Java 在单个循环中以声明方式分别按多个字段进行流分组

Question

我用谷歌搜索了它，但我发现大多数情况是按聚合字段分组或改变流的响应，但不是下面的情况：

我有一个 class User 字段 category 和 marketingChannel.

我必须以声明式的方式编写一个方法，该方法接受用户列表并根据 category 并且也分别基于 marketingChannel （即不是 groupingBy(... ,groupingBy(..)) ）。

我无法在一个循环中完成。这就是我必须要达到的目标。

我编写了几个方法如下：

import java.util.*;
import java.util.stream.*;
public class Main
{
    public static void main(String[] args) {
        List<User> users = User.createDemoList();
        imperative(users);
        declerativeMultipleLoop(users);
        declerativeMultipleColumn(users);
    }
    
    public static void imperative(List<User> users){
        Map<String, Integer> categoryMap = new HashMap<>();
        Map<String, Integer> channelMap = new HashMap<>();
        for(User user : users){
           Integer  value = categoryMap.getOrDefault(user.getCategory(), 0);
           categoryMap.put(user.getCategory(), value+1);
           value = channelMap.getOrDefault(user.getMarketingChannel(), 0);
           channelMap.put(user.getMarketingChannel(), value+1);
        }
        System.out.println("imperative");
        System.out.println(categoryMap);
        System.out.println(channelMap);
    }
    
    public static void declerativeMultipleLoop(List<User> users){
        Map<String, Long> categoryMap = users.stream()
        .collect(Collectors.groupingBy(
            User::getCategory, Collectors.counting()));
        Map<String, Long> channelMap = users.stream()
        .collect(Collectors.groupingBy(
            User::getMarketingChannel, Collectors.counting()));
        System.out.println("declerativeMultipleLoop");
        System.out.println(categoryMap);
        System.out.println(channelMap);
    }
    
    public static void declerativeMultipleColumn(List<User> users){
        Map<String, Map<String, Long>> map = users.stream()
        .collect(Collectors.groupingBy(
            User::getCategory,
            Collectors.groupingBy(User::getMarketingChannel, 
            Collectors.counting())));
       
        System.out.println("declerativeMultipleColumn");
        System.out.println("groupingBy category and marketChannel");
        System.out.println(map);
        
        Map<String, Long> categoryMap = new HashMap<>();
        Map<String, Long> channelMap = new HashMap<>();
        
        for (Map.Entry<String, Map<String, Long>> entry: map.entrySet()) {
            String category = entry.getKey();
            Integer count = entry.getValue().size();
            Long value = categoryMap.getOrDefault(category,0L);
            categoryMap.put(category, value+count);
            for (Map.Entry<String, Long> channelEntry : entry.getValue().entrySet()){
                String channel = channelEntry.getKey();
                Long channelCount = channelEntry.getValue();
                Long channelValue = channelMap.getOrDefault(channel,0L);
                channelMap.put(channel, channelValue+channelCount);
            }
        }
        System.out.println("After Implerative Loop on above.");
        System.out.println(categoryMap);
        System.out.println(channelMap);
    }
}
class User{
    private String name;
    private String category;
    private String marketChannel;
    
    public User(String name, String category, String marketChannel){
        this.name = name;
        this.category = category;
        this.marketChannel = marketChannel;
    }
    public String getName(){
        return this.name;
    }
    public String getCategory(){
        return this.category;
    }
    public String getMarketingChannel(){
        return this.marketChannel;
    }
    
     @Override
    public boolean equals(Object o) {
        if (this == o) return true;
        if (o == null || getClass() != o.getClass()) return false;
        User user = (User) o;
        return Objects.equals(name, user.name) &&
                Objects.equals(category, user.category) &&
                Objects.equals(marketChannel, user.marketChannel);
    }

    @Override
    public int hashCode() {
        return Objects.hash(name, category, marketChannel);
    }
    public static List<User> createDemoList(){
        return Arrays.asList(
            new User("a", "student","google"),
            new User("b", "student","bing"),
            new User("c", "business","google"),
            new User("d", "business", "direct")
            );
    }

The method declerativeMultipleLoop is declarative but it has a separate loop for each field. Complexity : O(noOfFields * No of users)

The problem is in declerativeMultipleColumn Method as I end up writing imperative code and multiple loops.

I want to write the above method in completely declarative and as efficient as possible. i.e Complexity : O(No of users)

示例输出：

imperative
{business=2, student=2}
{direct=1, google=2, bing=1}
declerativeMultipleLoop
{business=2, student=2}
{direct=1, google=2, bing=1}
declerativeMultipleColumn
groupingBy category and marketChannel
{business={direct=1, google=1}, student={google=1, bing=1}}
After Implerative Loop on above.
{business=2, student=2}
{direct=1, google=2, bing=1}

Answer 1

如果我理解您的要求，那就是使用产生 2 个单独地图的单个流操作。这将需要一个结构来保存地图和一个收集器来构建结构。类似于以下内容：

class Counts {
    public final Map<String, Integer> categoryCounts = new HashMap<>();
    public final Map<String, Integer> channelCounts = new HashMap<>();

    public static Collector<User,Counts,Counts> countsCollector() {
        return Collector.of(Counts::new, Counts::accept, Counts::combine, CONCURRENT, UNORDERED);
    }

    private Counts() { }

    private void accept(User user) {
        categoryCounts.merge(user.getCategory(), 1, Integer::sum);
        channelCounts.merge(user.getChannel(), 1, Integer::sum);
    }

    private Counts combine(Counts other) {
        other.categoryCounts.forEach((c, v) -> categoryCounts.merge(c, v, Integer::sum));
        other.channelCounts.forEach((c, v) -> channelCounts.merge(c, v, Integer::sum));
        return this;
    }
}

然后可以用作收集器：

Counts counts = users.stream().collect(Counts.countsCollector());
counts.categoryCounts.get("student")...

（仅供参考：在这种情况下，命令式和声明式之间的区别非常随意。对我来说，定义流操作感觉很程序化（与 Haskell 中的等价物相反）。

Answer 2

您可以compute two maps in a single forEach方法：

public static void main(String[] args) {
    List<User> users = Arrays.asList(
            new User("a", "student", "google"),
            new User("b", "student", "bing"),
            new User("c", "business", "google"),
            new User("d", "business", "direct"));
    Map<String, Integer> categoryMap = new HashMap<>();
    Map<String, Integer> channelMap = new HashMap<>();

    // group users into maps
    users.forEach(user -> {
        categoryMap.compute(user.getCategory(),
                (key, value) -> value == null ? 1 : value + 1);
        channelMap.compute(user.getChannel(),
                (key, value) -> value == null ? 1 : value + 1);
    });
    // output
    System.out.println(categoryMap); // {business=2, student=2}
    System.out.println(channelMap); // {direct=1, google=2, bing=1}
}

static class User {
    private final String name, category, channel;

    public User(String name, String category, String channel) {
        this.name = name;
        this.category = category;
        this.channel = channel;
    }

    public String getName() { return this.name; }
    public String getCategory() { return this.category; }
    public String getChannel() { return this.channel; }
}

Java 在单个循环中以声明方式分别按多个字段进行流分组

Java Stream Grouping by multiple fields individually in declarative way in single loop

java

declarative-programming

java-stream

groupingby