如果你细想想,就会发现,非零元节点如果没有指示位置的域,那么做加法和乘法时,为了确定节点的位置,每次都要遍历行和列的链表。因此,为了运算效率,这个域是必须的。为了看出十字链表和单链表的差异,我从单链表派生出十字链表,这需要先定义一种新的结构,如下:
class MatNode
{
public:
int data;
int row, col;
union { Node<MatNode> *down; List<MatNode> *downrow; };
};
另外,由于这样的十字链表是由多条单链表拼起来的,为了访问每条单链表的保护成员,要声明十字链表类为单链表类的友元。即在class List的声明中添加friend class Matrix;
稀疏矩阵的定义和实现
#ifndef Matrix_H
#define Matrix_H
#include "List.h"
class MatNode
{
public:
int data;
int row, col;
union { Node<MatNode> *down; List<MatNode> *downrow; };
MatNode(int value = 0, Node<MatNode> *p = NULL, int i = 0, int j = 0)
: data(value), down(p), row(i), col(j) {}
friend ostream & operator << (ostream & strm, MatNode &mtn)
{
strm << '(' << mtn.row << ',' << mtn.col << ')' << mtn.data;
return strm;
}
};
class Matrix : List<MatNode>
{
public:
Matrix() : row(0), col(0), num(0) {}
Matrix(int row, int col, int num) : row(row), col(col), num(num) {}
~Matrix() { MakeEmpty(); }
void MakeEmpty()
{
List<MatNode> *q;
while (first->data.downrow != NULL)
{
q = first->data.downrow;
first->data.downrow = q->first->data.downrow;
delete q;
}
List<MatNode>::MakeEmpty();
row = col = num = 0;
}
void Input()
{
if (!row) { cout << "输入矩阵行数:"; cin >> row; }
if (!col) { cout << "输入矩阵列数:"; cin >> col; }
if (!num) { cout << "输入非零个数:"; cin >> num; }
if (!row || !col || !num) return;
cout << endl << "请按顺序输入各个非零元素,以列序为主,输入0表示本列结束" << endl;
int i, j, k, v;//i行数 j列数 k个非零元 v非零值
Node<MatNode> *p = first, *t;
List<MatNode> *q;
for (j = 1; j <= col; j++) LastInsert(MatNode(0, NULL, 0, j));
for (i = 1; i <= row; i++)
{
q = new List<MatNode>;
q->first->data.row = i;
p->data.downrow = q;
p = q->first;
}
j = 1; q = first->data.downrow; First(); t = pNext();
for (k = 0; k < num; k++)
{
if (j > col) break;
cout << endl << "输入第" << j << "列非零元素" << endl;
cout << "行数:"; cin >> i;
if (i < 1 || i > row) { j++; k--; q = first->data.downrow; t = pNext(); continue; }
cout << "非零元素值"; cin >> v;
if (!v) { k--; continue; }
MatNode matnode(v, NULL, i, j);
p = new Node<MatNode>(matnode);
t->data.down = p; t = p;
while (q->first->data.row != i) q = q->first->data.downrow;
q->LastInsert(t);
}
}
void Print()
{
List<MatNode> *q = first->data.downrow;
cout << endl;
while (q != NULL)
{
cout << *q;
q = q->first->data.downrow;
}
}
Matrix & Add(Matrix &matB)
{
//初始化赋值辅助变量
if (row != matB.row || col != matB.col || matB.num == 0) return *this;
Node<MatNode> *pA, *pB;
Node<MatNode> **pAT = new Node<MatNode>*[col + 1];
Node<MatNode> **pBT = new Node<MatNode>*[matB.col + 1];
List<MatNode> *qA = pGetFirst()->data.downrow, *qB = matB.pGetFirst()->data.downrow;
First(); matB.First();
for (int j = 1; j <= col; j++)
{
pAT[j] = pNext();
pBT[j] = matB.pNext();
}
//开始
for (int i = 1; i <= row; i++)
{
qA->First(); qB->First();
pA = qA->pNext(); pB = qB->pNext();
while (pA != NULL && pB !=NULL)
{
if (pA->data.col == pB->data.col)
{
pA->data.data += pB->data.data;
pBT[pB->data.col]->data.down = pB->data.down; qB->Remove();
if (!pA->data.data)
{
pAT[pA->data.col]->data.down = pA->data.down;
qA->Remove();
}
else
{
pAT[pA->data.col] = pA;
qA->pNext();
}
}
else
{
if (pA->data.col > pB->data.col)
{
pBT[pB->data.col]->data.down = pB->data.down;
qB->pRemove();
pB->data.down = pAT[pB->data.col]->data.down;
pAT[pB->data.col]->data.down = pB;
pAT[pB->data.col] = pB;
qA->InsertBefore(pB);
}
else if (pA->data.col < pB->data.col)
{
pAT[pA->data.col] = pA;
qA->pNext();
}
}
pA = qA->pGet();pB = qB->pGet();
}
if (pA == NULL && pB != NULL)
{
qA->pGetPrior()->link = pB;
qB->pGetPrior()->link = NULL;
while (pB != NULL)
{
pBT[pB->data.col]->data.down = pB->data.down;
pB->data.down = pAT[pB->data.col]->data.down;
pAT[pB->data.col]->data.down = pB;
pAT[pB->data.col] = pB;
pB = pB->link;
}
}
if (pA !=NULL)
{
while (qA->pGet() != NULL)
{
pAT[pA->data.col] = pA;
qA->pNext();
}
}
qA = qA->first->data.downrow; qB = qB->first->data.downrow;
}
delete []pAT; delete []pBT;
return *this;
}
private:
int row, col, num;
};
#endif
【说明】对于十字链表来说,只要记住对每个节点的操作,要同时考虑它的两个指针域,那么,各种算法的理解都不是很难。比如说矩阵加法,“两个矩阵相加和两个一元多项式相加极为相似,所不同的是一元多项式只有一个变元(指数项),而矩阵中每个非零元有两个变元(行值和列值),每个节点既在行表中又在列表中,致使插入和删除节点时指针的修改稍为复杂,故需要更多的辅助指针。”(《数据结构(C语言版)》)其实private的row等可以放在首行的头节点里的,但为了清晰一点(本来就够乱了),我把他们单立出来了。另外,很多地方考虑不是很周全,要是不按照注明的要求使用,很容易就会出错。
【后记】按理说,十字链表应该不算是线性链式结构,按照原书的安排,放在链表这章不是很合适;《数据结构(C语言版)》将它和广义表放在一章还是合理的。其实十字链表不是很难,就是很烦人;并且,如果不是数值运算,基本很少用到矩阵,就算是用到矩阵运算,在矩阵规模不大的时候,可以用二维数组代替十字链表。从历届考研题来看,这部分几乎没有题,原因就是麻烦(你写起来麻烦,他批起来也麻烦)、不常用、算法固定没新意。所以,你要是闹心,这部分跳过去也可以。