commit 7f0b3a166a5865cfada2f03bfbc0c15079eef8ef
Author: Frederik Hanghøj Iversen <fhanghojiversen@zendesk.com>
Date:   Fri Nov 13 15:49:01 2020 +0100

    Hello world

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..5be0d00
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,4 @@
+*.hi
+*.o
+*.bin
+Main
\ No newline at end of file
diff --git a/Bloom.hs b/Bloom.hs
new file mode 100644
index 0000000..4586aaa
--- /dev/null
+++ b/Bloom.hs
@@ -0,0 +1,160 @@
+{-# language RoleAnnotations #-}
+{-# language GADTs #-}
+{-# language DataKinds #-}
+{-# language InstanceSigs #-}
+{-# language ScopedTypeVariables #-}
+{-# language StandaloneDeriving #-}
+{-# language DerivingStrategies #-}
+{-# language TypeApplications #-}
+{-# language TypeFamilies #-}
+{-# options_ghc -Wall -Werror #-}
+module Bloom (singleton, insert, Bloom, elem, assocs, hexdump) where
+
+import Prelude hiding (elem)
+import Data.Singletons
+import Data.Kind (Type)
+import Data.Array.Unboxed (UArray, listArray, (!))
+import qualified Data.Array.Unboxed as Array
+import GHC.Types (Nat)
+import GHC.TypeLits (KnownNat, natVal)
+import Data.Hashable (Hashable)
+import qualified Data.Hashable as Hashable
+import GHC.Exts (IsList(..))
+import Data.List (sort)
+import Data.Binary
+import Control.Monad (when)
+import Data.Bits
+import Data.Foldable (foldl')
+import Data.ByteString.Lazy (ByteString)
+import qualified Data.ByteString.Lazy as ByteString
+
+type role Bloom nominal nominal
+data Bloom :: Nat -> Type -> Type where
+  Bloom :: forall a n . Sing n -> UArray Int Bool -> Bloom n a
+
+deriving stock instance Show (Bloom n a)
+instance Eq (Bloom n a) where
+  Bloom _ a == Bloom _ b = a == b
+
+instance KnownNat n => Semigroup (Bloom n a) where
+  (<>) :: Bloom n a -> Bloom n a -> Bloom n a
+  Bloom _ a <> Bloom _ b = Bloom Sing c
+    where
+    c :: UArray Int Bool
+    c = listArray (0, pred n) $ f <$> [0..]
+    n :: Int
+    n = fromIntegral $ natVal @n Proxy
+    f :: Int -> Bool
+    f i = a ! i || b ! i
+
+instance KnownNat n => Monoid (Bloom n a) where
+  mempty :: Bloom n a
+  mempty = Bloom Sing c
+    where
+    c :: UArray Int Bool
+    c = listArray (0, pred n) $ repeat False
+    n :: Int
+    n = fromIntegral $ natVal @n Proxy
+
+instance KnownNat n => Binary (Bloom n a) where
+  -- This is not really a valid instance, because we don't know what
+  -- `n` and `a` was used to encode what we're `get`ing.
+  get = do
+    n <- get @Integer
+    let n' = natVal @n Proxy
+    when (n /= n') $ fail $ "Bloom.get: Wrong bucket size. Expected " <> show n' <> ". Got: " <> show n <> "."
+    Bloom Sing . listArray (0, fromInteger (pred n)) . unpack <$> get @ByteString
+  put b = do
+    put $ natVal @n Proxy
+    put @ByteString $ pack $ snd <$> assocs b
+
+singleton :: KnownNat n => Hashable a => a -> Bloom n a
+singleton a = insert a mempty
+
+insert :: forall n a . KnownNat n => Hashable a => a -> Bloom n a -> Bloom n a
+insert a (Bloom _ b) = Bloom Sing c
+  where
+  c :: UArray Int Bool
+  c = listArray (0, pred n) $ f <$> [0..n]
+  n :: Int
+  n = fromIntegral $ natVal @n Proxy
+  h :: Int
+  h = hash n a
+  f :: Int -> Bool
+  f i = b ! i || h == i
+
+hash :: Hashable a => Int -> a -> Int
+hash n = (`mod` n) . Hashable.hash
+
+instance (KnownNat n, Hashable a) => IsList (Bloom n a) where
+  type Item (Bloom n a) = a
+  fromList :: [] a -> Bloom n a
+  fromList = Bloom Sing . indicesToArray (pred n) . fmap (hash n)
+    where
+    n :: Int
+    n = fromIntegral $ natVal @n Proxy
+  toList = error "Unimplemented"
+
+indicesToArray :: Int -> [] Int -> UArray Int Bool
+indicesToArray n = listArray (0, n) . indicesToList
+
+indicesToList :: [] Int -> [] Bool
+indicesToList xs0 = take (succ $ maximum xs0) $ sort xs0 <| [0..]
+  where
+  (<|) :: [] Int -> [] Int -> [] Bool
+  [] <| _  = repeat False
+  xs@(x:xss) <| (y:yss) = if x <= y then True : (xss <| yss) else False : (xs<|yss)
+  _ <| _ = error "IMPOSSIBLE"
+
+elem :: forall a n . KnownNat n => Hashable a => a -> Bloom n a -> Bool
+elem a (Bloom _ r) = r ! hash n a
+  where
+  n = fromIntegral $ natVal @n Proxy
+
+assocs :: Bloom n a -> [(Int, Bool)]
+assocs (Bloom _ b) = Array.assocs b
+
+pack :: [] Bool -> ByteString
+pack = ByteString.pack . toWords
+
+unpack :: ByteString -> [] Bool
+unpack = (>>= fromWord) . ByteString.unpack
+
+toWords :: [] Bool -> [] Word8
+toWords = fmap toWord . chunksOf 8
+
+chunksOf :: Int -> [] a -> [[a]]
+chunksOf n xs = a : case b of [] -> [] ; _ -> chunksOf n b
+  where
+  a = take n xs
+  b = drop n xs
+
+toWord :: Bits b => [] Bool -> b
+toWord xs = foldl' (.|.) zeroBits $ zipWith shift (fmap conv xs) [0..]
+  where
+  conv b = if b then bit 0 else zeroBits
+
+fromWord :: Word8 -> [] Bool
+fromWord w = testBit w <$> [0..7]
+
+hexdump :: Bloom n a -> String
+hexdump = hexdump' . toWords . fmap snd . assocs
+
+hexdump' :: [] Word8 -> String
+hexdump' = unlines . zipWith (\n s -> step16 n <> s) [0..] . ls
+  where
+  ls :: [Word8] -> [String]
+  ls = fmap unwords . chunksOf 16 . fmap step8
+  step16 :: Word16 -> [] Char
+  step16 w0 = [r ! fromIntegral d0, r ! fromIntegral d1, r ! fromIntegral d2, r ! fromIntegral d3]
+    where
+    (d0, w1) = w0 `divMod` 0x10
+    (d1, w2) = w1 `divMod` 0x10
+    (d2, w3) = w2 `divMod` 0x10
+    (d3, _ ) = w3 `divMod` 0x10
+  step8 :: Word8 -> [] Char
+  step8 w = [r ! a, r ! b]
+    where
+    (a, b) = w `divMod` 0x10
+  r :: UArray Word8 Char
+  r = listArray (0, 15) (['0'..'9'] <> ['a'..'f'])
diff --git a/Main.hs b/Main.hs
new file mode 100644
index 0000000..5b97373
--- /dev/null
+++ b/Main.hs
@@ -0,0 +1,48 @@
+{-# language TypeApplications #-}
+{-# language DataKinds #-}
+{-# options_ghc -Wall -Werror #-}
+module Main (main) where
+
+import GHC.Exts (IsList(..))
+import Data.Foldable (traverse_)
+import System.Exit (ExitCode(ExitSuccess, ExitFailure), exitWith)
+import Bloom (Bloom)
+import qualified Bloom
+import qualified Data.Binary as Binary
+import qualified Data.ByteString.Lazy as ByteString
+import System.Directory
+import System.FilePath
+import System.Environment
+
+main :: IO ()
+main = do
+  b <- cachedLoad
+  input <- lines <$> getContents
+  let ws' = filter (not . (`Bloom.elem` b)) input
+  traverse_ putStrLn ws'
+  exitWith (if null ws' then ExitSuccess else ExitFailure 1)
+
+cachedLoad :: IO (Bloom Size String)
+cachedLoad = do
+  cache <- getCachePath
+  cacheExists <- doesFileExist cache
+  if cacheExists
+  then load cache
+  else do
+   b <- build
+   ByteString.writeFile cache $ Binary.encode b
+   pure b
+
+getCachePath :: IO FilePath
+getCachePath = (</> "bloom-spell/words.bin") <$> getEnv "XDG_DATA_HOME"
+
+load :: FilePath -> IO (Bloom Size String)
+load p = Binary.decode @(Bloom Size _) <$> ByteString.readFile p
+
+build :: IO (Bloom Size String)
+build = fromList @(Bloom Size _) . lines <$> readFile dict
+
+type Size = 0xfffff
+
+dict :: FilePath
+dict = "/usr/share/dict/words"
diff --git a/README b/README
new file mode 100644
index 0000000..d7cfb1f
--- /dev/null
+++ b/README
@@ -0,0 +1,3 @@
+A spell checker implemented using a bloom filter.
+
+My first time programming with (emulated) dependent types in Haskell.
\ No newline at end of file