MapReduce 中的 Reducer 值需要复制，否则需要修改？

Question

在 MapReduce 应用程序中，我有一个名为 AnonymousPair 的任意 WritableComparable 实现，我注意到了这一点

import com.google.common.collect.MinMaxPriorityQueue;

public static class MyReducer extends Reducer<LongWritable, AnonymousPair, LongWritable, Text> {
    @Override
    protected void reduce(LongWritable key, Iterable<AnonymousPair> values, Context context) throws IOException, InterruptedException { 
        // ...
        MinMaxPriorityQueue<AnonymousPair> pQueue = MinMaxPriorityQueue
                .orderedBy(new AnonymousPair().comparator())
                .maximumSize(Constants.MaxKeywords)
                .create();

        for(AnonymousPair val : values) {
            pQueue.add(new AnonymousPair(val)); // No problem with copy constructor
            // pQueue.add(val);                 // Wrong! Every element in pQueue will be the same 
        }
    }
}

如果我不使用 'copy constructor'，pQueue 中的每个元素最终都会相同。谁能帮我理解这个？谢谢！我的猜测是

对可能会被修改的 Reducer Values 元素的引用。它在文档中的某个地方，但我错过了。
我错误地使用了 Google Guava MinMaxPriorityQueue
或者我的 WritableComparable 实现有问题

我的 AnonymousPair 实现

    public static class AnonymousPair implements WritableComparable<AnonymousPair> {
        private String a = "";
        private Float b = 0f;
        public AnonymousPair() {}
        public AnonymousPair(String a, Float b) {this.a = a; this.b = b;}
        public AnonymousPair(AnonymousPair o) {this.a = o.a; this.b = o.b;}

        public Comparator<AnonymousPair> comparator() {return new AnonymousPairComparator();}

        class AnonymousPairComparator implements Comparator<AnonymousPair> {
            @Override
            public int compare(AnonymousPair o1, AnonymousPair o2) {
                Float diff = o1.b - o2.b;
                if(diff == 0) {
                    return 0;
                }
                if(diff < 0) {
                    return 1;   // Reverse order
                } else {
                    return -1;
                }

            }
        }

        @Override
        public int compareTo(AnonymousPair o) {
            int temp = this.a.compareTo(o.a);
            if(temp == 0) {
                return -this.b.compareTo(o.b);
            } else {
                return temp;
            }
        }

        // More overriding...
    }

Answer 1

见javadoc:

The framework will reuse the key and value objects that are passed into the reduce, therefore the application should clone the objects they want to keep a copy of. In many cases, all values are combined into zero or one value.

MapReduce 中的 Reducer 值需要复制，否则需要修改？

Reducer values in MapReduce need to be copied, otherwise subject to modification?

java

hadoop

mapreduce