Rewrite and clean up Seminar module

This commit is contained in:
David Pätzel 2019-10-22 06:49:17 +02:00
parent a702bb6431
commit 59b5e4de8b

View File

@ -1,11 +1,12 @@
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE NoImplicitPrelude #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE TemplateHaskell #-}
module Seminar where
import Data.List ((!!), lookup, zipWith3)
import Data.List.Extra (nub)
import Data.List ((!!), (\\), lookup, zipWith3)
import Data.List.Extra (nubOrd, nubOrdOn)
import Data.Random
import Data.Random.Distribution.Uniform
import qualified Data.Text as T
@ -14,91 +15,91 @@ import Pretty
import Protolude
import Test.QuickCheck hiding (sample, shuffle)
import Test.QuickCheck.Instances
import Test.QuickCheck.Monadic (assert, monadicIO)
data Name
= N
{ firstName :: Text,
lastName :: Text
}
instance Pretty Text where
pretty = identity
type Student = Text
type Topic = Text
instance Pretty (Maybe Topic) where
pretty (Just t) = t
pretty Nothing = "Kein Thema"
newtype Priorities = P { unP :: [(Student, [(Topic, Int)])] }
deriving (Eq, Show)
instance Arbitrary Name where
arbitrary = N <$> arbitrary <*> arbitrary
instance Pretty Name where
pretty (N f l) = f <> " " <> l
type EMail = Text
data Student
= S
{ name :: Name,
email :: EMail
}
deriving (Eq, Show)
instance Arbitrary Student where
arbitrary = S <$> arbitrary <*> arbitrary
instance Pretty Student where
pretty (S n e) = pretty n <> " <" <> e <> ">"
instance Arbitrary Priorities where
arbitrary = priorities <$> Test.QuickCheck.arbitrary
{-|
The seminar GA is about giving topics to students. If there are not enough
topics, some students might get assigned 'NoTopic'.
Smart constructor for priority lists.
Completes a priority list, that is, if students did not assign priorities to
certain topics, adds these topics to their respective priority lists having a
priority of one less than the lowest priority assigned by them.
In addition, throws out all but the first occurence of each topic in a student's
list (i.e. removes assignments of multiple priorities to one topic for each
student) as well as all but the first occurrence of each student.
-}
data T
= T Text
| NoT
deriving (Show)
{-|
'Topic' is not 'Maybe' because this 'Eq' instance is different ('Nothing' @==@
'Nothing' but 'NoTopic' @/=@ 'NoTopic').
-}
instance Eq T where
T x == T y = x == y
-- NoT == _ = False
-- _ == NoT = False
NoT == NoT = True
NoT == _ = False
_ == NoT = False
instance Arbitrary T where
arbitrary = oneof [T <$> arbitrary, return NoT]
instance Pretty T where
pretty (T s) = s
pretty NoT = "Kein Thema"
topicToMaybe (T x) = Just x
topicToMaybe NoT = Nothing
type Priorities = [(Student, [(T, Int)])]
students :: Priorities -> [Student]
students = fmap fst
topics :: Priorities -> [T]
topics p = topics' ++ padding
priorities :: [(Student, [(Topic, Int)])] -> Priorities
priorities p = P . nubOrdOn fst $ second priorities1 <$> p
where
padding = replicate (length (students p) - length topics') NoT
topics' = nub . join $ fmap fst . snd <$> p
priorities1 :: [(Topic, Int)] -> [(Topic, Int)]
priorities1 [] =
topics p `zip` repeat 1
priorities1 ps =
let tLacking = topics p \\ (fst <$> ps) :: [Topic]
pWorst = maximum (snd <$> ps) + 1 :: Int
in nubOrdOn fst $ ps ++ (tLacking `zip` repeat pWorst)
topics = nubOrd . concatMap (fmap fst . snd)
prop_priorities_allListsSameLength :: [(Student, [(Topic, Int)])] -> Bool
prop_priorities_allListsSameLength p =
case unP . priorities $ p of
[] -> True
(s : ss) -> all (((length . snd) s ==) . length . snd) ss
{-|
The sum of all priorities given by all students.
The students that assigned priorities to topics.
-}
sumOfAll p = (1 +) . sum $ (sum . fmap snd) . snd <$> p
students :: Priorities -> [Student]
students = fmap fst . unP
{-|
The priority of 'NoTopic' is 101. The priority of topics that have not been
assigned a priority by a student is one less than this value.
-}
prioOf :: Priorities -> Student -> T -> Int
prioOf _ _ NoT = 101
prioOf p s t = maybe (prioOf p s NoT - 1) identity $ lookup s p >>= lookup t
The topics students assigned priorities to.
type Assignment = [(Student, T)]
Since 'Priorities' objects are well-formed due to the smart constructor, we can
simply return the topics the first student assigned priorities to.
-}
topics :: Priorities -> [Topic]
topics (P []) = []
topics (P (s : _)) = fmap fst . snd $ s
{-|
The priority value given by a student to a topic.
-}
prioOf :: Priorities -> Student -> Topic -> Int
prioOf p s t = fromMaybe (lowestPriority p + 1) $ lookup s (unP p) >>= lookup t
prop_prioOf_empty = prioOf (P []) "S" "T" == 1
prop_prioOf_singletonFound =
prioOf (P [("S", [("Existing topic", 10)])]) "S" "Existing topic" == 10
prop_prioOf_singletonNotFound =
prioOf (P [("S", [("Existing topic", 10)])]) "S" "Non-existing topic" == 11
{-|
The lowest priority assigned by a student to a topic.
-}
lowestPriority :: Priorities -> Int
lowestPriority = fromMaybe 0 . maximumMay . fmap snd . join . fmap snd . unP
type Assignment = [(Student, Maybe Topic)]
data I = I Priorities Assignment
deriving (Eq, Show)
@ -107,20 +108,31 @@ instance Pretty I where
pretty i@(I p a) =
T.unlines (gene <$> a)
where
gene :: (Student, T) -> Text
gene :: (Student, Maybe Topic) -> Text
gene (s, t) =
pretty s <> ": " <> pretty t <> prio s t
prio :: Student -> T -> Text
prio s t = " (" <> show (prioOf p s t) <> ")"
prio :: Student -> Maybe Topic -> Text
prio s t = " (" <> show (prioOf' p s t) <> ")"
{-|
The priority value given by a student to a topic including the case of her not
receiving a topic.
-}
prioOf' :: Priorities -> Student -> Maybe Topic -> Int
prioOf' p s Nothing = lowestPriority p + 2
prioOf' p s (Just t) = prioOf p s t
instance Individual I where
new (I p _) =
sample $ I p . zip (nub $ students p) <$> shuffle (topics p)
sample $ I p . zip (nubOrd $ students p) <$> shuffle topics'
where
topics' = (Just <$> topics p) ++ padding
padding = replicate (length (students p) - length (topics p)) Nothing
fitness (I p a) =
return . negate . sum
$ fromIntegral . uncurry (prioOf p) <$> a
$ fromIntegral . uncurry (prioOf' p) <$> a
mutate (I p a) = do
x <- sample $ Uniform 0 (length a - 1)
@ -132,15 +144,13 @@ instance Individual I where
does not create an invalid offspring).
TODO Assumes that both individuals are based on the same priorities.
TODO Require type-wise that both Assignments are of the same length.
-}
crossover1 (I p a1) (I _ a2) = do
let l = fromIntegral $ min (length a1) (length a2) :: Double
x <- sample $ Uniform 0 l
let a1' = zipWith3 (f x) a1 a2 [0 ..]
let a2' = zipWith3 (f x) a2 a1 [0 ..]
if valid a1' && valid a2'
if valid p a1' && valid p a2'
then return . Just $ (I p a1', I p a2')
else return Nothing
where
@ -163,54 +173,58 @@ switch i' j' xs
in left ++ [(fst ei, snd ej)] ++ middle ++ [(fst ej, snd ei)] ++ right
| otherwise = xs
prop_switch_keepsValid i j xs = valid xs == valid (switch i j xs)
{-|
Whether the given assignment is valid (every student occurs at most once, as
does every topic).
does every topic; also, there is only no topic given to students if there are
less topics then students).
Assumes that the priorities are well-formed.
-}
valid :: Assignment -> Bool
valid a = unique students && (unique . filter (/= NoT)) topics
valid :: Priorities -> Assignment -> Bool
valid p a =
sort (students p) == sort studentsAssigned
-- either there are less students than topics …
&& ( length (students p) < length (topics p)
-- … or every existing topic needs to be assigned
|| sort (topics p) == (sort . catMaybes) topicsAssigned
)
where
students = fmap fst a
topics = fmap snd a
studentsAssigned = fmap fst a
topicsAssigned = fmap snd a
{-|
Whether the given list only contains unique elements.
-}
unique xs = length xs == (length . nub) xs
prop_new_valid p = monadicIO $ do
I _ a <- lift $ new (I p [])
assert $ valid p a
prop_valid_empty = valid []
prop_mutate_valid p = monadicIO $ do
a <- lift . new $ I p []
I _ a <- lift $ mutate a
assert $ valid p a
prop_valid_dupNoT =
forAll noDupsList $ \ss ->
valid (ss `zip` repeat NoT)
prop_valid_dupT =
forAll noDupsList $ \ss ->
forAll noDupsList $ \ts' ->
let ts = filter (/= NoT) ts'
in length ss > length ts && not (null ts)
==> not . valid
$ ss `zip` cycle ts
prop_valid_noDups =
forAll noDupsList $ \ss ->
forAll noDupsList $ \ts ->
valid $ ss `zip` ts
prop_crossover1_valid p = monadicIO $ do
a1 <- lift . new $ I p []
a2 <- lift . new $ I p []
asM <- lift $ crossover1 a1 a2
assert
$ case asM of
Just (I _ a1', I _ a2') -> valid p a1' && valid p a2'
Nothing -> True
{-|
Generator for lists fulfilling 'unique', that is, only containing unique
elements.
-}
noDupsList :: (Arbitrary a, Eq a) => Gen [a]
noDupsList = nub <$> arbitrary
noDupsList :: (Arbitrary a, Eq a, Ord a) => Gen [a]
noDupsList = nubOrd <$> arbitrary
prop_noDupsList = forAll (noDupsList :: Gen [Int]) unique
-- NEXT find out why after x thousand repetitions, sometimes, bad solutions
-- occur with duplicate students
-- probably due to error in 'switch' implementation
{-|
Whether the given list only contains unique elements.
-}
unique :: (Ord a) => [a] -> Bool
unique xs = length xs == (length . nubOrd) xs
return []
runTests = $quickCheckAll