6

在它的来源中Data/ByteString.hs说，该功能findSubstrings已被弃用，取而代之的是breakSubstring. 但是我认为findSubstrings使用 KMP 算法实现的算法比使用的算法更有效breakSubstring。有人知道为什么会这样做吗？

这是旧的实现：

{-# DEPRECATED findSubstrings "findSubstrings is deprecated in favour of breakSubstring." #-}

{-
{- This function uses the Knuth-Morris-Pratt string matching algorithm.  -}

findSubstrings pat@(PS _ _ m) str@(PS _ _ n) = search 0 0
where
  patc x = pat `unsafeIndex` x
  strc x = str `unsafeIndex` x

  -- maybe we should make kmpNext a UArray before using it in search?
  kmpNext = listArray (0,m) (-1:kmpNextL pat (-1))
  kmpNextL p _ | null p = []
  kmpNextL p j = let j' = next (unsafeHead p) j + 1
                     ps = unsafeTail p
                     x = if not (null ps) && unsafeHead ps == patc j'
                            then kmpNext Array.! j' else j'
                    in x:kmpNextL ps j'
  search i j = match ++ rest -- i: position in string, j: position in pattern
    where match = if j == m then [(i - j)] else []
          rest = if i == n then [] else search (i+1) (next (strc i) j + 1)
  next c j | j >= 0 && (j == m || c /= patc j) = next c (kmpNext Array.! j)
           | otherwise = j
-}

这是新的天真的：

findSubstrings :: ByteString -- ^ String to search for.
           -> ByteString -- ^ String to seach in.
           -> [Int]
findSubstrings pat str
    | null pat         = [0 .. length str]
    | otherwise        = search 0 str
where
    STRICT2(search)
    search n s
        | null s             = []
        | pat `isPrefixOf` s = n : search (n+1) (unsafeTail s)
        | otherwise          =     search (n+1) (unsafeTail s)
4

1 回答 1

4

改变的原因是 KMP 的实现实际上比 naive 版本效率低，后者是着眼于性能来实现的。

于 2011-06-13T16:37:09.373 回答