# Blow your mind

### From HaskellWiki

(Add case) |
|||

(43 intermediate revisions by 23 users not shown) | |||

Line 6: | Line 6: | ||

− | == List/String Operations == |
+ | == List/String operations == |

Line 14: | Line 14: | ||

words |
words |
||

− | takeWhile (not . null) . unfoldr (Just . (second $ drop 1) . break (==' ')) |
+ | unfoldr (\b -> fmap (const . (second $ drop 1) . break (==' ') $ b) . listToMaybe $ b) |

+ | |||

+ | takeWhile (not . null) . evalState (repeatM $ modify (drop 1) >> State (break (== ' '))) . (' ' :) |
||

+ | where repeatM = sequence . repeat |
||

fix (\f l -> if null l then [] else let (s,e) = break (==' ') l in s:f (drop 1 e)) |
fix (\f l -> if null l then [] else let (s,e) = break (==' ') l in s:f (drop 1 e)) |
||

Line 21: | Line 21: | ||

-- splitting in two (alternating) |
-- splitting in two (alternating) |
||

-- "1234567" -> ("1357", "246") |
-- "1234567" -> ("1357", "246") |
||

− | foldr (\a (x,y) -> (a:y,x)) ([],[]) |
+ | -- the lazy match with ~ is necessary for efficiency, especially enabling processing of infinite lists |

+ | foldr (\a ~(x,y) -> (a:y,x)) ([],[]) |
||

(map snd *** map snd) . partition (even . fst) . zip [0..] |
(map snd *** map snd) . partition (even . fst) . zip [0..] |
||

− | transpose . unfoldr (\a -> if null a then Nothing else Just $ splitAt 2 a) |
+ | transpose . unfoldr (\a -> toMaybe (null a) (splitAt 2 a)) |

-- this one uses the solution to the next problem in a nice way :) |
-- this one uses the solution to the next problem in a nice way :) |
||

+ | toMaybe b x = if b then Just x else Nothing |
||

+ | -- or generalize it: |
||

+ | -- toMaybe = (toMonadPlus :: Bool -> a -> Maybe a) |
||

+ | toMonadPlus b x = guard b >> return x |
||

-- splitting into lists of length N |
-- splitting into lists of length N |
||

-- "1234567" -> ["12", "34", "56", "7"] |
-- "1234567" -> ["12", "34", "56", "7"] |
||

− | unfoldr (\a -> if null a then Nothing else Just $ splitAt 2 a) |
+ | unfoldr (\a -> toMaybe (not $ null a) (splitAt 2 a)) |

takeWhile (not . null) . unfoldr (Just . splitAt 2) |
takeWhile (not . null) . unfoldr (Just . splitAt 2) |
||

+ | |||

+ | ensure :: MonadPlus m => (a -> Bool) -> a -> m a |
||

+ | ensure p x = guard (p x) >> return x |
||

+ | unfoldr (ensure (not . null . fst) . splitAt 2) |
||

-- sorting by a custom function |
-- sorting by a custom function |
||

-- length -> ["abc", "ab", "a"] -> ["a", "ab", "abc"] |
-- length -> ["abc", "ab", "a"] -> ["a", "ab", "abc"] |
||

− | comparing f x y = compare (f x) (f y) |
+ | comparing f = compare `on` f -- "comparing" is already defined in Data.Ord |

sortBy (comparing length) |
sortBy (comparing length) |
||

Line 46: | Line 54: | ||

-- comparing adjacent elements |
-- comparing adjacent elements |
||

− | rises xs = zipWith (<) xs (drop 1 xs) |
+ | rises xs = zipWith (<) xs (tail xs) |

-- lazy substring search |
-- lazy substring search |
||

-- "ell" -> "hello" -> True |
-- "ell" -> "hello" -> True |
||

substr a b = any (a `isPrefixOf`) $ tails b |
substr a b = any (a `isPrefixOf`) $ tails b |
||

+ | |||

+ | -- multiple splitAt's: |
||

+ | -- splitAts [2,5,0,3] [1..15] == [[1,2],[3,4,5,6,7],[],[8,9,10],[11,12,13,14,15]] |
||

+ | splitAts = foldr (\n r -> splitAt n >>> second r >>> uncurry (:)) return |
||

+ | |||

+ | -- frequency distribution |
||

+ | -- "abracadabra" -> fromList [('a',5),('b',2),('c',1),('d',1),('r',2)] |
||

+ | import Data.Map |
||

+ | histogram = fromListWith (+) . (`zip` repeat 1) |
||

+ | |||

+ | -- using arrows and sort |
||

+ | histogramArr = map (head&&&length) . group . sort |
||

+ | |||

+ | -- multidimensional zipWith |
||

+ | zip2DWith :: (a -> b -> c) -> [[a]] -> [[b]] -> [[c]] |
||

+ | zip2DWith = zipWith . zipWith |
||

+ | zip3DWith :: (a -> b -> c) -> [[[a]]] -> [[[b]]] -> [[[c]]] |
||

+ | zip3DWith = zipWith . zipWith . zipWith |
||

+ | -- etc. |
||

</haskell> |
</haskell> |
||

− | == Mathematical Series, etc == |
+ | == Mathematical sequences, etc == |

Line 61: | Line 88: | ||

product [1..6] |
product [1..6] |
||

− | foldl1 (*) [1..6] |
+ | foldl1 (*) [1..6] -- this won't work for 0; use "foldl (*) 1 [1..n]" instead |

(!!6) $ scanl (*) 1 [1..] |
(!!6) $ scanl (*) 1 [1..] |
||

Line 68: | Line 95: | ||

− | -- powers of two series |
+ | -- powers of two sequence |

iterate (*2) 1 |
iterate (*2) 1 |
||

Line 74: | Line 101: | ||

− | -- fibonacci series |
+ | -- fibonacci sequence |

unfoldr (\(f1,f2) -> Just (f1,(f2,f1+f2))) (0,1) |
unfoldr (\(f1,f2) -> Just (f1,(f2,f1+f2))) (0,1) |
||

fibs = 0:1:zipWith (+) fibs (tail fibs) |
fibs = 0:1:zipWith (+) fibs (tail fibs) |
||

− | fib = 0:scanl (+) 1 fib |
+ | fib = 0:scanl (+) 1 fib -- also seen as: fibs = fix ((0:) . scanl (+) 1) |

Line 93: | Line 120: | ||

unfoldr sieve [2..] where |
unfoldr sieve [2..] where |
||

sieve (p:x) = Just(p, [ n | n <- x, n `mod` p > 0 ]) |
sieve (p:x) = Just(p, [ n | n <- x, n `mod` p > 0 ]) |
||

+ | |||

+ | otherPrimes = nubBy (((>1).).gcd) [2..] |
||

-- or if you want to use the Sieve of Eratosthenes.. |
-- or if you want to use the Sieve of Eratosthenes.. |
||

Line 108: | Line 137: | ||

rats = iterate next 1 where |
rats = iterate next 1 where |
||

next x = recip (fromInteger n+1-y) where (n,y) = properFraction x |
next x = recip (fromInteger n+1-y) where (n,y) = properFraction x |
||

+ | |||

+ | -- another way |
||

+ | rats2 = fix ((1:) . (>>= \x -> [1+x, 1/(1+x)])) :: [Rational] |
||

</haskell> |
</haskell> |
||

[1] [http://web.comlab.ox.ac.uk/oucl/work/jeremy.gibbons/publications/index.html#rationals Gibbons, Lest, Bird - Enumerating the Rationals] |
[1] [http://web.comlab.ox.ac.uk/oucl/work/jeremy.gibbons/publications/index.html#rationals Gibbons, Lest, Bird - Enumerating the Rationals] |
||

− | == Monad Magic == |
+ | == Monad magic == |

+ | The list monad can be used for some amazing Prolog-ish search problems. |
||

<haskell> |
<haskell> |
||

Line 128: | Line 161: | ||

"12" >>= \a -> "45" >>= \b -> return [a,b] |
"12" >>= \a -> "45" >>= \b -> return [a,b] |
||

− | |||

-- all combinations of letters |
-- all combinations of letters |
||

(inits . repeat) ['a'..'z'] >>= sequence |
(inits . repeat) ['a'..'z'] >>= sequence |
||

− | |||

-- apply a list of functions to an argument |
-- apply a list of functions to an argument |
||

-- even -> odd -> 4 -> [True,False] |
-- even -> odd -> 4 -> [True,False] |
||

Line 139: | Line 170: | ||

sequence [even,odd] 4 |
sequence [even,odd] 4 |
||

− | + | ||

+ | -- all subsequences of a sequence/ aka powerset. |
||

+ | filterM (const [True, False]) |
||

-- apply a function to two other function the same argument |
-- apply a function to two other function the same argument |
||

Line 148: | Line 179: | ||

liftM2 (>>) putStrLn return "hello" |
liftM2 (>>) putStrLn return "hello" |
||

+ | fix ((1:) . (>>= \x -> [x+1, 1/(x+1)])) :: [Rational] |
||

+ | [1%1,2%1,1%2,3%1,1%3,3%2,2%3,4%1,1%4,4%3,3%4,5%2,2%5,5%3,3%5,5%1,1%5,5%4,4%5... |
||

-- forward function concatenation |
-- forward function concatenation |
||

Line 162: | Line 195: | ||

-- [still to categorize] |
-- [still to categorize] |
||

− | (id >>= (+) >>= (+) >>= (+)) 3 -- (3+3)+(3+3) = 12 |
+ | (id >>= (+) >>= (+) >>= (+)) 3 -- 3+3+3+3 = 12 |

+ | -- Note: need to import Control.Monad.Instances |
||

− | (join . liftM2) (*) (+3) 5 -- 64 |
+ | |

+ | -- Galloping horsemen |
||

+ | -- A large circular track has only one place where horsemen can pass; |
||

+ | -- many can pass at once there. Is it possible for nine horsemen to |
||

+ | -- gallop around it continuously, all at different constant speeds? |
||

+ | -- the following prints out possible speeds for 2 or more horsemen. |
||

+ | spd s = ' ': show s ++ '/': show (s+1) |
||

+ | ext (c,l) = [(tails.filter(\b->a*(a+1)`mod`(b-a)==0)$r,a:l) | (a:r)<-c] |
||

+ | put = putStrLn . ('1':) . concatMap spd . reverse . snd . head |
||

+ | main = mapM_ put . iterate (>>= ext) $ [(map reverse $ inits [1..],[])] |
||

+ | |||

+ | -- output: |
||

+ | 1 1/2 |
||

+ | 1 2/3 1/2 |
||

+ | 1 3/4 2/3 1/2 |
||

+ | 1 5/6 4/5 3/4 2/3 |
||

+ | 1 12/13 11/12 10/11 9/10 8/9 |
||

+ | 1 27/28 26/27 25/26 24/25 23/24 20/21 |
||

+ | 1 63/64 60/61 59/60 57/58 56/57 55/56 54/55 |
||

+ | 1 755/756 741/742 740/741 735/736 734/735 728/729 727/728 720/721 |
||

+ | 1 126224/126225 122759/122760 122549/122550 122528/122529 122451/122452 |
||

+ | 122444/122445 122374/122375 122304/122305 122264/122265 |
||

+ | |||

+ | |||

+ | double = join (+) |
||

+ | |||

+ | (join . liftM2) (*) (+3) 5 -- (5+3)*(5+3) = 64 |
||

+ | -- Note: need to import Control.Monad.Instances |
||

mapAccumL (\acc n -> (acc+n,acc+n)) 0 [1..10] -- interesting for fac, fib, ... |
mapAccumL (\acc n -> (acc+n,acc+n)) 0 [1..10] -- interesting for fac, fib, ... |
||

Line 172: | Line 205: | ||

do { Just x <- [Nothing, Just 5, Nothing, Just 6, Just 7, Nothing]; return x } |
do { Just x <- [Nothing, Just 5, Nothing, Just 6, Just 7, Nothing]; return x } |
||

</haskell> |
</haskell> |
||

− | |||

== Other == |
== Other == |
||

Line 183: | Line 215: | ||

| otherwise -> True |
| otherwise -> True |
||

+ | --or: |
||

+ | cond = foldr (uncurry if') -- see [1] below |
||

-- match a constructor |
-- match a constructor |
||

Line 189: | Line 223: | ||

case a of Just{} -> True |
case a of Just{} -> True |
||

_ -> False |
_ -> False |
||

+ | |||

+ | |||

+ | -- spreadsheet magic |
||

+ | -- requires import Control.Monad.Instances |
||

+ | let loeb x = fmap ($ loeb x) x in |
||

+ | loeb [ (!!5), const 3, liftM2 (+) (!!0) (!!1), (*2) . (!!2), length, const 17] |
||

Line 204: | Line 244: | ||

-} |
-} |
||

</haskell> |
</haskell> |
||

+ | |||

+ | [1]: see [[Case]] and [[If-then-else]]. |
||

=== Polynomials === |
=== Polynomials === |
||

− | In abstract algebra you learn that polynomials can be used the same way integers are used given the right assumptions about their coefficients and roots. Specifically, polynomials support addition, subtraction, multiplication and sometimes division. It also turns out that one way to think of polynomials is that they are just lists of numbers (their coefficients). Here is one way to use lists to model polynomials. Since polynomials can support the same operations as integers, we model polynomials by making a list of numbers an instance of the Num type class. |
+ | In abstract algebra you learn that polynomials can be used the same way integers are used given the right assumptions about their coefficients and roots. Specifically, polynomials support addition, subtraction, multiplication and sometimes division. It also turns out that one way to think of polynomials is that they are just lists of numbers (their coefficients). |

− | <haskell> |
+ | instance Num a => Num [a] where -- (1) |

− | -- First we tell Haskell that we want to make lists (or [a]) an instance of Num. |
||

− | -- We refer to this instance of the Num type class as Num [a]. |
||

− | -- If you tried to use just: |
||

− | -- "instance Num [a] where" |
||

− | -- You'd get errors because the element type a is too general, too unconstrainted |
||

− | -- for what we need. |
||

− | -- So we add constraints to "a" by saying "Num a", this means whatever "a" is, it |
||

− | -- must be in the Num type class. |
||

− | instance Num a => Num [a] where |
+ | (f:fs) + (g:gs) = f+g : fs+gs -- (2) |

− | -- Next, we have to implement all the operations that instances of Num support. |
+ | fs + [] = fs -- (3a) |

− | -- A minimal set of operations is +, *, negate, abs, signum and fromInteger. |
+ | [] + gs = gs -- (3b) |

− | xs + ys = zipWith' (+) xs ys |
+ | |

− | where |
+ | (f:fs) * (g:gs) = f*g : [f]*gs + fs*(g:gs) -- (4) |

− | zipWith' f [] ys = ys |
+ | _ * _ = [] -- (5) |

− | zipWith' f xs [] = xs |
+ | |

− | zipWith' f (x:xs) (y:ys) = f x y : zipWith' f xs ys |
+ | abs = undefined -- I can't think of a sensible definition |

− | -- We define a new version of zipWith that returns a list as long as the longest |
+ | signum = map signum |

− | -- of the two lists it is given. If we did not do this then when we add polynomials |
+ | fromInteger n = [fromInteger n] |

− | -- the result would be truncated to the length of the shorter polynomial. |
+ | negate = map (\x -> -x) |

− | xs * ys = foldl1 (+) (padZeros partialProducts) |
+ | |

− | where |
+ | ====Explanation==== |

− | partialProducts = map (\x -> [x*y | y <- ys]) xs |
+ | (1) puts lists into type class Num, the class to which operators + and * belong, provided the list elements are in class Num. |

− | padZeros = map (\(z,zs) -> replicate z 0 ++ zs) . (zip [0..]) |
+ | |

− | -- This function is sort of hard to explain.... basically [1,2,3] should correspond |
+ | Lists are ordered by increasing powers. Thus <tt>f:fs</tt> means <tt>f+x*fs</tt> in algebraic notation. (2) and (4) follow from these algebraic identities: |

− | -- to the polynomial 1 + 2x + 3x^2. partialProducts does the steps of the multiplication |
+ | |

− | -- just like you would by hand when multiplying polynomials. |
+ | (f+x*fs) + (g+x*gs) = f+g + x*(fs+gs) |

− | -- padZeros takes a list of polynomials and creates tuples of the form |
+ | (f+x*fs) * (g+x*gs) = f*g + x*(f*gs + fs*(g+x*gs)) |

− | -- (offset, poly). If you notice when you add the partial products by hand |
+ | |

− | -- that you have to shift the partial products to the left on each new line. |
+ | (3) and (5) handle list ends. |

− | -- we accomplish this by padding by zeros at the beginning of the partial product. |
+ | |

− | -- Finally we use foldl1 to sum the partial products. Since they are polynomials |
+ | The bracketed <tt>[f]</tt> in (4) avoids mixed arithmetic, which Haskell doesn't support. |

− | -- They are added by the definition of plus we already gave. |
+ | |

− | negate xs = map negate xs |
+ | ====Comments==== |

− | abs xs = map abs xs -- is this reasonable? |
+ | |

− | signum xs = fromIntegral ((length xs)-1) |
+ | The methods are qualitatively different from ordinary array-based methods; there is no vestige of subscripting or counting of terms. |

− | -- signum isn't really defined for polynomials, but polynomials do have a concept |
+ | |

− | -- of degree. We might as well reuse signum as the degree of the |
+ | The methods are suitable for on-line computation. Only |

− | -- the polynomial. Notice that constants have degree zero. |
+ | <i>n</i> terms of each input must be seen before the <i>n</i>-th term |

− | fromInteger x = [fromInteger x] |
+ | of output is produced. |

− | -- This definition of fromInteger seems cyclical, it is left |
+ | |

− | -- as an exercise to the reader to figure out why it is correct :) |
+ | Thus the methods work on infinite series as well as polynomials. |

− | </haskell> |
+ | |

− | The reader is encouraged to write a simple pretty printer that takes into account the many special cases of displaying a polynomial. For example, [1,3,-2, 0, 1,-1,0] should display as: <tt>-x^5 + x^4 - 2x^2 + 3x + 1</tt> |
+ | Integer power comes for free. This example tests the cubing of (1+x): |

+ | |||

+ | [1, 1]^3 == [1, 3, 3, 1] |
||

+ | |||

+ | |||

+ | This gives us the infinite list of rows of Pascal's triangle: |
||

+ | |||

+ | pascal = map ([1,1]^) [0..] |
||

+ | |||

+ | For example, |
||

− | Other execrises for the reader include writing <haskell>polyApply :: (Num a) => [a] -> a -> a </haskell> which evaluates the polynomial at a specific value or writing a differentiation function. |
+ | take 5 pascal -- [[1], [1,1], [1,2,1], [1,3,3,1], [1,4,6,4,1]] |

− | See also [[Pointfree]]. |
+ | See also |

+ | * [[Pointfree]] |
||

+ | * [http://darcs.haskell.org/numericprelude/src/MathObj/Polynomial.hs NumericPrelude: Polynomials] |
||

+ | * [[Add polynomials]] |
||

+ | * Solve differential equations in terms of [http://www.haskell.org/pipermail/haskell-cafe/2004-May/006192.html power series]. |
||

[[Category:Idioms]] |
[[Category:Idioms]] |
||

+ | [[Category:Mathematics]] |

## Revision as of 01:24, 7 October 2011

Useful Idioms that will blow your mind (unless you already know them :)

This collection is supposed to be comprised of short, useful, cool, magical examples, which should incite the reader's curiosity and (hopefully) lead him to a deeper understanding of advanced Haskell concepts. At a later time I might add explanations to the more obscure solutions. I've also started providing several alternatives to give more insight into the interrelations of solutions.

More examples are always welcome, especially "obscure" monadic ones.

## Contents |

## 1 List/String operations

-- split at whitespace -- "hello world" -> ["hello","world"] words unfoldr (\b -> fmap (const . (second $ drop 1) . break (==' ') $ b) . listToMaybe $ b) takeWhile (not . null) . evalState (repeatM $ modify (drop 1) >> State (break (== ' '))) . (' ' :) where repeatM = sequence . repeat fix (\f l -> if null l then [] else let (s,e) = break (==' ') l in s:f (drop 1 e)) -- splitting in two (alternating) -- "1234567" -> ("1357", "246") -- the lazy match with ~ is necessary for efficiency, especially enabling processing of infinite lists foldr (\a ~(x,y) -> (a:y,x)) ([],[]) (map snd *** map snd) . partition (even . fst) . zip [0..] transpose . unfoldr (\a -> toMaybe (null a) (splitAt 2 a)) -- this one uses the solution to the next problem in a nice way :) toMaybe b x = if b then Just x else Nothing -- or generalize it: -- toMaybe = (toMonadPlus :: Bool -> a -> Maybe a) toMonadPlus b x = guard b >> return x -- splitting into lists of length N -- "1234567" -> ["12", "34", "56", "7"] unfoldr (\a -> toMaybe (not $ null a) (splitAt 2 a)) takeWhile (not . null) . unfoldr (Just . splitAt 2) ensure :: MonadPlus m => (a -> Bool) -> a -> m a ensure p x = guard (p x) >> return x unfoldr (ensure (not . null . fst) . splitAt 2) -- sorting by a custom function -- length -> ["abc", "ab", "a"] -> ["a", "ab", "abc"] comparing f = compare `on` f -- "comparing" is already defined in Data.Ord sortBy (comparing length) map snd . sortBy (comparing fst) . map (length &&& id) -- the so called "Schwartzian Transform" for computationally more expensive -- functions. -- comparing adjacent elements rises xs = zipWith (<) xs (tail xs) -- lazy substring search -- "ell" -> "hello" -> True substr a b = any (a `isPrefixOf`) $ tails b -- multiple splitAt's: -- splitAts [2,5,0,3] [1..15] == [[1,2],[3,4,5,6,7],[],[8,9,10],[11,12,13,14,15]] splitAts = foldr (\n r -> splitAt n >>> second r >>> uncurry (:)) return -- frequency distribution -- "abracadabra" -> fromList [('a',5),('b',2),('c',1),('d',1),('r',2)] import Data.Map histogram = fromListWith (+) . (`zip` repeat 1) -- using arrows and sort histogramArr = map (head&&&length) . group . sort -- multidimensional zipWith zip2DWith :: (a -> b -> c) -> [[a]] -> [[b]] -> [[c]] zip2DWith = zipWith . zipWith zip3DWith :: (a -> b -> c) -> [[[a]]] -> [[[b]]] -> [[[c]]] zip3DWith = zipWith . zipWith . zipWith -- etc.

## 2 Mathematical sequences, etc

-- factorial -- 6 -> 720 product [1..6] foldl1 (*) [1..6] -- this won't work for 0; use "foldl (*) 1 [1..n]" instead (!!6) $ scanl (*) 1 [1..] fix (\f n -> if n <= 0 then 1 else n * f (n-1)) -- powers of two sequence iterate (*2) 1 unfoldr (\z -> Just (z,2*z)) 1 -- fibonacci sequence unfoldr (\(f1,f2) -> Just (f1,(f2,f1+f2))) (0,1) fibs = 0:1:zipWith (+) fibs (tail fibs) fib = 0:scanl (+) 1 fib -- also seen as: fibs = fix ((0:) . scanl (+) 1) -- pascal triangle pascal = iterate (\row -> zipWith (+) ([0] ++ row) (row ++ [0])) [1] -- prime numbers -- example of a memoising caf (??) primes = sieve [2..] where sieve (p:x) = p : sieve [ n | n <- x, n `mod` p > 0 ] unfoldr sieve [2..] where sieve (p:x) = Just(p, [ n | n <- x, n `mod` p > 0 ]) otherPrimes = nubBy (((>1).).gcd) [2..] -- or if you want to use the Sieve of Eratosthenes.. diff [] l = l diff l [] = l diff xl@(x:xs) yl@(y:ys) | x < y = x:diff xs yl | x > y = diff xl ys | otherwise = diff xs ys esieve [] = [] esieve (p:ps) = p:esieve (diff ps (iterate (+p) p)) eprimes = esieve [2..] -- enumerating the rationals (see [1]) rats :: [Rational] rats = iterate next 1 where next x = recip (fromInteger n+1-y) where (n,y) = properFraction x -- another way rats2 = fix ((1:) . (>>= \x -> [1+x, 1/(1+x)])) :: [Rational]

[1] Gibbons, Lest, Bird - Enumerating the Rationals

## 3 Monad magic

The list monad can be used for some amazing Prolog-ish search problems.

-- all combinations of a list of lists. -- these solutions are all pretty much equivalent in that they run -- in the List Monad. the "sequence" solution has the advantage of -- scaling to N sublists. -- "12" -> "45" -> ["14", "15", "24", "25"] sequence ["12", "45"] [[x,y] | x <- "12", y <- "45"] do { x <- "12"; y <- "45"; return [x,y] } "12" >>= \a -> "45" >>= \b -> return [a,b] -- all combinations of letters (inits . repeat) ['a'..'z'] >>= sequence -- apply a list of functions to an argument -- even -> odd -> 4 -> [True,False] map ($4) [even,odd] sequence [even,odd] 4 -- all subsequences of a sequence/ aka powerset. filterM (const [True, False]) -- apply a function to two other function the same argument -- (lifting to the Function Monad (->)) -- even 4 && odd 4 -> False liftM2 (&&) even odd 4 liftM2 (>>) putStrLn return "hello" fix ((1:) . (>>= \x -> [x+1, 1/(x+1)])) :: [Rational] [1%1,2%1,1%2,3%1,1%3,3%2,2%3,4%1,1%4,4%3,3%4,5%2,2%5,5%3,3%5,5%1,1%5,5%4,4%5... -- forward function concatenation (*3) >>> (+1) $ 2 foldl1 (flip (.)) [(+1),(*2)] 500 -- perform functions in/on a monad, lifting fmap (+2) (Just 2) liftM2 (+) (Just 4) (Just 2) -- [still to categorize] (id >>= (+) >>= (+) >>= (+)) 3 -- 3+3+3+3 = 12 -- Note: need to import Control.Monad.Instances -- Galloping horsemen -- A large circular track has only one place where horsemen can pass; -- many can pass at once there. Is it possible for nine horsemen to -- gallop around it continuously, all at different constant speeds? -- the following prints out possible speeds for 2 or more horsemen. spd s = ' ': show s ++ '/': show (s+1) ext (c,l) = [(tails.filter(\b->a*(a+1)`mod`(b-a)==0)$r,a:l) | (a:r)<-c] put = putStrLn . ('1':) . concatMap spd . reverse . snd . head main = mapM_ put . iterate (>>= ext) $ [(map reverse $ inits [1..],[])] -- output: 1 1/2 1 2/3 1/2 1 3/4 2/3 1/2 1 5/6 4/5 3/4 2/3 1 12/13 11/12 10/11 9/10 8/9 1 27/28 26/27 25/26 24/25 23/24 20/21 1 63/64 60/61 59/60 57/58 56/57 55/56 54/55 1 755/756 741/742 740/741 735/736 734/735 728/729 727/728 720/721 1 126224/126225 122759/122760 122549/122550 122528/122529 122451/122452 122444/122445 122374/122375 122304/122305 122264/122265 double = join (+) (join . liftM2) (*) (+3) 5 -- (5+3)*(5+3) = 64 -- Note: need to import Control.Monad.Instances mapAccumL (\acc n -> (acc+n,acc+n)) 0 [1..10] -- interesting for fac, fib, ... do f <- [not, not]; d <- [True, False]; return (f d) -- [False,True,False,True] do { Just x <- [Nothing, Just 5, Nothing, Just 6, Just 7, Nothing]; return x }

## 4 Other

-- simulating lisp's cond case () of () | 1 > 2 -> True | 3 < 4 -> False | otherwise -> True --or: cond = foldr (uncurry if') -- see [1] below -- match a constructor -- this is better than applying all the arguments, because this way the -- data type can be changed without touching the code (ideally). case a of Just{} -> True _ -> False -- spreadsheet magic -- requires import Control.Monad.Instances let loeb x = fmap ($ loeb x) x in loeb [ (!!5), const 3, liftM2 (+) (!!0) (!!1), (*2) . (!!2), length, const 17] {- TODO, IDEAS: more fun with monad, monadPlus (liftM, ap, guard, when) fun with arrows (second, first, &&&, ***) liftM, ap lazy search (searching as traversal of lazy structures) innovative data types (i.e. having fun with Maybe sequencing) LINKS: bananas, envelopes, ... (generic traversal) why functional fp matters (lazy search, ...) -}

[1]: see Case and If-then-else.

### 4.1 Polynomials

In abstract algebra you learn that polynomials can be used the same way integers are used given the right assumptions about their coefficients and roots. Specifically, polynomials support addition, subtraction, multiplication and sometimes division. It also turns out that one way to think of polynomials is that they are just lists of numbers (their coefficients).

instance Num a => Num [a] where -- (1)

(f:fs) + (g:gs) = f+g : fs+gs -- (2) fs + [] = fs -- (3a) [] + gs = gs -- (3b)

(f:fs) * (g:gs) = f*g : [f]*gs + fs*(g:gs) -- (4) _ * _ = [] -- (5)

abs = undefined -- I can't think of a sensible definition signum = map signum fromInteger n = [fromInteger n] negate = map (\x -> -x)

#### 4.1.1 Explanation

(1) puts lists into type class Num, the class to which operators + and * belong, provided the list elements are in class Num.

Lists are ordered by increasing powers. Thus `f:fs` means `f+x*fs` in algebraic notation. (2) and (4) follow from these algebraic identities:

(f+x*fs) + (g+x*gs) = f+g + x*(fs+gs) (f+x*fs) * (g+x*gs) = f*g + x*(f*gs + fs*(g+x*gs))

(3) and (5) handle list ends.

The bracketed `[f]` in (4) avoids mixed arithmetic, which Haskell doesn't support.

#### 4.1.2 Comments

The methods are qualitatively different from ordinary array-based methods; there is no vestige of subscripting or counting of terms.

The methods are suitable for on-line computation. Only
*n* terms of each input must be seen before the *n*-th term
of output is produced.

Thus the methods work on infinite series as well as polynomials.

Integer power comes for free. This example tests the cubing of (1+x):

[1, 1]^3 == [1, 3, 3, 1]

This gives us the infinite list of rows of Pascal's triangle:

pascal = map ([1,1]^) [0..]

For example,

take 5 pascal -- [[1], [1,1], [1,2,1], [1,3,3,1], [1,4,6,4,1]]

See also

- Pointfree
- NumericPrelude: Polynomials
- Add polynomials
- Solve differential equations in terms of power series.