An advanced, purely functional programming language

Declarative, statically typed code.

primes = filterPrime [2..]
  where filterPrime (p:xs) =
          p : filterPrime [x | x <- xs, x `mod` p /= 0]

Try it!


Statically typed

Every expression in Haskell has a type which is determined at compile time. All the types composed together by function application have to match up. If they don't, the program will be rejected by the compiler. Types become not only a form of guarantee, but a language for expressing the construction of programs.

Click to expand

All Haskell values have a type:

char = 'a'    :: Char
int = 123     :: Int
fun = isDigit :: Char -> Bool

You have to pass the right type of values to functions, or the compiler will reject the program:

Type error
isDigit 1

You can decode bytes into text:

bytes = Crypto.Hash.SHA1.hash "hello" :: ByteString
text = decodeUtf8 bytes               :: Text

But you cannot decode Text, which is already a vector of Unicode points:

Type error
doubleDecode = decodeUtf8 (decodeUtf8 bytes)

Purely functional

Every function in Haskell is a function in the mathematical sense (i.e., "pure"). Even side-effecting IO operations are but a description of what to do, produced by pure code. There are no statements or instructions, only expressions which cannot mutate variables (local or global) nor access state like time or random numbers.

Click to expand

The following function takes an integer and returns an integer. By the type it cannot do any side-effects whatsoever, it cannot mutate any of its arguments.

square :: Int -> Int
square x = x * x

The following string concatenation is okay:

"Hello: " ++ "World!" 

The following string concatenation is a type error:

Type error
"Name: " ++ getLine

Because getLine has type IO String and not String, like "Name: " is. So by the type system you cannot mix and match purity with impurity.

Type inference

You don't have to explicitly write out every type in a Haskell program. Types will be inferred by unifying every type bidirectionally. However, you can write out types if you choose, or ask the compiler to write them for you for handy documentation.

Click to expand

This example has a type signature for every binding:

main :: IO ()
main = do line :: String <- getLine
          print (parseDigit line)
  where parseDigit :: String -> Maybe Int
        parseDigit ((c :: Char) : _) =
          if isDigit c
             then Just (ord c - ord '0')
             else Nothing

But you can just write:

main = do line <- getLine
          print (parseDigit line)
  where parseDigit (c : _) =
          if isDigit c
             then Just (ord c - ord '0')
             else Nothing

You can also use inference to avoid wasting time explaining what you want:

do ss <- decode "[\"Hello!\",\"World!\"]"
   is <- decode "[1,2,3]"
   return (zipWith (\s i -> s ++ " " ++ show (i + 5)) ss is)
 => Just ["Hello! 6","World! 7"]

Types give a parser specification for free, the following input is not accepted:

do ss <- decode "[1,2,3]"
   is <- decode "[null,null,null]"
   return (zipWith (\s i -> s ++ " " ++ show (i + 5)) ss is)
 => Nothing


Haskell lends itself well to concurrent programming due to its explicit handling of effects. Its flagship compiler, GHC, comes with a high-performance parallel garbage collector and light-weight concurrency library containing a number of useful concurrency primitives and abstractions.

Click to expand

Easily launch threads and communicate with the standard library:

main = do
  done <- newEmptyMVar
  forkIO (do putStrLn "I'm one thread!"
             putMVar done "Done!")
  second <- forkIO (do threadDelay 100000
                       putStrLn "I'm another thread!")
  killThread second
  msg <- takeMVar done
  putStrLn msg

Use an asynchronous API for threads:

do a1 <- async (getURL url1)
  a2 <- async (getURL url2)
  page1 <- wait a1
  page2 <- wait a2

Atomic threading with software transactional memory:

transfer :: Account -> Account -> Int -> IO ()
transfer from to amount =
  atomically (do deposit to amount
                 withdraw from amount)

Atomic transactions must be repeatable, so arbitrary IO is disabled in the type system:

Type error
main = atomically (putStrLn "Hello!")


Functions don't evaluate their arguments. This means that programs can compose together very well, with the ability to write control constructs (such as if/else) just by writing normal functions. The purity of Haskell code makes it easy to fuse chains of functions together, allowing for performance benefits.

Click to expand

Define control structures easily:

when p m = if p then m else return ()
main = do args <- getArgs
          when (null args)
               (putStrLn "No args specified!") 

If you notice a repeated expression pattern, like

if c then t else False

you can give this a name, like

and c t = if c then t else False

and then use it with the same effect as the original expression.

Get code re-use by composing lazy functions. It's quite natural to express the any function by reusing the map and or functions:

any :: (a -> Bool) -> [a] -> Bool
any p = or . map p

Reuse the recursion patterns in map, filter, foldr, etc.


Open source contribution to Haskell is very active with a wide range of packages available on the public package servers.

Click to expand

There are 6,954 packages freely available. Here is a sample of the most common ones:

bytestring Binary data base Prelude, IO, threads
network Networking text Unicode text
parsec Parser library directory File/directory
hspec RSpec-like tests attoparsec Fast parser
monad-logger Logging persistent Database ORM
template-haskell Meta-programming tar Tar archives
snap Web framework time Date, time, etc.
happstack Web framework yesod Web framework
containers Maps, graphs, sets fsnotify Watch filesystem
hint Interpret Haskell unix UNIX bindings
SDL SDL binding OpenGL OpenGL graphics system
criterion Benchmarking pango Text rendering
cairo Cairo graphics statistics Statistical analysis
gtk Gtk+ library glib GLib library
test-framework Testing framework resource-pool Resource pooling
conduit Streaming I/O mwc-random High-quality randoms
QuickCheck Property testing stm Atomic threading
blaze-html Markup generation cereal Binary parsing/printing
xml XML parser/printer http-client HTTP client engine
zlib zlib/gzip/raw yaml YAML parser/printer
pandoc Markup conversion binary Serialization
tls TLS/SSL zip-archive Zip compression
warp Web server text-icu Text encodings
vector Vectors async Async concurrency
pipes Streaming IO scientific Arbitrary-prec. nums
process Launch processes aeson JSON parser/printer
dlist Difflists syb Generic prog.


DataDog provides powerful, customizable 24/7 metrics and monitoring integration for all of, and complains loudly for us when things go wrong.

Fastly's Next Generation CDN provides low latency access for all of's downloads and highest traffic services, including the primary Hackage server, Haskell Platform downloads, and more.

Equinix Metal provides compute, storage, and networking resources, powering almost all of in several regions around the world. powers, and lets us easily tell you when we broke something.

Galois provides infrastructure, funds, administrative resources and has historically hosted critical infrastructure, as well as helping the Haskell community at large with their work.

DreamHost has teamed up to provide with redundant, scalable object-storage through their Dream Objects service.

Hosted and managed by, a 501(c)(3) non-profit.

Psst! Looking for the wiki?

This is the new Haskell home page! The wiki has moved to