【问题标题】:Java 8 Streams: How to call once the Collection.stream() method and retrieve an array of several aggregate values with different fieldsJava 8 Streams:如何调用一次 Collection.stream() 方法并检索具有不同字段的多个聚合值的数组
【发布时间】:2016-01-05 21:49:33
【问题描述】:

我从 Java 8 中的 Stream API 开始。

这是我使用的 Person 对象:

public class Person {

    private String firstName;
    private String lastName;
    private int age;
    private double height;
    private double weight;

    public Person(String firstName, String lastName, int age, double height, double weight) {
      this.firstName = firstName;
      this.lastName = lastName;
      this.age = age;
      this.height = height;
      this.weight = weight;
    }

    public String getFirstName() {
      return firstName;
    }
    public String getLastName() {
      return lastName;
    }
    public int getAge() {
      return age;
    }
    public double getHeight() {
      return height;
    }
    public double getWeight() {
      return weight;
    }

  }

这是我的代码,它初始化一个对象列表 Person 并获取按特定名字过滤的对象数量、最大年龄和最小高度、平均体重,最后创建一个包含这些值的对象数组:

List<Person> personsList = new ArrayList<Person>();

personsList.add(new Person("John", "Doe", 25, 1.80, 80));
personsList.add(new Person("Jane", "Doe", 30, 1.69, 60));
personsList.add(new Person("John", "Smith", 35, 174, 70));

long count = personsList.stream().filter(p -> p.getFirstName().equals("John")).count();
int maxAge = personsList.stream().mapToInt(Person::getAge).max().getAsInt();
double minHeight = personsList.stream().mapToDouble(Person::getHeight).min().getAsDouble();
double avgWeight = personsList.stream().mapToDouble(Person::getWeight).average().getAsDouble();

Object[] result = new Object[] { count, maxAge, minHeight, avgWeight };
System.out.println(Arrays.toString(result));

是否可以一次调用stream() 方法并直接返回对象数组?

Object[] result = personsList.stream()...count()...max()...min()...average()

我之前问过非常相似的问题:Java 8 Streams: How to call once the Collection.stream() method and retrieve an array of several aggregate values 但这次我不能使用summaryStatistics() 方法,因为我使用不同的字段(年龄、身高、体重)来检索聚合值。


编辑 2016-01-07

我测试了TriCoreTagir Valeev的解决方案,并计算了每个解决方案的运行时间。

看来TriCore的方案比Tagir Valeev的效率更高。

Tagir Valeev 的解决方案与我的解决方案(使用多个流)相比似乎并没有节省太多时间。

这是我的测试类:

public class StreamTest {

  public static class Person {

    private String firstName;
    private String lastName;
    private int age;
    private double height;
    private double weight;

    public Person(String firstName, String lastName, int age, double height, double weight) {
      this.firstName = firstName;
      this.lastName = lastName;
      this.age = age;
      this.height = height;
      this.weight = weight;
    }

    public String getFirstName() {
      return firstName;
    }

    public String getLastName() {
      return lastName;
    }

    public int getAge() {
      return age;
    }

    public double getHeight() {
      return height;
    }

    public double getWeight() {
      return weight;
    }

  }

  public static abstract class Process {

    public void run() {
      StopWatch timer = new StopWatch();
      timer.start();
      doRun();
      timer.stop();
      System.out.println(timer.getTime());
    }

    protected abstract void doRun();

  }

  public static void main(String[] args) {
    List<Person> personsList = new ArrayList<Person>();

    for (int i = 0; i < 1000000; i++) {
      int age = random(15, 60);
      double height = random(1.50, 2.00);
      double weight = random(50.0, 100.0);
      personsList.add(new Person(randomString(10, Mode.ALPHA), randomString(10, Mode.ALPHA), age, height, weight));
    }

    personsList.add(new Person("John", "Doe", 25, 1.80, 80));
    personsList.add(new Person("Jane", "Doe", 30, 1.69, 60));
    personsList.add(new Person("John", "Smith", 35, 174, 70));
    personsList.add(new Person("John", "T", 45, 179, 99));

    // Query with mutiple Streams
    new Process() {
      protected void doRun() {
        queryJava8(personsList);
      }
    }.run();

    // Query with 'TriCore' method
    new Process() {
      protected void doRun() {
        queryJava8_1(personsList);
      }
    }.run();

    // Query with 'Tagir Valeev' method
    new Process() {
      protected void doRun() {
        queryJava8_2(personsList);
      }
    }.run();
  }

  // --------------------
  // JAVA 8
  // --------------------

  private static void queryJava8(List<Person> personsList) {
    long count = personsList.stream().filter(p -> p.getFirstName().equals("John")).count();
    int maxAge = personsList.stream().mapToInt(Person::getAge).max().getAsInt();
    double minHeight = personsList.stream().mapToDouble(Person::getHeight).min().getAsDouble();
    double avgWeight = personsList.stream().mapToDouble(Person::getWeight).average().getAsDouble();

    Object[] result = new Object[] { count, maxAge, minHeight, avgWeight };
    System.out.println("Java8: " + Arrays.toString(result));

  }

  // --------------------
  // JAVA 8_1 - TriCore
  // --------------------

  private static void queryJava8_1(List<Person> personsList) {
    Object[] objects = personsList.stream().collect(Collector.of(() -> new PersonStatistics(p -> p.getFirstName().equals("John")),
        PersonStatistics::accept, PersonStatistics::combine, PersonStatistics::toStatArray));
    System.out.println("Java8_1: " + Arrays.toString(objects));
  }

  public static class PersonStatistics {
    private long firstNameCounter;
    private int maxAge = Integer.MIN_VALUE;
    private double minHeight = Double.MAX_VALUE;
    private double totalWeight;
    private long total;
    private final Predicate<Person> firstNameFilter;

    public PersonStatistics(Predicate<Person> firstNameFilter) {
      Objects.requireNonNull(firstNameFilter);
      this.firstNameFilter = firstNameFilter;
    }

    public void accept(Person p) {
      if (this.firstNameFilter.test(p)) {
        firstNameCounter++;
      }

      this.maxAge = Math.max(p.getAge(), maxAge);
      this.minHeight = Math.min(p.getHeight(), minHeight);
      this.totalWeight += p.getWeight();
      this.total++;
    }

    public PersonStatistics combine(PersonStatistics personStatistics) {
      this.firstNameCounter += personStatistics.firstNameCounter;
      this.maxAge = Math.max(personStatistics.maxAge, maxAge);
      this.minHeight = Math.min(personStatistics.minHeight, minHeight);
      this.totalWeight += personStatistics.totalWeight;
      this.total += personStatistics.total;

      return this;
    }

    public Object[] toStatArray() {
      return new Object[] { firstNameCounter, maxAge, minHeight, total == 0 ? 0 : totalWeight / total };
    }
  }

  // --------------------
  // JAVA 8_2 - Tagir Valeev
  // --------------------

  private static void queryJava8_2(List<Person> personsList) {
    // @formatter:off
    Collector<Person, ?, Object[]> collector = multiCollector(
            filtering(p -> p.getFirstName().equals("John"), Collectors.counting()),
            Collectors.collectingAndThen(Collectors.mapping(Person::getAge, Collectors.maxBy(Comparator.naturalOrder())), Optional::get),
            Collectors.collectingAndThen(Collectors.mapping(Person::getHeight, Collectors.minBy(Comparator.naturalOrder())), Optional::get),
            Collectors.averagingDouble(Person::getWeight)
    );
    // @formatter:on

    Object[] result = personsList.stream().collect(collector);
    System.out.println("Java8_2: " + Arrays.toString(result));
  }

  /**
   * Returns a collector which combines the results of supplied collectors
   * into the Object[] array.
   */
  @SafeVarargs
  public static <T> Collector<T, ?, Object[]> multiCollector(Collector<T, ?, ?>... collectors) {
    @SuppressWarnings("unchecked")
    Collector<T, Object, Object>[] cs = (Collector<T, Object, Object>[]) collectors;
    // @formatter:off
      return Collector.<T, Object[], Object[]> of(
          () -> Stream.of(cs).map(c -> c.supplier().get()).toArray(),
          (acc, t) -> IntStream.range(0, acc.length).forEach(
              idx -> cs[idx].accumulator().accept(acc[idx], t)),
          (acc1, acc2) -> IntStream.range(0, acc1.length)
              .mapToObj(idx -> cs[idx].combiner().apply(acc1[idx], acc2[idx])).toArray(),
          acc -> IntStream.range(0, acc.length)
              .mapToObj(idx -> cs[idx].finisher().apply(acc[idx])).toArray());
     // @formatter:on
  }

  /**
   * filtering() collector (which will be added in JDK-9, see JDK-8144675)
   */
  public static <T, A, R> Collector<T, A, R> filtering(Predicate<? super T> filter, Collector<T, A, R> downstream) {
    BiConsumer<A, T> accumulator = downstream.accumulator();
    Set<Characteristics> characteristics = downstream.characteristics();
    return Collector.of(downstream.supplier(), (acc, t) -> {
      if (filter.test(t))
        accumulator.accept(acc, t);
    } , downstream.combiner(), downstream.finisher(), characteristics.toArray(new Collector.Characteristics[characteristics.size()]));
  }

  // --------------------
  // HELPER METHODS
  // --------------------

  public static enum Mode {
    ALPHA,
    ALPHANUMERIC,
    NUMERIC
  }

  private static String randomString(int length, Mode mode) {
    StringBuffer buffer = new StringBuffer();
    String characters = "";

    switch (mode) {
      case ALPHA:
        characters = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ";
        break;

      case ALPHANUMERIC:
        characters = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890";
        break;

      case NUMERIC:
        characters = "1234567890";
        break;
    }

    int charactersLength = characters.length();

    for (int i = 0; i < length; i++) {
      double index = Math.random() * charactersLength;
      buffer.append(characters.charAt((int) index));
    }
    return buffer.toString();
  }

  private static int random(int min, int max) {
    Random rand = new Random();
    return rand.nextInt((max - min) + 1) + min;
  }

  private static double random(double min, double max) {
    return min + Math.random() * (max - min);
  }

}

【问题讨论】:

  • 您的Object[] result 应包含哪些值?如果我有像p1(21, 180, 80), p2(30, 178, 100), p3(40, 168, 70) 这样的人,结果应该是什么?
  • @NicolasDosSantos:答案是否定的,这是不可能的。您可能能够将这些操作中的几个合并为一个Collector,但对于这么多操作,您最好只是吸收它并按照自己的方式进行操作在这里做。
  • @LouisWasserman:你甚至没有尝试!当然有可能。要么使用库,要么自己编写Collector-combining 胶水代码

标签: java java-8 java-stream


【解决方案1】:

这里是收集器

public class PersonStatistics {
    private long firstNameCounter;
    private int maxAge = Integer.MIN_VALUE;
    private double minHeight = Double.MAX_VALUE;
    private double totalWeight;
    private long total;
    private final Predicate<Person> firstNameFilter;

    public PersonStatistics(Predicate<Person> firstNameFilter) {
        Objects.requireNonNull(firstNameFilter);
        this.firstNameFilter = firstNameFilter;
    }

    public void accept(Person p) {
        if (this.firstNameFilter.test(p)) {
            firstNameCounter++;
        }

        this.maxAge = Math.max(p.getAge(), maxAge);
        this.minHeight = Math.min(p.getHeight(), minHeight);
        this.totalWeight += p.getWeight();
        this.total++;
    }

    public PersonStatistics combine(PersonStatistics personStatistics) {
        this.firstNameCounter += personStatistics.firstNameCounter;
        this.maxAge = Math.max(personStatistics.maxAge, maxAge);
        this.minHeight = Math.min(personStatistics.minHeight, minHeight);
        this.totalWeight += personStatistics.totalWeight;
        this.total += personStatistics.total;

        return this;
    }

    public Object[] toStatArray() {
        return new Object[]{firstNameCounter, maxAge, minHeight, total == 0 ? 0 : totalWeight / total};
    }
}

您可以按如下方式使用此收集器

public class PersonMain {
    public static void main(String[] args) {
        List<Person> personsList = new ArrayList<>();

        personsList.add(new Person("John", "Doe", 25, 180, 80));
        personsList.add(new Person("Jane", "Doe", 30, 169, 60));
        personsList.add(new Person("John", "Smith", 35, 174, 70));
        personsList.add(new Person("John", "T", 45, 179, 99));

        Object[] objects = personsList.stream().collect(Collector.of(
                () -> new PersonStatistics(p -> p.getFirstName().equals("John")),
                PersonStatistics::accept,
                PersonStatistics::combine,
                PersonStatistics::toStatArray));
        System.out.println(Arrays.toString(objects));
    }
}

【讨论】:

    【解决方案2】:

    使用标准 JDK 8 API 解决这个问题有点棘手,它没有提供很多组合 Collector 类型的方法。如果你愿意使用像jOOλ 这样的第三方库,你可以写:

    Tuple4<Long, Optional<Integer>, Optional<Double>, Optional<Double>> result =
    Seq.seq(personsList)
       .collect(
           filter(p -> p.getFirstName().equals("John"), count()),
           max(Person::getAge),
           min(Person::getHeight),
           avg(Person::getWeight)
       );
    
    System.out.println(result);
    

    以上产量:

    (2, Optional[35], Optional[1.8], Optional[75.0])
    

    注意,它使用了新的Agg.filter() 方法,类似于JDK 9 的Collectors.filtering() 方法,工作原理如下:

    public static <T, A, R> Collector<T, A, R> filter(
        Predicate<? super T> predicate, Collector<T, A, R> downstream) {
        return Collector.of(
            downstream.supplier(),
            (c, t) -> {
                if (predicate.test(t))
                    downstream.accumulator().accept(c, t);
            }, 
            downstream.combiner(),
            downstream.finisher()
        );
    } 
    

    collect(collector1, collector2, ...) 是如何工作的?

    如果不想使用上述第三方库,可以自己编写Collector组合实用程序。将两个收集器组合成一个Tuple2 收集器的示例:

    static <T, A1, A2, D1, D2> Collector<T, Tuple2<A1, A2>, Tuple2<D1, D2>> collectors(
        Collector<T, A1, D1> collector1
      , Collector<T, A2, D2> collector2
    ) {
        return Collector.<T, Tuple2<A1, A2>, Tuple2<D1, D2>>of(
            () -> tuple(
                collector1.supplier().get()
              , collector2.supplier().get()
            ),
            (a, t) -> {
                collector1.accumulator().accept(a.v1, t);
                collector2.accumulator().accept(a.v2, t);
            },
            (a1, a2) -> tuple(
                collector1.combiner().apply(a1.v1, a2.v1)
              , collector2.combiner().apply(a1.v2, a2.v2)
            ),
            a -> tuple(
                collector1.finisher().apply(a.v1)
              , collector2.finisher().apply(a.v2)
            )
        );
    }
    

    免责声明:我为 jOOλ 背后的公司工作。

    【讨论】:

    • 实际上,如果没有filtering 收集器,这个答案是错误的,因为它过滤然后收集到少数收集器,而不是在一个(计数)收集器中收集和过滤。
    • @Xaerxess:你说得对,非常感谢。修复了答案(had to add a new feature for this!
    【解决方案3】:

    如果没有第三方库,您可以创建一个通用收集器,它将任意数量的指定收集器的结果组合到单个 Object[] 数组中:

    /**
     * Returns a collector which combines the results of supplied collectors
     * into the Object[] array.
     */
    @SafeVarargs
    public static <T> Collector<T, ?, Object[]> multiCollector(
            Collector<T, ?, ?>... collectors) {
        @SuppressWarnings("unchecked")
        Collector<T, Object, Object>[] cs = (Collector<T, Object, Object>[]) collectors;
        return Collector.<T, Object[], Object[]> of(
            () -> Stream.of(cs).map(c -> c.supplier().get()).toArray(),
            (acc, t) -> IntStream.range(0, acc.length).forEach(
                idx -> cs[idx].accumulator().accept(acc[idx], t)),
            (acc1, acc2) -> IntStream.range(0, acc1.length)
                .mapToObj(idx -> cs[idx].combiner().apply(acc1[idx], acc2[idx])).toArray(),
            acc -> IntStream.range(0, acc.length)
                .mapToObj(idx -> cs[idx].finisher().apply(acc[idx])).toArray());
    }
    

    对于您的具体问题,您还需要一个 filtering() 收集器(将在 JDK-9 中添加,请参阅 JDK-8144675):

    public static <T, A, R> Collector<T, A, R> filtering(
            Predicate<? super T> filter, Collector<T, A, R> downstream) {
        BiConsumer<A, T> accumulator = downstream.accumulator();
        Set<Characteristics> characteristics = downstream.characteristics();
        return Collector.of(downstream.supplier(), (acc, t) -> {
            if(filter.test(t)) accumulator.accept(acc, t);
        }, downstream.combiner(), downstream.finisher(), 
            characteristics.toArray(new Collector.Characteristics[characteristics.size()]));
    }
    

    现在您可以构建一个收集器来生成最终结果:

    Collector<Person, ?, Object[]> collector = 
        multiCollector(
            filtering(p -> p.getFirstName().equals("John"), counting()),
            collectingAndThen(mapping(Person::getAge, 
                maxBy(Comparator.naturalOrder())), Optional::get),
            collectingAndThen(mapping(Person::getHeight, 
                minBy(Comparator.naturalOrder())), Optional::get),
            averagingDouble(Person::getWeight));
    
    Object[] result = personsList.stream().collect(collector);
    System.out.println(Arrays.toString(result));
    

    【讨论】:

      猜你喜欢
      • 1970-01-01
      • 1970-01-01
      • 1970-01-01
      • 2020-12-08
      • 1970-01-01
      • 1970-01-01
      • 2020-01-18
      • 1970-01-01
      • 2015-04-15
      相关资源
      最近更新 更多