📜  使用就地排序算法对对象进行排序

📅  最后修改于: 2021-06-25 22:09:08             🧑  作者: Mango

给定一个红色蓝色黄色对象的数组,任务是使用就地排序算法对数组进行排序,以使所有蓝色对象出现在所有红色对象之前,所有红色对象出现在所有对象之前。黄色物体。

例子:

方法:首先分别使用哈希表的蓝色红色黄色的对象的值映射到1,23。现在,只要需要比较两个对象,就使用这些映射值。因此,该算法将对对象数组进行排序,以便首先显示所有蓝色对象(映射到值1),然后显示所有红色对象(映射到值2),然后显示所有黄色对象(映射到值3)。

下面是上述方法的实现:

C++
// C++ implementation of the approach
#include 
using namespace std;
  
// Partition function which will partition
// the array and into two parts
int partition(vector& objects, int l, int r,
            unordered_map& hash)
{
    int j = l - 1;
  
    int last_element = hash[objects[r]];
  
    for (int i = l; i < r; i++) {
  
        // Compare hash values of objects
        if (hash[objects[i]] <= last_element) {
            j++;
            swap(objects[i], objects[j]);
        }
    }
  
    j++;
  
    swap(objects[j], objects[r]);
  
    return j;
}
  
// Classic quicksort algorithm
void quicksort(vector& objects, int l, int r, 
                    unordered_map& hash)
{
    if (l < r) {
        int mid = partition(objects, l, r, hash);
        quicksort(objects, l, mid - 1, hash);
        quicksort(objects, mid + 1, r, hash);
    }
}
  
// Function to sort and print the objects
void sortObj(vector& objects)
{
  
    // Create a hash table
    unordered_map hash;
  
    // As the sorting order is blue objects,
    // red objects and then yellow objects
    hash["blue"] = 1;
    hash["red"] = 2;
    hash["yellow"] = 3;
  
    // Quick sort function
    quicksort(objects, 0, int(objects.size() - 1), hash);
  
    // Printing the sorted array
    for (int i = 0; i < objects.size(); i++)
        cout << objects[i] << " ";
}
  
// Driver code
int main()
{
  
    // Let's represent objects as strings
    vector objects{ "red", "blue",
                            "red", "yellow", "blue" };
  
    sortObj(objects);
  
    return 0;
}


Java
// Java implementation of the approach
import java.util.*;
class GFG 
{
  
// Partition function which will partition
// the array and into two parts
static int partition(Vector objects, int l, int r,
                        Map hash)
{
    int j = l - 1;
  
    int last_element = hash.get(objects.get(r));
  
    for (int i = l; i < r; i++) 
    {
  
        // Compare hash values of objects
        if (hash.get(objects.get(i)) <= last_element) 
        {
            j++;
            Collections.swap(objects, i, j);
        }
    }
  
    j++;
  
    Collections.swap(objects, j, r);
  
    return j;
}
  
// Classic quicksort algorithm
static void quicksort(Vector objects, int l, int r,
                         Map hash)
{
    if (l < r)
    {
        int mid = partition(objects, l, r, hash);
        quicksort(objects, l, mid - 1, hash);
        quicksort(objects, mid + 1, r, hash);
    }
}
  
// Function to sort and print the objects
static void sortObj(Vector objects)
{
  
    // Create a hash table
    Map hash = new HashMap<>();
  
    // As the sorting order is blue objects,
    // red objects and then yellow objects
    hash. put("blue", 1);
    hash. put("red", 2);
    hash. put("yellow", 3);
  
    // Quick sort function
    quicksort(objects, 0, objects.size() - 1, hash);
  
    // Printing the sorted array
    for (int i = 0; i < objects.size(); i++)
        System.out.print(objects.get(i) + " ");
}
  
// Driver code
public static void main(String []args)
{
    // Let's represent objects as strings
    Vector objects = new Vector<>(Arrays.asList( "red", "blue",
                                                         "red", "yellow", 
                                                         "blue" ));
  
    sortObj(objects);
}
}
  
// This code is contributed by PrinciRaj1992


Python3
# Python3 implementation of the approach
  
# Partition function which will partition
# the array and into two parts
objects = []
hash = dict()
  
def partition(l, r):
    global objects, hash
    j = l - 1
  
    last_element = hash[objects[r]]
  
    for i in range(l, r):
  
        # Compare hash values of objects
        if (hash[objects[i]] <= last_element):
            j += 1
            (objects[i], 
             objects[j]) = (objects[j], 
                            objects[i])
  
    j += 1
  
    (objects[j], 
     objects[r]) = (objects[r], 
                    objects[j])
  
    return j
  
# Classic quicksort algorithm
def quicksort(l, r):
    if (l < r):
        mid = partition(l, r)
        quicksort(l, mid - 1)
        quicksort(mid + 1, r)
  
# Function to sort and print the objects
def sortObj():
    global objects, hash
  
    # As the sorting order is blue objects,
    # red objects and then yellow objects
    hash["blue"] = 1
    hash["red"] = 2
    hash["yellow"] = 3
  
    # Quick sort function
    quicksort(0, int(len(objects) - 1))
  
    # Printing the sorted array
    for i in objects:
        print(i, end = " ")
  
# Driver code
  
# Let's represent objects as strings
objects = ["red", "blue", "red", 
               "yellow", "blue"]
  
sortObj()
  
# This code is contributed 
# by Mohit Kumar


C#
// C# implementation of the approach
using System;
using System.Collections.Generic;
  
class GFG 
{
  
// Partition function which will partition
// the array and into two parts
static int partition(List objects, int l, int r,
                           Dictionary hash)
{
    int j = l - 1;
    String temp;
    int last_element = hash[objects[r]];
  
    for (int i = l; i < r; i++) 
    {
  
        // Compare hash values of objects
        if (hash[objects[i]] <= last_element) 
        {
            j++;
            temp = objects[i];
            objects[i] = objects[j];
            objects[j] = temp;
        }
    }
  
    j++;
  
    temp = objects[r];
    objects[r] = objects[j];
    objects[j] = temp;
  
    return j;
}
  
// Classic quicksort algorithm
static void quicksort(List objects, int l, int r,
                            Dictionary hash)
{
    if (l < r)
    {
        int mid = partition(objects, l, r, hash);
        quicksort(objects, l, mid - 1, hash);
        quicksort(objects, mid + 1, r, hash);
    }
}
  
// Function to sort and print the objects
static void sortObj(List objects)
{
  
    // Create a hash table
    Dictionary hash = new Dictionary();
  
    // As the sorting order is blue objects,
    // red objects and then yellow objects
    hash.Add("blue", 1);
    hash.Add("red", 2);
    hash.Add("yellow", 3);
  
    // Quick sort function
    quicksort(objects, 0, objects.Count - 1, hash);
  
    // Printing the sorted array
    for (int i = 0; i < objects.Count; i++)
        Console.Write(objects[i] + " ");
}
  
// Driver code
public static void Main(String []args)
{
    // Let's represent objects as strings
    List objects = new List{"red", "blue",
                                            "red", "yellow", 
                                            "blue"};
  
    sortObj(objects);
}
}
  
// This code is contributed by Rajput-Ji


输出:
blue blue red red yellow

如果您希望与行业专家一起参加现场课程,请参阅《 Geeks现场课程》和《 Geeks现场课程美国》。