假设我们有两个堆栈,没有其他临时变量。

是否有可能“构造”一个队列数据结构只使用两个堆栈?


当前回答

使用O(1) dequeue(),这与pythonquick的答案相同:

// time: O(n), space: O(n)
enqueue(x):
    if stack.isEmpty():
        stack.push(x)
        return
    temp = stack.pop()
    enqueue(x)
    stack.push(temp)

// time: O(1)
x dequeue():
    return stack.pop()

使用O(1) enqueue()(这在本文中没有提到,所以这个答案),它也使用回溯来冒泡并返回最底部的项。

// O(1)
enqueue(x):
    stack.push(x)

// time: O(n), space: O(n)
x dequeue():
    temp = stack.pop()
    if stack.isEmpty():
        x = temp
    else:
        x = dequeue()
        stack.push(temp)
    return x

显然,这是一个很好的编码练习,因为它效率很低,但仍然很优雅。

其他回答

您必须从第一个堆栈中取出所有元素来获取底部元素。然后在每次“出队列”操作时将它们都放回第二个堆栈。

保持2个堆栈,让我们称之为收件箱和发件箱。

排队:

将新元素推到收件箱上

出列:

如果发件箱为空,则通过弹出收件箱中的每个元素并将其推入发件箱来重新填充它 弹出并返回发件箱中的顶部元素

使用这种方法,每个元素只在每个堆栈中存在一次——这意味着每个元素将被压入两次,弹出两次,从而给出平摊常数时间操作。

下面是Java中的实现:

public class Queue<E>
{

    private Stack<E> inbox = new Stack<E>();
    private Stack<E> outbox = new Stack<E>();

    public void queue(E item) {
        inbox.push(item);
    }

    public E dequeue() {
        if (outbox.isEmpty()) {
            while (!inbox.isEmpty()) {
               outbox.push(inbox.pop());
            }
        }
        return outbox.pop();
    }

}

使用堆栈实现队列的以下操作。

push(x)——将元素x推到队列的后面。

pop()——从队列前面移除元素。

peek()——获取前端元素。

empty()——返回队列是否为空。

class MyQueue {

  Stack<Integer> input;
  Stack<Integer> output;

  /** Initialize your data structure here. */
  public MyQueue() {
    input = new Stack<Integer>();
    output = new Stack<Integer>();
  }

  /** Push element x to the back of queue. */
  public void push(int x) {
    input.push(x);
  }

  /** Removes the element from in front of queue and returns that element. */
  public int pop() {
    peek();
    return output.pop();
  }

  /** Get the front element. */
  public int peek() {
    if(output.isEmpty()) {
        while(!input.isEmpty()) {
            output.push(input.pop());
        }
    }
    return output.peek();
  }

  /** Returns whether the queue is empty. */
  public boolean empty() {
    return input.isEmpty() && output.isEmpty();
  }
}

我将在Go中回答这个问题,因为Go在其标准库中没有丰富的集合。

由于堆栈真的很容易实现,我想我应该尝试使用两个堆栈来完成一个双端队列。为了更好地理解我是如何得到我的答案的,我将实现分为两部分,第一部分希望更容易理解,但它是不完整的。

type IntQueue struct {
    front       []int
    back        []int
}

func (q *IntQueue) PushFront(v int) {
    q.front = append(q.front, v)
}

func (q *IntQueue) Front() int {
    if len(q.front) > 0 {
        return q.front[len(q.front)-1]
    } else {
        return q.back[0]
    }
}

func (q *IntQueue) PopFront() {
    if len(q.front) > 0 {
        q.front = q.front[:len(q.front)-1]
    } else {
        q.back = q.back[1:]
    }
}

func (q *IntQueue) PushBack(v int) {
    q.back = append(q.back, v)
}

func (q *IntQueue) Back() int {
    if len(q.back) > 0 {
        return q.back[len(q.back)-1]
    } else {
        return q.front[0]
    }
}

func (q *IntQueue) PopBack() {
    if len(q.back) > 0 {
        q.back = q.back[:len(q.back)-1]
    } else {
        q.front = q.front[1:]
    }
}

它基本上是两个堆栈,我们允许堆栈的底部相互操纵。我还使用了STL命名约定,其中堆栈的传统push、pop、peek操作都有一个front/back前缀,无论它们是指队列的前面还是后面。

上面代码的问题是它没有非常有效地使用内存。事实上,它会不断增长,直到空间耗尽。这太糟糕了。解决这个问题的方法是尽可能重用堆栈空间的底部。我们必须引入一个偏移量来跟踪这一点,因为围棋中的切片一旦收缩就不能在前面生长。

type IntQueue struct {
    front       []int
    frontOffset int
    back        []int
    backOffset  int
}

func (q *IntQueue) PushFront(v int) {
    if q.backOffset > 0 {
        i := q.backOffset - 1
        q.back[i] = v
        q.backOffset = i
    } else {
        q.front = append(q.front, v)
    }
}

func (q *IntQueue) Front() int {
    if len(q.front) > 0 {
        return q.front[len(q.front)-1]
    } else {
        return q.back[q.backOffset]
    }
}

func (q *IntQueue) PopFront() {
    if len(q.front) > 0 {
        q.front = q.front[:len(q.front)-1]
    } else {
        if len(q.back) > 0 {
            q.backOffset++
        } else {
            panic("Cannot pop front of empty queue.")
        }
    }
}

func (q *IntQueue) PushBack(v int) {
    if q.frontOffset > 0 {
        i := q.frontOffset - 1
        q.front[i] = v
        q.frontOffset = i
    } else {
        q.back = append(q.back, v)
    }
}

func (q *IntQueue) Back() int {
    if len(q.back) > 0 {
        return q.back[len(q.back)-1]
    } else {
        return q.front[q.frontOffset]
    }
}

func (q *IntQueue) PopBack() {
    if len(q.back) > 0 {
        q.back = q.back[:len(q.back)-1]
    } else {
        if len(q.front) > 0 {
            q.frontOffset++
        } else {
            panic("Cannot pop back of empty queue.")
        }
    }
}

有很多小函数,但6个函数中有3个只是另一个的镜像。

public class QueueUsingStacks<T>
{
    private LinkedListStack<T> stack1;
    private LinkedListStack<T> stack2;

    public QueueUsingStacks()
    {
        stack1=new LinkedListStack<T>();
        stack2 = new LinkedListStack<T>();

    }
    public void Copy(LinkedListStack<T> source,LinkedListStack<T> dest )
    {
        while(source.Head!=null)
        {
            dest.Push(source.Head.Data);
            source.Head = source.Head.Next;
        }
    }
    public void Enqueue(T entry)
    {

       stack1.Push(entry);
    }
    public T Dequeue()
    {
        T obj;
        if (stack2 != null)
        {
            Copy(stack1, stack2);
             obj = stack2.Pop();
            Copy(stack2, stack1);
        }
        else
        {
            throw new Exception("Stack is empty");
        }
        return obj;
    }

    public void Display()
    {
        stack1.Display();
    }


}

对于每一个入队列操作,我们都将其添加到stack1的顶部。每次出队列时,我们都将stack1的内容清空到stack2中,并删除堆栈顶部的元素。出队列的时间复杂度是O(n),因为我们必须将stack1复制到stack2。队列的时间复杂度与常规堆栈相同