-----------------------------------------------------------------------------
-- |
-- Module    : Data.SBV.Tools.Polynomial
-- Copyright : (c) Levent Erkok
-- License   : BSD3
-- Maintainer: erkokl@gmail.com
-- Stability : experimental
--
-- Implementation of polynomial arithmetic
-----------------------------------------------------------------------------

{-# LANGUAGE DataKinds            #-}
{-# LANGUAGE FlexibleContexts     #-}
{-# LANGUAGE FlexibleInstances    #-}
{-# LANGUAGE PatternGuards        #-}
{-# LANGUAGE TypeFamilies         #-}
{-# LANGUAGE UndecidableInstances #-}

{-# OPTIONS_GHC -Wall -Werror #-}

module Data.SBV.Tools.Polynomial (
        -- * Polynomial arithmetic and CRCs
        Polynomial(..), crc, crcBV, ites, mdp, addPoly
        ) where

import Data.Bits  (Bits(..))
import Data.List  (genericTake)
import Data.Maybe (fromJust, fromMaybe)
import Data.Word  (Word8, Word16, Word32, Word64)

import Data.SBV.Core.Data
import Data.SBV.Core.Sized
import Data.SBV.Core.Model

import GHC.TypeLits

-- | Implements polynomial addition, multiplication, division, and modulus operations
-- over GF(2^n).  NB. Similar to 'sQuotRem', division by @0@ is interpreted as follows:
--
--     @x `pDivMod` 0 = (0, x)@
--
-- for all @x@ (including @0@)
--
-- Minimal complete definition: 'pMult', 'pDivMod', 'showPolynomial'
class (Num a, Bits a) => Polynomial a where
 -- | Given bit-positions to be set, create a polynomial
 -- For instance
 --
 --     @polynomial [0, 1, 3] :: SWord8@
 -- 
 -- will evaluate to @11@, since it sets the bits @0@, @1@, and @3@. Mathematicans would write this polynomial
 -- as @x^3 + x + 1@. And in fact, 'showPoly' will show it like that.
 polynomial :: [Int] -> a
 -- | Add two polynomials in GF(2^n).
 pAdd  :: a -> a -> a
 -- | Multiply two polynomials in GF(2^n), and reduce it by the irreducible specified by
 -- the polynomial as specified by coefficients of the third argument. Note that the third
 -- argument is specifically left in this form as it is usally in GF(2^(n+1)), which is not available in our
 -- formalism. (That is, we would need SWord9 for SWord8 multiplication, etc.) Also note that we do not
 -- support symbolic irreducibles, which is a minor shortcoming. (Most GF's will come with fixed irreducibles,
 -- so this should not be a problem in practice.)
 --
 -- Passing [] for the third argument will multiply the polynomials and then ignore the higher bits that won't
 -- fit into the resulting size.
 pMult :: (a, a, [Int]) -> a
 -- | Divide two polynomials in GF(2^n), see above note for division by 0.
 pDiv  :: a -> a -> a
 -- | Compute modulus of two polynomials in GF(2^n), see above note for modulus by 0.
 pMod  :: a -> a -> a
 -- | Division and modulus packed together.
 pDivMod :: a -> a -> (a, a)
 -- | Display a polynomial like a mathematician would (over the monomial @x@), with a type.
 showPoly :: a -> String
 -- | Display a polynomial like a mathematician would (over the monomial @x@), the first argument
 -- controls if the final type is shown as well.
 showPolynomial :: Bool -> a -> String

 {-# MINIMAL pMult, pDivMod, showPolynomial #-}
 polynomial = foldr (flip setBit) 0
 pAdd       = xor
 pDiv x y   = fst (pDivMod x y)
 pMod x y   = snd (pDivMod x y)
 showPoly   = showPolynomial False

instance Polynomial Word8   where {showPolynomial   = sp;           pMult = lift polyMult; pDivMod = liftC polyDivMod}
instance Polynomial Word16  where {showPolynomial   = sp;           pMult = lift polyMult; pDivMod = liftC polyDivMod}
instance Polynomial Word32  where {showPolynomial   = sp;           pMult = lift polyMult; pDivMod = liftC polyDivMod}
instance Polynomial Word64  where {showPolynomial   = sp;           pMult = lift polyMult; pDivMod = liftC polyDivMod}
instance Polynomial SWord8  where {showPolynomial b = liftS (sp b); pMult = polyMult;      pDivMod = polyDivMod}
instance Polynomial SWord16 where {showPolynomial b = liftS (sp b); pMult = polyMult;      pDivMod = polyDivMod}
instance Polynomial SWord32 where {showPolynomial b = liftS (sp b); pMult = polyMult;      pDivMod = polyDivMod}
instance Polynomial SWord64 where {showPolynomial b = liftS (sp b); pMult = polyMult;      pDivMod = polyDivMod}

instance (KnownNat n, IsNonZero n) => Polynomial (SWord n) where {showPolynomial b = liftS (sp b); pMult = polyMult;      pDivMod = polyDivMod}

lift :: SymVal a => ((SBV a, SBV a, [Int]) -> SBV a) -> (a, a, [Int]) -> a
lift f (x, y, z) = fromJust $ unliteral $ f (literal x, literal y, z)
liftC :: SymVal a => (SBV a -> SBV a -> (SBV a, SBV a)) -> a -> a -> (a, a)
liftC f x y = let (a, b) = f (literal x) (literal y) in (fromJust (unliteral a), fromJust (unliteral b))
liftS :: SymVal a => (a -> String) -> SBV a -> String
liftS f s
  | Just x <- unliteral s = f x
  | True                  = show s

-- | Pretty print as a polynomial
sp :: Bits a => Bool -> a -> String
sp st a
 | null cs = '0' : t
 | True    = foldr (\x y -> sh x ++ " + " ++ y) (sh (last cs)) (init cs) ++ t
 where t | st   = " :: GF(2^" ++ show n ++ ")"
         | True = ""
       n  = fromMaybe (error "SBV.Polynomial.sp: Unexpected non-finite usage!") (bitSizeMaybe a)
       is = [n-1, n-2 .. 0]
       cs = map fst $ filter snd $ zip is (map (testBit a) is)
       sh 0 = "1"
       sh 1 = "x"
       sh i = "x^" ++ show i

-- | Add two polynomials
addPoly :: [SBool] -> [SBool] -> [SBool]
addPoly xs    []      = xs
addPoly []    ys      = ys
addPoly (x:xs) (y:ys) = x .<+> y : addPoly xs ys

-- | Run down a boolean condition over two lists. Note that this is
-- different than zipWith as shorter list is assumed to be filled with
-- sFalse at the end (i.e., zero-bits); which nicely pads it when
-- considered as an unsigned number in little-endian form.
ites :: SBool -> [SBool] -> [SBool] -> [SBool]
ites s xs ys
 | Just t <- unliteral s
 = if t then xs else ys
 | True
 = go xs ys
 where go []     []     = []
       go []     (b:bs) = ite s sFalse b : go [] bs
       go (a:as) []     = ite s a sFalse : go as []
       go (a:as) (b:bs) = ite s a b : go as bs

-- | Multiply two polynomials and reduce by the third (concrete) irreducible, given by its coefficients.
-- See the remarks for the 'pMult' function for this design choice
polyMult :: SFiniteBits a => (SBV a, SBV a, [Int]) -> SBV a
polyMult (x, y, red)
  | isReal x
  = error $ "SBV.polyMult: Received a real value: " ++ show x
  | not (isBounded x)
  = error $ "SBV.polyMult: Received infinite precision value: " ++ show x
  | True
  = fromBitsLE $ genericTake sz $ r ++ repeat sFalse
  where (_, r) = mdp ms rs
        ms = genericTake (2*sz) $ mul (blastLE x) (blastLE y) [] ++ repeat sFalse
        rs = genericTake (2*sz) $ [fromBool (i `elem` red) |  i <- [0 .. foldr max 0 red] ] ++ repeat sFalse
        sz = intSizeOf x
        mul _  []     ps = ps
        mul as (b:bs) ps = mul (sFalse:as) bs (ites b (as `addPoly` ps) ps)

polyDivMod :: SFiniteBits a => SBV a -> SBV a -> (SBV a, SBV a)
polyDivMod x y
   | isReal x
   = error $ "SBV.polyDivMod: Received a real value: " ++ show x
   | not (isBounded x)
   = error $ "SBV.polyDivMod: Received infinite precision value: " ++ show x
   | True
   = ite (y .== 0) (0, x) (adjust d, adjust r)
   where adjust xs = fromBitsLE $ genericTake sz $ xs ++ repeat sFalse
         sz        = intSizeOf x
         (d, r)    = mdp (blastLE x) (blastLE y)

-- conservative over-approximation of the degree
degree :: [SBool] -> Int
degree xs = walk (length xs - 1) $ reverse xs
  where walk n []     = n
        walk n (b:bs)
         | Just t <- unliteral b
         = if t then n else walk (n-1) bs
         | True
         = n -- over-estimate

-- | Compute modulus/remainder of polynomials on bit-vectors.
mdp :: [SBool] -> [SBool] -> ([SBool], [SBool])
mdp xs ys = go (length ys - 1) (reverse ys)
  where degTop  = degree xs
        go _ []     = error "SBV.Polynomial.mdp: Impossible happened; exhausted ys before hitting 0"
        go n (b:bs)
         | n == 0   = (reverse qs, rs)
         | True     = let (rqs, rrs) = go (n-1) bs
                      in (ites b (reverse qs) rqs, ites b rs rrs)
         where degQuot = degTop - n
               ys' = replicate degQuot sFalse ++ ys
               (qs, rs) = divx (degQuot+1) degTop xs ys'

-- return the element at index i; if not enough elements, return sFalse
-- N.B. equivalent to '(xs ++ repeat sFalse) !! i', but more efficient
idx :: [SBool] -> Int -> SBool
idx []     _ = sFalse
idx (x:_)  0 = x
idx (_:xs) i = idx xs (i-1)

divx :: Int -> Int -> [SBool] -> [SBool] -> ([SBool], [SBool])
divx n _ xs _ | n <= 0 = ([], xs)
divx n i xs ys'        = (q:qs, rs)
  where q        = xs `idx` i
        xs'      = ites q (xs `addPoly` ys') xs
        (qs, rs) = divx (n-1) (i-1) xs' (tail ys')

-- | Compute CRCs over bit-vectors. The call @crcBV n m p@ computes
-- the CRC of the message @m@ with respect to polynomial @p@. The
-- inputs are assumed to be blasted big-endian. The number
-- @n@ specifies how many bits of CRC is needed. Note that @n@
-- is actually the degree of the polynomial @p@, and thus it seems
-- redundant to pass it in. However, in a typical proof context,
-- the polynomial can be symbolic, so we cannot compute the degree
-- easily. While this can be worked-around by generating code that
-- accounts for all possible degrees, the resulting code would
-- be unnecessarily big and complicated, and much harder to reason
-- with. (Also note that a CRC is just the remainder from the
-- polynomial division, but this routine is much faster in practice.)
--
-- NB. The @n@th bit of the polynomial @p@ /must/ be set for the CRC
-- to be computed correctly. Note that the polynomial argument @p@ will
-- not even have this bit present most of the time, as it will typically
-- contain bits @0@ through @n-1@ as usual in the CRC literature. The higher
-- order @n@th bit is simply assumed to be set, as it does not make
-- sense to use a polynomial of a lesser degree. This is usually not a problem
-- since CRC polynomials are designed and expressed this way.
--
-- NB. The literature on CRC's has many variants on how CRC's are computed.
-- We follow the following simple procedure:
--
--     * Extend the message @m@ by adding @n@ 0 bits on the right
--
--     * Divide the polynomial thus obtained by the @p@
--
--     * The remainder is the CRC value.
--
-- There are many variants on final XOR's, reversed polynomials etc., so
-- it is essential to double check you use the correct /algorithm/.
crcBV :: Int -> [SBool] -> [SBool] -> [SBool]
crcBV n m p = take n $ go (replicate n sFalse) (m ++ replicate n sFalse)
  where mask = drop (length p - n) p
        go c []     = c
        go c (b:bs) = go next bs
          where c' = drop 1 c ++ [b]
                next = ite (head c) (zipWith (.<+>) c' mask) c'

-- | Compute CRC's over polynomials, i.e., symbolic words. The first
-- 'Int' argument plays the same role as the one in the 'crcBV' function.
crc :: (SFiniteBits a, SFiniteBits b) => Int -> SBV a -> SBV b -> SBV b
crc n m p
  | isReal m || isReal p
  = error $ "SBV.crc: Received a real value: " ++ show (m, p)
  | not (isBounded m) || not (isBounded p)
  = error $ "SBV.crc: Received an infinite precision value: " ++ show (m, p)
  | True
  = fromBitsBE $ replicate (sz - n) sFalse ++ crcBV n (blastBE m) (blastBE p)
  where sz = intSizeOf p