需求：将表中数据按照name聚合，并且count进行累加

name	count
Jan	1
Jan	2
Feb	3
Feb	1
Mar	1
Mar	5

预期结果：

name	count
Jan	3
Feb	7
Mar	13

使用idea的maxCompute studio新增UDAF

然后自动生成未实现的方法，我的字段name是string，count是bigint
所以@Resolve("string,bigint->bigint")

新增一个class用来存储字段

        private String  name;
        private Long count;
        @Override
        public void write(DataOutput out) throws IOException {
            out.writeUTF(name);
            out.writeLong(count);
        }

        @Override
        public void readFields(DataInput in) throws IOException {
            name =  in.readUTF();
            count = in.readLong();
        }
    }

这样newBuffer就可以这样写了

    public Writable newBuffer() {
        return new MyBuffer();
    }

还需要一个Map来存储key（name）和value（count），一个long类型的参数存储累加的值

    Long old_count = 0L;//存储累加值
    private LongWritable ret = new LongWritable();//存储输出值

完整代码参考：

public class UDAFTest extends Aggregator {
    private static class MyBuffer implements Writable {
        private String  name;
        private Long count;
        @Override
        public void write(DataOutput out) throws IOException {
            out.writeUTF(name);
            out.writeLong(count);
        }

        @Override
        public void readFields(DataInput in) throws IOException {
            name =  in.readUTF();
            count = in.readLong();
        }
    }



    @Override
    public Writable newBuffer() {
        return new MyBuffer();
    }
    Map<String,Long> map = new LinkedHashMap<>();
    Long old_count = 0L;
    @Override
    public void iterate(Writable buffer, Writable[] args) throws UDFException {
        String arg = String.valueOf(args[0]);
        Long cnt = Long.parseLong(String.valueOf(args[1]));
        MyBuffer buf = (MyBuffer) buffer;

        if (arg != null) {
            if(map.containsKey(arg)){
                Long newcnt = map.get(arg);
                old_count = cnt+newcnt;
                map.put(arg,old_count);
            }else {
                map.put(arg,old_count+cnt);
            }
        }
        buf.name = arg;
        buf.count = map.get(arg);

    }
    private LongWritable ret = new LongWritable();
    @Override
    public Writable terminate(Writable arg0) throws UDFException {
        MyBuffer buffer = (MyBuffer) arg0;
        ret.set(buffer.count);
        return ret;
    }

    @Override
    public void merge(Writable buffer, Writable partial) throws UDFException {
        MyBuffer buf = (MyBuffer) buffer;
        MyBuffer p = (MyBuffer) partial;
        buf.name = p.name;
        buf.count = p.count;
    }


}

然后通过maxCompute studio发布下

发布名为test20191119，这样就可以在Dataworks中调用了。

其中原表数据：