java - 建议将递归 Diff 算法转换为迭代算法？

Question

您好!

我从头开始写了一个递归差异算法。它找到两个字符串之间的“最佳匹配”，使差异最小化，并打印出两个字符串，并以 CAPS 表示任何差异。它“按原样”工作正常，只是效率很低。我已经盯着它看了一天半了，试图找到让它迭代的方法，或者至少减少它达到的堆栈深度，但我已经无计可施了，希望这里有一个敏锐的头脑会比我更清楚地看到解决方案。

下面是代码的主要内容。引用的 MergePoint 类只是一个简单的“链表”样式节点，它包含一个“原始索引”整数、一个“更改后的索引”整数和一个“下一个”MergePoint。 MergePoint 列表表示每个数组中已“合并”的一系列索引。当链完成时，链中未表示的任何索引都是插入/删除。 NullObject 对象是 MergePoint 的扩展，回想起来，创建它并不是绝对必要的，基本上可以将其视为常规“null”。

任何意见/建议将不胜感激。

public class StringCompare 
{
    public static int[][]       mergeList       = new int[0][0];
    public static MergePoint    NULL            = NullObject.getNull();
    public static int           maxMerged       = 0;
    public static int           minClusterSize  = -1;

    public static void diff(String orig, String alt)
    {
        String[] original = orig.toUpperCase().split(" ");
        String[] altered  = alt.toUpperCase().split(" ");

        for(int i = 0; i < altered.length; i++)
        {
            merge(original, altered, 0, i, NULL, NULL, 0, 0);
        }


        for(int i = 0; i < mergeList.length; i++)
        {
            or[mergeList[i][0]] = or[mergeList[i][0]].toLowerCase();
            al[mergeList[i][1]] = al[mergeList[i][1]].toLowerCase();
        }

        printStringArray(or);
        printStringArray(al);
    }

    private void printStringArray(String[] arr)
    {
        for(String word : arr)
        {
            System.out.print(word.trim() + " ");
        }

        System.out.println();
    }

    private static void merge(String[] original, String[] altered, int indexInOriginal, int indexInAltered, MergePoint head, MergePoint tail, int listSize, int clusters)
    {
        if (indexInOriginal >= original.length)
        {           
            if (listSize > 0)
            {
                if (((listSize == maxMerged) && (clusters < minClusterSize)) ||
                    (listSize > maxMerged))
                {
                    storeMergePoints(head, listSize, clusters);
                }
            }
        }
        else if (indexInAltered >= altered.length)
        {
            if (tail != NULL)
            {
                merge(original, altered, (indexInOriginal + 1), (tail.indexInNew() + 1), head, tail, listSize, clusters);
            }
            else
            {
                merge(original, altered, (indexInOriginal + 1), 0, head, tail, listSize, 0);
            }
        }
        else
        {
            if(original[indexInOriginal].equals(altered[indexInAltered]))
            {
                MergePoint  mergePoint  = new MergePoint(indexInOriginal, indexInAltered);
                MergePoint  bookMark    = NULL;             
                int         newClusters = clusters;

                if (indexInOriginal != (tail.indexInOriginal() + 1))
                {
                    newClusters++;
                }

                if (indexInAltered != (tail.indexInNew() + 1))
                {
                    newClusters++;
                }

                if (head == NULL)
                {
                    head = mergePoint;
                    tail = head;
                }
                else
                {
                    tail.setNext(mergePoint);

                    bookMark    = tail;                 
                    tail        = tail.next();
                }

                merge(original, altered, (indexInOriginal + 1), (indexInAltered + 1), head, tail, (listSize + 1), newClusters);

                if (bookMark == NULL)
                {
                    merge(original, altered, indexInOriginal, (indexInAltered + 1), NULL, NULL, 0, 0);
                }
                else
                {
                    bookMark.setNext(NULL);

                    merge(original, altered, indexInOriginal, (indexInAltered + 1), head, bookMark, listSize, newClusters);
                }
            }
            else
            {
                merge(original, altered, indexInOriginal, (indexInAltered + 1), head, tail, listSize, clusters);
            }
        }
    }

    public static void storeMergePoints(MergePoint current, int size, int clusters)
    {       
        mergeList       = new int[size][2];
        maxMerged       = size;
        minClusterSize  = clusters;

        for(int i = 0; i < size; i++)
        {
            mergeList[i][0] = current.indexInOriginal();
            mergeList[i][1] = current.indexInNew();
            current         = current.next();
        }       
    }
}

Answer 1

为了用迭代替换递归，您可以考虑用您控制的堆栈对象替换 JVM 堆栈:java.util.Stack 非常适合这种情况。只需在每次迭代时 push() 和 pop() 您的数据到该堆栈上，而不是让方法自行调用。