Angelos.hs

{-# LANGUAGE OverloadedStrings #-}

import           Control.Applicative
import           Text.HTML.Scalpel

import           Data.Char
import           Data.List
import           Data.Ord


wikiUrl = "https://en.wikipedia.org/wiki/Haskell_(programming_language)?action=render"
paragraphs :: IO (Maybe [String])
paragraphs = scrapeURL wikiUrl $ chroots anySelector $ text "p" <|> text "h3"


-- Remember that haskell is a lazy language and infinite structures
-- are common. `take (length x - 1) x` will die for `x = [1..]` as it
-- will get stuck calculating `length [1..]`.
shortenBy1 :: String -> String
shortenBy1 []     = []
shortenBy1 [_]    = []
shortenBy1 (x:xs) = x:shortenBy1 xs

removePunc :: String -> String
removePunc = filter $ not . isPunctuation


-- A style that isn't better or worse (called point free) is to
-- compose functions rather than appy functions to the arguments. The
-- two are equivalent (and often only the latter is possible, eg if
-- the arguments are used in more than one place) but point-free
-- trains you in a certain way of thinking that haskell is designed
-- for. Namely that functions are not special in any way to the
-- language. Also at least sometimes you avoid having to things.
--
-- Finally you can make your intent clearer by making your types more
-- general.
countWords :: Ord a => [a] -> [(a,Int)]
countWords = map (\w -> (head w, length w)) . group . sort
histogram :: Ord a => [a] -> [(a,Int)]
histogram = sortBy (comparing snd) . countWords
tokenize :: String -> [String]
tokenize = filter (not . null) . words . map toLower . removePunc
main = do
  paragsM <- paragraphs
  case paragsM of
    Just parags -> do
      let wordHisto = histogram $ tokenize =<< parags
      print $ head wordHisto
    Nothing -> fail "Couldn't parse site.."