]> Git — Sourcephile - gargantext.git/blob - src/Data/Gargantext/Prelude.hs
app/Main & connectGargandb
[gargantext.git] / src / Data / Gargantext / Prelude.hs
1 {-# OPTIONS_GHC -fno-warn-name-shadowing #-}
2 {-# OPTIONS_GHC -fno-warn-type-defaults #-}
3 {-# LANGUAGE NoImplicitPrelude #-}
4
5 {-
6 TODO: import head impossible from Protolude: why ?
7 -}
8
9 module Data.Gargantext.Prelude
10 ( module Data.Gargantext.Prelude
11 , module Protolude
12 , headMay
13 )
14 where
15
16 import Protolude ( Bool(True, False), Int, Double, Integer
17 , Fractional, Num, Maybe, Floating, Char
18 , Ord, Integral, Foldable, RealFrac, Monad, filter
19 , reverse, map, zip, drop, take, zipWith
20 , sum, fromIntegral, length, fmap
21 , takeWhile, sqrt, undefined, identity
22 , abs, maximum, minimum, return, snd, truncate
23 , (+), (*), (/), (-), (.), (>=), ($), (**), (^)
24 )
25
26 -- TODO import functions optimized in Utils.Count
27 -- import Protolude hiding (head, last, all, any, sum, product, length)
28 -- import Data.Gargantext.Utils.Count
29
30 import qualified Data.List as L hiding (head, sum)
31 import qualified Control.Monad as M
32 import qualified Data.Map as Map
33 import qualified Data.Vector as V
34 import Safe (headMay)
35
36
37 pf :: (a -> Bool) -> [a] -> [a]
38 pf = filter
39
40 pr :: [a] -> [a]
41 pr = reverse
42
43 pm :: (a -> b) -> [a] -> [b]
44 pm = map
45
46 pm2 :: (t -> b) -> [[t]] -> [[b]]
47 pm2 fun = pm (pm fun)
48
49 pz :: [a] -> [b] -> [(a, b)]
50 pz = zip
51
52 pd :: Int -> [a] -> [a]
53 pd = drop
54
55 ptk :: Int -> [a] -> [a]
56 ptk = take
57
58 pzw :: (a -> b -> c) -> [a] -> [b] -> [c]
59 pzw = zipWith
60
61 -- Exponential Average
62 eavg :: [Double] -> Double
63 eavg (x:xs) = a*x + (1-a)*(eavg xs)
64 where a = 0.70
65 eavg [] = 0
66
67 -- Simple Average
68 mean :: Fractional a => [a] -> a
69 mean xs = if L.null xs then 0.0
70 else sum xs / fromIntegral (length xs)
71
72 sumMaybe :: Num a => [Maybe a] -> Maybe a
73 sumMaybe = fmap sum . M.sequence
74
75 variance :: Floating a => [a] -> a
76 variance xs = mean $ pm (\x -> (x - m) ** 2) xs where
77 m = mean xs
78
79 deviation :: [Double] -> Double
80 deviation = sqrt . variance
81
82 movingAverage :: Fractional b => Int -> [b] -> [b]
83 movingAverage steps xs = pm mean $ chunkAlong steps 1 xs
84
85 ma :: [Double] -> [Double]
86 ma = movingAverage 3
87
88
89 -- | Function to split a range into chunks
90 chunkAlong :: Int -> Int -> [a] -> [[a]]
91 chunkAlong a b l = only (while dropAlong)
92 where
93 only = pm (take a)
94 while = takeWhile (\x -> length x >= a)
95 dropAlong = L.scanl (\x _y -> drop b x) l ([1..] :: [Integer])
96
97 -- | Optimized version (Vector)
98 chunkAlong' :: Int -> Int -> V.Vector a -> V.Vector (V.Vector a)
99 chunkAlong' a b l = only (while dropAlong)
100 where
101 only = V.map (V.take a)
102 while = V.takeWhile (\x -> V.length x >= a)
103 dropAlong = V.scanl (\x _y -> V.drop b x) l (V.fromList [1..])
104
105 -- | TODO Inverse of chunk ? unchunkAlong ?
106 unchunkAlong :: Int -> Int -> [[a]] -> [a]
107 unchunkAlong = undefined
108
109
110 -- splitAlong [2,3,4] ("helloworld" :: [Char]) == ["he", "llo", "worl", "d"]
111 splitAlong :: [Int] -> [Char] -> [[Char]]
112 splitAlong _ [] = [] -- No list? done
113 splitAlong [] xs = [xs] -- No place to split at? Return the remainder
114 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
115
116 takeWhileM :: (Monad m) => (a -> Bool) -> [m a] -> m [a]
117 takeWhileM _ [] = return []
118 takeWhileM p (a:as) = do
119 v <- a
120 if p v
121 then do
122 vs <- takeWhileM p as
123 return (v:vs)
124 else return []
125
126 -- SUMS
127 -- To select the right algorithme according to the type:
128 -- https://github.com/mikeizbicki/ifcxt
129
130 sumSimple :: Num a => [a] -> a
131 sumSimple = L.foldl' (+) 0
132
133 -- | https://en.wikipedia.org/wiki/Kahan_summation_algorithm
134 sumKahan :: Num a => [a] -> a
135 sumKahan = snd . L.foldl' go (0,0)
136 where
137 go (c,t) i = ((t'-t)-y,t')
138 where
139 y = i-c
140 t' = t+y
141
142 -- | compute part of the dict
143 count2map :: (Ord k, Foldable t) => t k -> Map.Map k Double
144 count2map xs = Map.map (/ (fromIntegral (length xs))) (count2map' xs)
145
146 -- | insert in a dict
147 count2map' :: (Ord k, Foldable t) => t k -> Map.Map k Double
148 count2map' xs = L.foldl' (\x y -> Map.insertWith' (+) y 1 x) Map.empty xs
149
150
151 trunc :: (RealFrac a, Integral c, Integral b) => b -> a -> c
152 trunc n = truncate . (* 10^n)
153
154 trunc' :: Int -> Double -> Double
155 trunc' n x = fromIntegral $ truncate $ (x * 10^n)
156
157
158 bool2int :: Num a => Bool -> a
159 bool2int b = case b of
160 True -> 1
161 False -> 0
162
163 bool2double :: Bool -> Double
164 bool2double bool = case bool of
165 True -> 1.0
166 False -> 0.0
167
168
169
170 -- Normalizing && scaling data
171 scale :: [Double] -> [Double]
172 scale = scaleMinMax
173
174 scaleMinMax :: [Double] -> [Double]
175 scaleMinMax xs = pm (\x -> (x - mi / (ma - mi + 1) )) xs'
176 where
177 ma = maximum xs'
178 mi = minimum xs'
179 xs' = pm abs xs
180
181 scaleNormalize :: [Double] -> [Double]
182 scaleNormalize xs = pm (\x -> (x - v / (m + 1))) xs'
183 where
184 v = variance xs'
185 m = mean xs'
186 xs' = pm abs xs
187
188
189
190 normalize :: [Double] -> [Double]
191 normalize as = normalizeWith identity as
192
193 normalizeWith :: Fractional b => (a -> b) -> [a] -> [b]
194 normalizeWith extract bs = pm (\x -> x/(sum bs')) bs'
195 where
196 bs' = pm extract bs
197
198 -- Zip functions to add
199 zipFst :: ([b] -> [a]) -> [b] -> [(a, b)]
200 zipFst f xs = zip (f xs) xs
201
202 zipSnd :: ([a] -> [b]) -> [a] -> [(a, b)]
203 zipSnd f xs = zip xs (f xs)