本文主要讲解:Java 8 Stream之Collectors.groupingBy()分组示例
Collectors.groupingBy() 分组之常见用法
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list */ public void groupingByCity() { Map<String, List<Employee>> map = employees.stream().collect(Collectors.groupingBy(Employee::getCity)); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之统计每个分组的count
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list统计count */ public void groupingByCount() { Map<String, Long> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.counting())); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之统计分组平均值
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list并计算分组年龄平均值 */ public void groupingByAverage() { Map<String, Double> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.averagingInt(Employee::getAge))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之统计分组总值
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list并计算分组销售总值 */ public void groupingBySum() { Map<String, Long> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.summingLong(Employee::getAmount))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); // 对Map按照分组销售总值逆序排序 Map<String, Long> sortedMap = new LinkedHashMap<>(); map.entrySet().stream().sorted(Map.Entry.<String, Long> comparingByValue().reversed()) .forEachOrdered(e -> sortedMap.put(e.getKey(), e.getValue())); sortedMap.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之Join分组List
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list并通过join操作连接分组list中的对象的name 属性使用逗号分隔 */ public void groupingByString() { Map<String, String> map = employees.stream().collect(Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.joining(", ", "Names: [", "]")))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之转换分组结果List -> List
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list,将List转化为name的List */ public void groupingByList() { Map<String, List<String>> map = employees.stream().collect( Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.toList()))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.stream().forEach(item -> { System.out.println("item = " + item); }); }); }
Collectors.groupingBy() 分组之转换分组结果List -> Set
功能代码:
/** * 使用java8 stream groupingBy操作,按城市分组list,将List转化为name的Set */ public void groupingBySet() { Map<String, Set<String>> map = employees.stream().collect( Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.toSet()))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.stream().forEach(item -> { System.out.println("item = " + item); }); }); }
Collectors.groupingBy() 分组之使用对象分组List
功能代码:
/** * 使用java8 stream groupingBy操作,通过Object对象的成员分组List */ public void groupingByObject() { Map<Manage, List<Employee>> map = employees.stream().collect(Collectors.groupingBy(item -> { return new Manage(item.getName()); })); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); }
Collectors.groupingBy() 分组之使用两个成员分组List
功能代码:
/** * 使用java8 stream groupingBy操作, 基于city 和name 实现多次分组 */ public void groupingBys() { Map<String, Map<String, List<Employee>>> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.groupingBy(Employee::getName))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.forEach((i, j) -> { System.out.println(i + " = " + j); }); }); }
Collectors.groupingBy() 分组之按月分组
//根据日期字段的 yyyy-MM 进行分组 Map<String, List<SomeEntity>> monthMap = someEntityList.stream().collect(Collectors.groupingBy(p -> cn.hutool.core.date.DateUtil.format(p.getOrderTime(), "yyyy-MM")));
自定义Distinct对结果去重
功能代码
/** * 使用java8 stream groupingBy操作, 基于Distinct 去重数据 */ public void groupingByDistinct() { List<Employee> list = employees.stream().filter(distinctByKey(Employee :: getCity)) .collect(Collectors.toList());; list.stream().forEach(item->{ System.out.println("city = " + item.getCity()); }); } /** * 自定义重复key 规则 * @param keyExtractor * @return */ private static <T> Predicate<T> distinctByKey(Function<? super T, ?> keyExtractor) { Set<Object> seen = ConcurrentHashMap.newKeySet(); return t -> seen.add(keyExtractor.apply(t)); }
完整源代码:
package com.stream; import java.util.ArrayList; import java.util.Arrays; import java.util.LinkedHashMap; import java.util.List; import java.util.Map; import java.util.Random; import java.util.Set; import java.util.concurrent.ConcurrentHashMap; import java.util.function.Function; import java.util.function.Predicate; import java.util.stream.Collectors; /** * Java 8 Stream 之groupingBy 分组讲解 * * @author zzg * */ public class Java8GroupBy { List<Employee> employees = new ArrayList<Employee>(); /** * 数据初始化 */ public void init() { List<String> citys = Arrays.asList("湖南", "湖北", "江西", "广西 "); for (int i = 0; i < 10; i++) { Random random = new Random(); Integer index = random.nextInt(4); Employee employee = new Employee(citys.get(index), "姓名" + i, (random.nextInt(4) * 10 - random.nextInt(4)), (random.nextInt(4) * 1000 - random.nextInt(4))); employees.add(employee); } } /** * 使用java8 stream groupingBy操作,按城市分组list */ public void groupingByCity() { Map<String, List<Employee>> map = employees.stream().collect(Collectors.groupingBy(Employee::getCity)); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作,按城市分组list统计count */ public void groupingByCount() { Map<String, Long> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.counting())); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作,按城市分组list并计算分组年龄平均值 */ public void groupingByAverage() { Map<String, Double> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.averagingInt(Employee::getAge))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作,按城市分组list并计算分组销售总值 */ public void groupingBySum() { Map<String, Long> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.summingLong(Employee::getAmount))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); // 对Map按照分组销售总值逆序排序 Map<String, Long> sortedMap = new LinkedHashMap<>(); map.entrySet().stream().sorted(Map.Entry.<String, Long> comparingByValue().reversed()) .forEachOrdered(e -> sortedMap.put(e.getKey(), e.getValue())); sortedMap.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作,按城市分组list并通过join操作连接分组list中的对象的name 属性使用逗号分隔 */ public void groupingByString() { Map<String, String> map = employees.stream().collect(Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.joining(", ", "Names: [", "]")))); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作,按城市分组list,将List转化为name的List */ public void groupingByList() { Map<String, List<String>> map = employees.stream().collect( Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.toList()))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.stream().forEach(item -> { System.out.println("item = " + item); }); }); } /** * 使用java8 stream groupingBy操作,按城市分组list,将List转化为name的Set */ public void groupingBySet() { Map<String, Set<String>> map = employees.stream().collect( Collectors.groupingBy(Employee::getCity, Collectors.mapping(Employee::getName, Collectors.toSet()))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.stream().forEach(item -> { System.out.println("item = " + item); }); }); } /** * 使用java8 stream groupingBy操作,通过Object对象的成员分组List */ public void groupingByObject() { Map<Manage, List<Employee>> map = employees.stream().collect(Collectors.groupingBy(item -> { return new Manage(item.getName()); })); map.forEach((k, v) -> { System.out.println(k + " = " + v); }); } /** * 使用java8 stream groupingBy操作, 基于city 和name 实现多次分组 */ public void groupingBys() { Map<String, Map<String, List<Employee>>> map = employees.stream() .collect(Collectors.groupingBy(Employee::getCity, Collectors.groupingBy(Employee::getName))); map.forEach((k, v) -> { System.out.println(k + " = " + v); v.forEach((i, j) -> { System.out.println(i + " = " + j); }); }); } /** * 使用java8 stream groupingBy操作, 基于Distinct 去重数据 */ public void groupingByDistinct() { List<Employee> list = employees.stream().filter(distinctByKey(Employee :: getCity)) .collect(Collectors.toList());; list.stream().forEach(item->{ System.out.println("city = " + item.getCity()); }); } /** * 自定义重复key 规则 * @param keyExtractor * @return */ private static <T> Predicate<T> distinctByKey(Function<? super T, ?> keyExtractor) { Set<Object> seen = ConcurrentHashMap.newKeySet(); return t -> seen.add(keyExtractor.apply(t)); } public static void main(String[] args) { // TODO Auto-generated method stub Java8GroupBy instance = new Java8GroupBy(); instance.init(); instance.groupingByCity(); instance.groupingByCount(); instance.groupingByAverage(); instance.groupingBySum(); instance.groupingByString(); instance.groupingByList(); instance.groupingBySet(); instance.groupingByObject(); instance.groupingBys(); instance.groupingByDistinct(); } class Employee { private String city; private String name; private Integer age; private Integer amount; public String getCity() { return city; } public void setCity(String city) { this.city = city; } public String getName() { return name; } public void setName(String name) { this.name = name; } public Integer getAge() { return age; } public void setAge(Integer age) { this.age = age; } public Integer getAmount() { return amount; } public void setAmount(Integer amount) { this.amount = amount; } public Employee(String city, String name, Integer age, Integer amount) { super(); this.city = city; this.name = name; this.age = age; this.amount = amount; } public Employee() { super(); } } class Manage { private String name; public String getName() { return name; } public void setName(String name) { this.name = name; } public Manage(String name) { super(); this.name = name; } public Manage() { super(); } } }
标签:Stream,stream,Collectors,map,--,groupingBy,分组,public From: https://www.cnblogs.com/jasonBin/p/17410880.html