{-# LANGUAGE OverloadedLists #-}

module Clustering.FrequentItemSet.BruteForceSpec where

import Data.Either (fromRight)
import Data.Function (($))
import Data.Functor ((<$>))
import Data.Int (Int)
import Data.Ord (Ord)
import Data.Ratio ((%))
import Logic
import Logic.Theory.Arithmetic
import Logic.Theory.Ord
import Numeric.Probability (probability)
import Test.Syd
import Prelude (Num, undefined)

import Clustering.FrequentItemSet.BruteForce

exampleTakeakiUnoDB :: Ord item => Num item => Transactions item
exampleTakeakiUnoDB =
  [ [1, 2, 5, 6, 7]
  , [2, 3, 4, 5]
  , [1, 2, 7, 8, 9]
  , [1, 7, 9]
  , [2, 7, 9]
  , [2, 7, 9] -- Copy-paste typo on the original example
  , [1, 9] -- Add this to increase the support of [1,9] because the original example is wrong…
  , [2]
  ]

spec :: Spec
spec = do
  describe "frequentItemSets" do
    it "computes Takeaki Uno's example" $
      -- From https://research.nii.ac.jp/~uno/code/lcm.html#IntroductionstoFrequentItemsetMining
      letName 3 \(\x -> x / fromRight undefined (prove (x > zero)) -> minSupp) ->
        -- Alas, this is not a zero-cost `coerce`
        unName <$> allFrequentItemSets @Int (nameless exampleTakeakiUnoDB) minSupp
          `shouldBe` [ []
                     , [1]
                     , [1, 7]
                     , [1, 9]
                     , [2]
                     , [2, 7]
                     , [2, 7, 9]
                     , [2, 9]
                     , [7]
                     , [7, 9]
                     , [9]
                     ]
  describe "associationRules" do
    proba75 <- liftIO (probability (75 % 100))
    it "computes Takeaki Uno's example" do
      let db = nameless exampleTakeakiUnoDB
      let minSupp = let x = nameless 2 in x / fromRight undefined (prove (x > zero))
      goldenPrettyShowInstance
        "tests/Clustering/FrequentItemSet/BruteForce/associationRules/TakeakiUno.golden"
        [ associationRules fis db (nameless proba75)
        | fis <- allFrequentItemSets @Int db minSupp
        ]