packages feed

liquidhaskell-0.8.10.7: docs/slides/NEU14/01_AbstractRefinements-blank.hs

{-@ LIQUID "--short-names"    @-}
{-@ LIQUID "--no-warnings"    @-}
{-@ LIQUID "--no-termination" @-}

module AbstractRefinements (
    listMax
  , insertSort
  , insertSort'
  ) where


import Data.Set hiding (insert, foldr,size) 
import Prelude hiding (map, foldr)



-----------------------------------------------------------------------
-- | 0. Abstract Refinements 
-----------------------------------------------------------------------


-- Warmup: How shall we type listMax?

{-@ listMax  :: forall <p :: Int -> Prop>. {v:[Int<p>] | len v > 0}
             -> Int<p> @-}
listMax     :: [Int] -> Int
listMax xs  = foldr1 max xs 



-- Lets define a few different subsets of Int

{-@ type Even = {v: Int | v mod 2 = 0} @-}
{-@ type Odd  = {v: Int | v mod 2 /= 0} @-}
{-@ type RGB  = {v: Int | 0 <= v && v <= 255}  @-}


-- compute the largest of some lists

{-@ xE :: Even @-}
xE = listMax [0, 1] 



{-@ xO :: Odd @-}
xO = listMax [1, 21, 4001, 961] 



{-@ xR :: RGB @-}
xR = listMax [1, 21, 41, 61] 


-- Why do we get the errors? How do we fix it?









-----------------------------------------------------------------------
-- | 1. Abstract Refinement from List's Type 
-----------------------------------------------------------------------

{-@ data List a <p :: a -> a -> Prop> =
        N
      | C { x :: a, xs :: List<p> (a<p x>) }
  @-}










-----------------------------------------------------------------------
-- | 2. Instantiating Abstract Refinements 
-----------------------------------------------------------------------


{-@ type IncrList a = List <{\x1 x2 -> x1 <= x2}> a@-} 
{-@ type DecrList a = List <{\x1 x2 -> x1 >= x2}> a@-} 
{-@ type DiffList a = List <{\x1 x2 -> x1 /= x2}> a@-} 

ups, downs  :: List Int

{-@ ups :: IncrList Int @-}
ups   = 1 `C` 2 `C` 3 `C` N

{-@ downs :: DecrList Int @-}
downs = 10 `C` 8 `C` 6 `C` N




-----------------------------------------------------------------------
-- | 3. Insertion Sort: Revisited
-----------------------------------------------------------------------

{-@ insert         :: _ -> xs:_ -> {v:_ | size v = 1 + size xs} @-}
insert x N         = x `C` N
insert x (C y ys)
  | x <= y         = x `C` y `C` ys
  | otherwise      = y `C` insert x ys 



{-@ insertSort      :: xs:List a -> {v:IncrList a | EqSize v xs} @-}
insertSort N        = N
insertSort (C x xs) = insert x (insertSort xs)











-----------------------------------------------------------------------
-- | 3. Insertion Sort: using a `foldr` 
-----------------------------------------------------------------------


{-@ insertSort' :: xs:List a -> {v:IncrList a | true } @-}
insertSort' xs = foldr insert N xs 














-----------------------------------------------------------------------
-- | 4. But, there are limits...
-----------------------------------------------------------------------

-- how big is the list returned by insertSort' ?


-- Hmm. Thats a bummer... How do we type `foldr` to verify the above?











-----------------------------------------------------------------------
-- | 5. Induction, as an Abstract Refinement 
-----------------------------------------------------------------------


ifoldr = undefined

{- insertSort' :: xs:List a -> {v:IncrList a | size v = size xs} -}







-----------------------------------------------------------------------
-- | 6. But can you prove that you've permuted the input?
-----------------------------------------------------------------------


-- Lets reason about the set of elements in a container

-- measure elems

{-@ predicate EqSize X Y = size X = size Y @-}

-- predicate EqElems


-----------------------------------------------------------------------
-- | Old definitions from 00_Refinements.hs
-----------------------------------------------------------------------

data List a = N | C a (List a)

infixr 9 `C`

{-@ measure size @-}
size          :: List a -> Int
size (C x xs) = 1 + size xs 
size N        = 0

foldr f acc N        = acc
foldr f acc (C x xs) = f x (foldr f acc xs)