@Description(name = "letters", value = "_FUNC_(expr) - 返回该列中所有字符串的字符总数")
public class TotalNumOfLettersGenericUDAF extends AbstractGenericUDAFResolver {

    @Override
    public GenericUDAFEvaluator getEvaluator(TypeInfo[] parameters)
            throws SemanticException {
        if (parameters.length != 1) {
            throw new UDFArgumentTypeException(parameters.length - 1,
                    "Exactly one argument is expected.");
        }
        
        ObjectInspector oi = TypeInfoUtils.getStandardJavaObjectInspectorFromTypeInfo(parameters[0]);
        
        if (oi.getCategory() != ObjectInspector.Category.PRIMITIVE){
            throw new UDFArgumentTypeException(0,
                            "Argument must be PRIMITIVE, but "
                            + oi.getCategory().name()
                            + " was passed.");
        }
        
        PrimitiveObjectInspector inputOI = (PrimitiveObjectInspector) oi;
        
        if (inputOI.getPrimitiveCategory() != PrimitiveObjectInspector.PrimitiveCategory.STRING){
            throw new UDFArgumentTypeException(0,
                            "Argument must be String, but "
                            + inputOI.getPrimitiveCategory().name()
                            + " was passed.");
        }
        
        return new TotalNumOfLettersEvaluator();
    }

    public static class TotalNumOfLettersEvaluator extends GenericUDAFEvaluator {

        PrimitiveObjectInspector inputOI;
        ObjectInspector outputOI;
        PrimitiveObjectInspector integerOI;
        
        int total = 0;

        @Override
        public ObjectInspector init(Mode m, ObjectInspector[] parameters)
                throws HiveException {
        	
            assert (parameters.length == 1);
            super.init(m, parameters);
           
             //map阶段读取sql列，输入为String基础数据格式
            if (m == Mode.PARTIAL1 || m == Mode.COMPLETE) {
                inputOI = (PrimitiveObjectInspector) parameters[0];
            } else {
			//其余阶段，输入为Integer基础数据格式
            	integerOI = (PrimitiveObjectInspector) parameters[0];
            }

             // 指定各个阶段输出数据格式都为Integer类型
            outputOI = ObjectInspectorFactory.getReflectionObjectInspector(Integer.class,
                    ObjectInspectorOptions.JAVA);
            return outputOI;

        }

        /**
         * 存储当前字符总数的类
         */
        static class LetterSumAgg implements AggregationBuffer {
            int sum = 0;
            void add(int num){
            	sum += num;
            }
        }

        @Override
        public AggregationBuffer getNewAggregationBuffer() throws HiveException {
            LetterSumAgg result = new LetterSumAgg();
            return result;
        }

        @Override
        public void reset(AggregationBuffer agg) throws HiveException {
        	LetterSumAgg myagg = new LetterSumAgg();
        }
        
        private boolean warned = false;

        @Override
        public void iterate(AggregationBuffer agg, Object[] parameters)
                throws HiveException {
            assert (parameters.length == 1);
            if (parameters[0] != null) {
                LetterSumAgg myagg = (LetterSumAgg) agg;
                Object p1 = ((PrimitiveObjectInspector) inputOI).getPrimitiveJavaObject(parameters[0]);
                myagg.add(String.valueOf(p1).length());
            }
        }

        @Override
        public Object terminatePartial(AggregationBuffer agg) throws HiveException {
            LetterSumAgg myagg = (LetterSumAgg) agg;
            total += myagg.sum;
            return total;
        }

        @Override
        public void merge(AggregationBuffer agg, Object partial)
                throws HiveException {
            if (partial != null) {
                
                LetterSumAgg myagg1 = (LetterSumAgg) agg;
                
                Integer partialSum = (Integer) integerOI.getPrimitiveJavaObject(partial);
                
                LetterSumAgg myagg2 = new LetterSumAgg();
                
                myagg2.add(partialSum);
                myagg1.add(myagg2.sum);
            }
        }

        @Override
        public Object terminate(AggregationBuffer agg) throws HiveException {
            LetterSumAgg myagg = (LetterSumAgg) agg;
            total = myagg.sum;
            return myagg.sum;
        }

    }
}

代码说明

AggregationBuffer 允许我们保存中间结果，通过定义我们的buffer，我们可以处理任何格式的数据，在代码例子中字符总数保存在AggregationBuffer 。

/**
* 保存当前字符总数的类
*/
static class LetterSumAgg implements AggregationBuffer {
	int sum = 0;
	void add(int num){
		sum += num;
	}
}

这意味着UDAF在不同的mapreduce阶段会接收到不同的输入。Iterate读取我们表中的一行（或者准确来说是表），然后输出其他数据格式的聚集结果。

artialAggregation合并这些聚集结果到另外相同格式的新的聚集结果，然后最终的reducer取得这些聚集结果然后输出最终结果（该结果或许与接收数据的格式不一致）。

在init()方法中我们指定输入为string，结果输出格式为integer，还有，部分聚集结果输出格式为integer（保存在aggregation buffer中）；terminate()与terminatePartial()两者输出一个integer。

// init方法中根据不同的mode指定输出数据的格式objectinspector
if (m == Mode.PARTIAL1 || m == Mode.COMPLETE) {
	inputOI = (PrimitiveObjectInspector) parameters[0];
} else {
	integerOI = (PrimitiveObjectInspector) parameters[0];
}

// 不同model阶段的输出数据格式
outputOI = ObjectInspectorFactory.getReflectionObjectInspector(Integer.class,
                    ObjectInspectorOptions.JAVA);

iterate()函数读取到每行中列的字符串，计算与保存该字符串的长度

public void iterate(AggregationBuffer agg, Object[] parameters)
	throws HiveException {
	...
	Object p1 = ((PrimitiveObjectInspector) inputOI).getPrimitiveJavaObject(parameters[0]);
	myagg.add(String.valueOf(p1).length());
	}
}

Merge函数增加部分聚集总数到AggregationBuffer

public void merge(AggregationBuffer agg, Object partial)
      	throws HiveException {
	if (partial != null) {
                
		LetterSumAgg myagg1 = (LetterSumAgg) agg;
                
		Integer partialSum = (Integer) integerOI.getPrimitiveJavaObject(partial);
                
		LetterSumAgg myagg2 = new LetterSumAgg();
                
		myagg2.add(partialSum);
		myagg1.add(myagg2.sum);
	}
}

Terminate()函数返回AggregationBuffer中的内容，这里产生了最终结果。

public Object terminate(AggregationBuffer agg) throws HiveException {
	LetterSumAgg myagg = (LetterSumAgg) agg;
	total = myagg.sum;
	return myagg.sum;
}

使用自定义函数

ADD JAR ./hive-extension-examples-master/target/hive-extensions-1.0-SNAPSHOT-jar-with-dependencies.jar;
CREATE TEMPORARY FUNCTION letters as ‘com.matthewrathbone.example.TotalNumOfLettersGenericUDAF‘;

SELECT letters(name) FROM people;
OK
44
Time taken: 20.688 seconds