java以元素频率顺序遍历多集的最简单方法？

Question

考虑这个例子，打印出一些设备类型统计信息。（“DeviceType”是具有dozenish值的枚举。）

Multiset<DeviceType> histogram = getDeviceStats();
for (DeviceType type : histogram.elementSet()) {
    System.out.println(type + ": " + histogram.count(type));
}

按频率顺序打印不同元素的最简单、最优雅的方法是什么（最常见的类型优先）

快速查看一下^{}接口，没有现成的方法，而且Guava的Multiset实现（HashMultiset，TreeMultiset等）似乎也没有自动保持元素的频率顺序

Answer 1

我刚刚在番石榴中添加了这个特性，请参见here了解Javadoc

编辑：根据原始问题Multisets.copyHighestCountFirst()的用法示例：

Multiset<DeviceType> histogram = getDeviceStats();
for (DeviceType type : Multisets.copyHighestCountFirst(histogram).elementSet()) {
    System.out.println(type + ": " + histogram.count(type));
}

Answer 2

使用ForwardingMultiSet的实现：

（EntryComp来自seanizer'sanswer）

enum EntryComp implements Comparator<Multiset.Entry<?>> {
    DESCENDING {
        @Override
        public int compare(final Entry<?> a, final Entry<?> b) {
            return Ints.compare(b.getCount(), a.getCount());
        }
    },
    ASCENDING {
        @Override
        public int compare(final Entry<?> a, final Entry<?> b) {
            return Ints.compare(a.getCount(), b.getCount());
        }
    },
}

public class FreqSortMultiSet<E> extends ForwardingMultiset<E> {
    Multiset<E> delegate;
    EntryComp comp;

    public FreqSortMultiSet(Multiset<E> delegate, boolean ascending) {
        this.delegate = delegate;
        if (ascending)
            this.comp = EntryComp.ASCENDING;
        else
            this.comp = EntryComp.DESCENDING;
    }

    @Override
    protected Multiset<E> delegate() {
        return delegate;
    }

    @Override
    public Set<Entry<E>> entrySet() {
        TreeSet<Entry<E>> sortedEntrySet = new TreeSet<Entry<E>>(comp);
        sortedEntrySet.addAll(delegate.entrySet());
        return sortedEntrySet;
    }

    @Override
    public Set<E> elementSet() {
        Set<E> sortedEntrySet = new LinkedHashSet<E>();
        for (Entry<E> en : entrySet())
            sortedEntrySet.add(en.getElement());
        return sortedEntrySet;
    }

    public static <E> FreqSortMultiSet<E> create(boolean ascending) {
        return new FreqSortMultiSet<E>(HashMultiset.<E> create(), ascending);
    }

    /*
     * For Testing
     * public static void main(String[] args) {
        Multiset<String> s = FreqSortMultiSet.create(false);
        s.add("Hello");
        s.add("Hello");
        s.setCount("World", 3);
        s.setCount("Bye", 5);
        System.out.println(s.entrySet());
    }*/

}

Answer 3

下面是一个返回List条目的方法，按频率排序（更新：使用一个标志切换升序/降序，并使用番石榴最喜欢的玩具：Enum Singleton Pattern，如Effective Java，第3项所示）：

private enum EntryComp implements Comparator<Multiset.Entry<?>>{
    DESCENDING{
        @Override
        public int compare(final Entry<?> a, final Entry<?> b){
            return Ints.compare(b.getCount(), a.getCount());
        }
    },
    ASCENDING{
        @Override
        public int compare(final Entry<?> a, final Entry<?> b){
            return Ints.compare(a.getCount(), b.getCount());
        }
    },
}

public static <E> List<Entry<E>> getEntriesSortedByFrequency(
    final Multiset<E> ms, final boolean ascending){
    final List<Entry<E>> entryList = Lists.newArrayList(ms.entrySet());
    Collections.sort(entryList, ascending
        ? EntryComp.ASCENDING
        : EntryComp.DESCENDING);
    return entryList;
}

测试代码：

final Multiset<String> ms =
    HashMultiset.create(Arrays.asList(
        "One",
        "Two", "Two",
        "Three", "Three", "Three",
        "Four", "Four", "Four", "Four"
    ));

System.out.println("ascending:");
for(final Entry<String> entry : getEntriesSortedByFrequency(ms, true)){
    System.out.println(MessageFormat.format("{0} ({1})",
        entry.getElement(), entry.getCount()));
}

System.out.println("descending:");
for(final Entry<String> entry : getEntriesSortedByFrequency(ms, false)){
    System.out.println(MessageFormat.format("{0} ({1})",
        entry.getElement(), entry.getCount()));
}

输出：

ascending:
One (1)
Two (2)
Three (3)
Four (4)
descending:
Four (4)
Three (3)
Two (2)
One (1)

Answer 4

因为it is not yet implemented，我想你可以创建一个Map，key=type，value=count。然后对地图进行排序——参见here