.
Left
), and thus returns
Nothing
in this case. Also the
mdo
notation proves useful, since it maintains a set of variables for you in a safe manner.

Compare

mdo x <- f x y z
    y <- g x y z
    z <- h x y z
    return (x+y+z)

and

mfix
   (\ ~( ~(x,y,z), _) ->
      do x <- f x y z
         y <- g x y z
         z <- h x y z
         return ((x,y,z),x+y+z))


4 See also