如何删除Lisp代码中的冗余?

时间:2016-01-20 19:59:23

标签: coding-style lisp common-lisp

我试图在Common Lisp中提出快速排序的实现,这是我到目前为止所做的:

(defun quick-sort (list)
  (if (cdr list)
    (let ((pivot (car list)))
      (append (quick-sort (remove-if-not (lambda (n) (< n pivot)) list))
              (remove-if-not (lambda (n) (= n pivot)) list)
              (quick-sort (remove-if-not (lambda (n) (> n pivot)) list))))
    list))

显然它可行,但我认为该代码中有太多重复。基本上,我们有三次:

(remove-if-not (lambda (n) (< n pivot)) list)

这三个来电的唯一不同方式是> vs = vs <

因此我的问题是:如何删除冗余并使代码更具可读性和更紧凑?

当然,我可以将事物提取到函数中,例如:

(defun which (operator pivot list )
  (remove-if-not (lambda (n) (funcall operator n pivot)) list))

(defun quick-sort (list)
  (if (cdr list)
    (let ((pivot (car list)))
      (append (quick-sort (which #'< pivot list))
              (which #'= pivot list)
              (quick-sort (which #'> pivot list))))
    list))

但不知何故,我并不确定这是否是最佳方法。一次又一次地移交pivotlist仍然感觉很笨拙。

我也有使用flet的想法,它使函数的实际主体更具可读性,但只将复杂性移到其他地方:

(defun quick-sort (list)
  (if (cdr list)
    (let ((pivot (car list)))
      (flet ((left () (remove-if-not (lambda (n) (< n pivot)) list))
             (middle () (remove-if-not (lambda (n) (= n pivot)) list))
             (right () (remove-if-not (lambda (n) (> n pivot)) list)))
        (append (quick-sort (left))
                (middle)
                (quick-sort (right)))))
    list))

还有其他方法吗?

1 个答案:

答案 0 :(得分:16)

如果你把它写成本地函数,你就不必传递额外的参数,因为它们在范围内。

(defun quick-sort (list)
  (if (rest list)
      (let ((pivot (first list)))
        (flet ((filter (operator)
                 (remove-if-not (lambda (n) (funcall operator n pivot)) list)))
          (append (quick-sort (filter #'<))
                  (filter #'=)
                  (quick-sort (filter #'>)))))
    list))

稍微紧凑的版本:

(defun quick-sort (list &aux (pivot (first list)))
  (flet ((filter (operator)
           (remove-if-not (lambda (n) (funcall operator n pivot)) list)))
    (and list
         (nconc (quick-sort (filter #'<))
                (filter #'=)
                (quick-sort (filter #'>))))))

由于Common Lisp支持多个值,您也可以一次性在一个函数中对列表进行分区,并将列表作为值返回:

(defun partition (list pivot)
  (loop for e in list
        when (< e pivot) collect e into smaller else
        when (> e pivot) collect e into larger else
        when (= e pivot) collect e into same
        finally (return (values smaller same larger))))

(defun quick-sort (list)
  (if (rest list)
      (multiple-value-bind (smaller same larger)
          (partition list (first list))
        (append (quick-sort smaller) same (quick-sort larger)))
    list))

当新分配列表时,NCONC是可能的。由于REMOVE-IF-NOT是非破坏性的(与DELETE-IF-NOT比较),NCONC很好。由于LOOP收集了新列表,NCONC再次正常。

这是一个真正简单的就地Quicksort而不是矢量。请注意,Quicksort实际上就是这种意思。使用列表的版本不是真正的Quicksort。

(defun partition (array left right &aux (i (1- left))
                                        (j right)
                                        (v (aref array right)))
  (loop do (loop do (incf i) until (>= (aref array i) v))
           (loop do (decf j) until (or (zerop j)
                                       (<= (aref array j) v)))
           (rotatef (aref array i) (aref array j))
        until (<= j i))
  (rotatef (aref array j) (aref array i) (aref array right))
  i)

(defun quicksort (array &optional (left 0) (right (1- (length array))))
  (if (> right left)
    (let ((i (partition array left right)))
      (quicksort array left (1- i))
      (quicksort array (1+ i) right))
    array))

此版本基于Sedgewick的代码。