src/Gargantext/Prelude.hs

   1 {-|
   2 Module      : Gargantext.Prelude
   3 Description : Specific Prelude of the project
   4 Copyright   : (c) CNRS, 2017-Present
   5 License     : AGPL + CECILL v3
   6 Maintainer  : team@gargantext.org
   7 Stability   : experimental
   8 Portability : POSIX
   9
  10 Here is a longer description of this module, containing some
  11 commentary with @some markup@.
  12 -}
  13
  14 {-# OPTIONS_GHC -fno-warn-name-shadowing #-}
  15 {-# OPTIONS_GHC -fno-warn-type-defaults  #-}
  16
  17 {-# LANGUAGE     NoImplicitPrelude       #-}
  18 {-# LANGUAGE     OverloadedStrings       #-}
  19
  20 module Gargantext.Prelude
  21   ( module Gargantext.Prelude
  22   , module Protolude
  23   , headMay, lastMay
  24   , module GHC.Err.Located
  25   , module Text.Show
  26   , module Text.Read
  27   , cs
  28   , module Data.Maybe
  29   , sortWith
  30   )
  31   where
  32
  33 import GHC.Exts (sortWith)
  34 import GHC.Err.Located (undefined)
  35 import Control.Monad.IO.Class (MonadIO)
  36 import Data.Maybe (isJust, fromJust, maybe)
  37 import Protolude ( Bool(True, False), Int, Int64, Double, Integer
  38                  , Fractional, Num, Maybe(Just,Nothing)
  39                  , Enum, Bounded, Float
  40                  , Floating, Char, IO
  41                  , pure, (>>=), (=<<), (<*>), (<$>)
  42                  , putStrLn
  43                  , head, flip
  44                  , Ord, Integral, Foldable, RealFrac, Monad, filter
  45                  , reverse, map, mapM, zip, drop, take, zipWith
  46                  , sum, fromIntegral, length, fmap, foldl, foldl'
  47                  , takeWhile, sqrt, identity
  48                  , abs, min, max, maximum, minimum, return, snd, truncate
  49                  , (+), (*), (/), (-), (.), ($), (&), (**), (^), (<), (>), log
  50                  , Eq, (==), (>=), (<=), (<>), (/=)
  51                  , (&&), (||), not, any, all
  52                  , fst, snd, toS
  53                  , elem, die, mod, div, const, either
  54                  , curry, uncurry, repeat
  55                  , otherwise, when
  56                  , IO()
  57                  , compare
  58                  , on
  59                  , panic
  60                  )
  61
  62 -- TODO import functions optimized in Utils.Count
  63 -- import Protolude hiding (head, last, all, any, sum, product, length)
  64 -- import Gargantext.Utils.Count
  65 import qualified Data.List     as L hiding (head, sum)
  66 import qualified Control.Monad as M
  67
  68 import Data.Map (Map)
  69 import qualified Data.Map as M
  70
  71 import Data.Map.Strict (insertWith)
  72 import qualified Data.Vector as V
  73 import Safe (headMay, lastMay)
  74 import Text.Show (Show(), show)
  75 import Text.Read (Read())
  76 import Data.String.Conversions (cs)
  77
  78
  79 printDebug :: (Show a, MonadIO m) => [Char] -> a -> m ()
  80 printDebug msg x = putStrLn $ msg <> " " <> show x
  81 -- printDebug _ _ = pure ()
  82
  83
  84 map2 :: (t -> b) -> [[t]] -> [[b]]
  85 map2 fun = map (map fun)
  86
  87
  88 -- Some Statistics sugar functions
  89 -- Exponential Average
  90 eavg :: [Double] -> Double
  91 eavg (x:xs) = a*x + (1-a)*(eavg xs)
  92   where a = 0.70
  93 eavg [] = 0
  94
  95 -- Simple Average
  96 mean :: Fractional a => [a] -> a
  97 mean xs = if L.null xs then 0.0
  98                        else sum xs / fromIntegral (length xs)
  99
 100
 101 sumMaybe :: Num a => [Maybe a] -> Maybe a
 102 sumMaybe = fmap sum . M.sequence
 103
 104 variance :: Floating a => [a] -> a
 105 variance xs = mean $ map (\x -> (x - m) ** 2) xs where
 106     m = mean xs
 107
 108 deviation :: [Double] -> Double
 109 deviation = sqrt . variance
 110
 111 movingAverage :: (Eq b, Fractional b) => Int -> [b] -> [b]
 112 movingAverage steps xs = map mean $ chunkAlong steps 1 xs
 113
 114 ma :: [Double] -> [Double]
 115 ma = movingAverage 3
 116
 117 -- | splitEvery n == chunkAlong n n
 118 splitEvery :: Int -> [a] -> [[a]]
 119 splitEvery _ [] = []
 120 splitEvery n xs =
 121   let (h,t) = L.splitAt n xs
 122   in  h : splitEvery n t
 123
 124 type Grain = Int
 125 type Step  = Int
 126
 127 -- | Function to split a range into chunks
 128 -- if   step == grain then linearity
 129 -- elif step < grain then overlapping
 130 -- else dotted with holes
 131 chunkAlong :: Eq a => Grain -> Step -> [a] -> [[a]]
 132 chunkAlong a b l = case a > 0 && b > 0 of
 133     True  -> chunkAlong_ a b l
 134     False -> panic "ChunkAlong: Parameters should be > 0 and Grain > Step"
 135
 136 chunkAlong_ :: Eq a => Int -> Int -> [a] -> [[a]]
 137 chunkAlong_ a b l = filter (/= []) $ only (while dropAlong)
 138     where
 139         only      = map       (take a)
 140         while     = takeWhile (\x -> length x >= a)
 141         dropAlong = L.scanl   (\x _y -> drop b x) l ([1..] :: [Integer])
 142
 143 -- | Optimized version (Vector)
 144 chunkAlong' :: Int -> Int -> V.Vector a -> V.Vector (V.Vector a)
 145 chunkAlong' a b l = only (while  dropAlong)
 146     where
 147         only      = V.map       (V.take a)
 148         while     = V.takeWhile (\x -> V.length x >= a)
 149         dropAlong = V.scanl     (\x _y -> V.drop b x) l (V.fromList [1..])
 150
 151 -- | TODO Inverse of chunk ? unchunkAlong ?
 152 -- unchunkAlong :: Int -> Int -> [[a]] -> [a]
 153 -- unchunkAlong = undefined
 154
 155
 156 -- splitAlong [2,3,4] ("helloworld" :: [Char]) == ["he", "llo", "worl", "d"]
 157 splitAlong :: [Int] -> [Char] -> [[Char]]
 158 splitAlong _ [] = [] -- No list? done
 159 splitAlong [] xs = [xs] -- No place to split at? Return the remainder
 160 splitAlong (x:xs) ys = take x ys : splitAlong xs (drop x ys) -- take until our split spot, recurse with next split spot and list remainder
 161
 162 takeWhileM :: (Monad m) => (a -> Bool) -> [m a] -> m [a]
 163 takeWhileM _ [] = return []
 164 takeWhileM p (a:as) = do
 165     v <- a
 166     if p v
 167         then do
 168             vs <- takeWhileM p as
 169             return (v:vs)
 170         else return []
 171
 172 -- SUMS
 173 -- To select the right algorithme according to the type:
 174 -- https://github.com/mikeizbicki/ifcxt
 175
 176 sumSimple :: Num a => [a] -> a
 177 sumSimple = L.foldl' (+) 0
 178
 179 -- | https://en.wikipedia.org/wiki/Kahan_summation_algorithm
 180 sumKahan :: Num a => [a] -> a
 181 sumKahan = snd . L.foldl' go (0,0)
 182     where
 183         go (c,t) i = ((t'-t)-y,t')
 184             where
 185                 y  = i-c
 186                 t' = t+y
 187
 188 -- | compute part of the dict
 189 count2map :: (Ord k, Foldable t) => t k -> Map k Double
 190 count2map xs = M.map (/ (fromIntegral (length xs))) (count2map' xs)
 191
 192 -- | insert in a dict
 193 count2map' :: (Ord k, Foldable t) => t k -> Map k Double
 194 count2map' xs = L.foldl' (\x y -> insertWith (+) y 1 x) M.empty xs
 195
 196
 197 trunc :: (RealFrac a, Integral c, Integral b) => b -> a -> c
 198 trunc n = truncate . (* 10^n)
 199
 200 trunc' :: Int -> Double -> Double
 201 trunc' n x = fromIntegral $ truncate $ (x * 10^n)
 202
 203
 204 ------------------------------------------------------------------------
 205 bool2num :: Num a => Bool -> a
 206 bool2num True  = 1
 207 bool2num False = 0
 208
 209 bool2double :: Bool -> Double
 210 bool2double = bool2num
 211
 212 bool2int :: Bool -> Int
 213 bool2int = bool2num
 214 ------------------------------------------------------------------------
 215
 216 -- Normalizing && scaling data
 217 scale :: [Double] -> [Double]
 218 scale = scaleMinMax
 219
 220 scaleMinMax :: [Double] -> [Double]
 221 scaleMinMax xs = map (\x -> (x - mi / (ma - mi + 1) )) xs'
 222     where
 223         ma  = maximum xs'
 224         mi  = minimum xs'
 225         xs' = map abs xs
 226
 227 scaleNormalize :: [Double] -> [Double]
 228 scaleNormalize xs = map (\x -> (x - v / (m + 1))) xs'
 229     where
 230         v = variance xs'
 231         m = mean     xs'
 232         xs' = map abs xs
 233
 234 normalize :: [Double] -> [Double]
 235 normalize as = normalizeWith identity as
 236
 237 normalizeWith :: Fractional b => (a -> b) -> [a] -> [b]
 238 normalizeWith extract bs = map (\x -> x/(sum bs')) bs'
 239     where
 240         bs' = map extract bs
 241
 242 -- Zip functions to add
 243 zipFst :: ([b] -> [a]) -> [b] -> [(a, b)]
 244 zipFst  f xs = zip (f xs) xs
 245
 246 zipSnd :: ([a] -> [b]) -> [a] -> [(a, b)]
 247 zipSnd f xs = zip xs (f xs)
 248
 249 -- | maximumWith
 250 maximumWith :: (Ord a1, Foldable t) => (a2 -> a1) -> t a2 -> a2
 251 maximumWith f = L.maximumBy (compare `on` f)
 252