合并排序比插入排序更快的方式让我困惑。

Question

我的脚被Haskell的排序算法弄湿了。我实现了插入排序和合并排序

insert_sort :: (Ord a, Show a) => [a] -> [a]
insert_sort keys = foldr f [] keys
           where f key []        = [key]
                 f key acc       = insert key acc
                 insert y []     = [y]
                 insert y (x:xs)
                     | x < y     = x : insert y xs
                     | otherwise = y : x : xs

merge_sort :: (Ord a, Show a) => [a] -> [a]
merge_sort (x:[]) = [x]
merge_sort keys   = merge  (merge_sort (take len keys)) (merge_sort (drop len keys))
      where len         = length keys `div` 2
            merge :: [a] -> [a] -> [a]
            merge (x:xs) []     = (x:xs)
            merge []     (y:ys) = (y:ys)
            merge (x:xs) (y:ys) = if x <= y
                                  then x : merge (xs) (y:ys)
                                  else y : merge (x:xs) ys

下面是我比较它们的效率的方法：

insert_sort $ take 100000 $ randomRs (1,100000) $ mkStdGen 1 ::[Int]
merge_sort $ take 100000 $ randomRs (1,100000) $ mkStdGen 1 ::[Int]

两种方法都在延迟很短时间后开始打印结果，但是合并排序打印的速度要快得多。正如我们所知道的，合并排序比大数据集的插入排序快得多.我认为，这可以通过他们如何给出结果(比如长延迟和短延迟)来显示，而不是如何打印结果。是因为我在插入排序中使用了foldr吗？现场后面是什么？

编辑：Thx家伙。自从我开始学习Haskell以来，我就听说过懒惰评估，但是我已经掌握了它的诀窍。有人会用一个小数据集来说明更多吗，比如5,2,6,3,1,4？既然第一个元素是最后出现的，那么在用foldr完成排序之前，如何能够输出结果？

Answer 1

场景背后是懒惰的评价。排序列表的开始是在排序完成之前确定的，因此可以在工作完成之前输出。由于合并更快，合并排序列表打印得更快。

应请求:如何对[5,2,6,3,1,4]进行排序。为了简洁起见，我使用insert_sort = foldr ins []。

insert_sort [5,2,6,3,1,4]
  = foldr ins [] [5,2,6,3,1,4]
  = 5 `ins` foldr ins [] [2,6,3,1,4]
  = 5 `ins` 2 `ins` [6,3,1,4] ...
  = 5 `ins` 2 `ins` 6 `ins` 3 `ins` 1 `ins` 4 `ins` []
  = 5 `ins` 2 `ins` 6 `ins` 3 `ins` 1 `ins` (4:[])
  = 5 `ins` 2 `ins` 6 `ins` 3 `ins` (1:4:[])
  = 5 `ins` 2 `ins` 6 `ins` (1 : (3 `ins` (4:[])))
  = 5 `ins` 2 `ins` (1 : (6 `ins` (3 `ins` (4:[]))))
  = 5 `ins` (1 : (2 `ins` (6 `ins` (3 `ins` (4:[])))))
  = 1 : (5 `ins` (2 `ins` (6 `ins` (3 `ins` (4:[])))))  -- now 1 can be output
  = 1 : (5 `ins` (2 `ins` (6 `ins` (3:4:[]))))
  = 1 : (5 `ins` (2 `ins` (3 : (6 `ins` (4:[])))))
  = 1 : (5 `ins` (2 : (3 : (6 `ins` (4:[])))))
  = 1 : 2 : (5 `ins` (3 : (6 `ins` (4:[]))))            -- now 2 can be output
  = 1 : 2 : 3 : (5 `ins` (6 `ins` (4:[])))              -- now 3
  = 1 : 2 : 3 : (5 `ins` (4:6:[]))
  = 1 : 2 : 3 : 4 : (5 `ins` (6:[]))                    -- now 4
  = 1 : 2 : 3 : 4 : 5 : 6 : []                          -- done

和合并排序(缩写：merge = mg，merge_sort = ms)：

merge_sort [5,2,6,3,1,4]
  = mg (ms [5,2,6]) (ms [3,1,4])
  = mg (mg (ms [5]) (ms [2,6])) (mg (ms [3]) (ms [1,4]))
  = mg (mg [5] (mg [2] [6])) (mg [3] (mg [1] [4]))
  = mg (mg [5] [2,6]) (mg [3] [1,4])
  = mg (2 : mg [5] [6]) (1 : mg [3] [4])
  = 1 : mg (2 : mg [5] [6]) (mg [3] [4])                -- now 1 can be output
  = 1 : mg (2 : mg [5] [6]) [3,4]
  = 1 : 2 : mg (mg [5] [6]) [3,4]                       -- now 2 can be output
  = 1 : 2 : mg [5,6] [3,4]
  = 1 : 2 : 3 : mg [5,6] [4]                            -- now 3
  = 1 : 2 : 3 : 4 : mg [5,6] []                         -- now 4
  = 1 : 2 : 3 : 4 : 5 : 6 : []                          -- now 5 and 6

诚然，我走了几条捷径，但哈斯克尔并不是唯一一个懒惰的人。