如何写hive的udf函数？

最近感受了hive的udf函数的强大威力了，不仅可以使用很多已经有的udf函数，还可以自己定义符合业务场景的udf函数，下面就说一下如何写udf/udaf/udtf函数，算是一个入门介绍吧。

First, you need to create a new class that extends UDF, with one or more methods named evaluate.

package com.example.hive.udf;

import org.apache.hadoop.hive.ql.exec.UDF;
import org.apache.hadoop.io.Text;

public final class Lower extends UDF {
  public Text evaluate(final Text s) {
    if (s == null) { return null; }
    return new Text(s.toString().toLowerCase());
  }
}

After compiling your code to a jar, you need to add this to the hive classpath.

add jar my_jar.jar;

Once hive is started up with your jars in the classpath, the final step is to register your function

create temporary function my_lower as 'com.example.hive.udf.Lower';

上面主要描述了实现一个udf的过程，首先自然是实现一个UDF函数，然后编译为jar并加入到hive的classpath中，最后创建一个临时变量名字让hive中调用。

下面这个表格可以更加清晰的看出udf/udaf/udtf之间的区别

Show几个例子：

1） UDF （参考：http://svn.apache.org/repos/asf/hive/trunk/contrib/src/java/org/apache/hadoop/hive/contrib/udf/）

package org.apache.hadoop.hive.contrib.udf.example;

import org.apache.hadoop.hive.ql.exec.UDF;

/**
 * UDFExampleAdd.
 *
 */
public class UDFExampleAdd extends UDF {

  public Integer evaluate(Integer... a) {
    int total = 0;
    for (Integer element : a) {
      if (element != null) {
        total += element;
      }
    }
    return total;
  }

  public Double evaluate(Double... a) {
    double total = 0;
    for (Double element : a) {
      if (element != null) {
        total += element;
      }
    }
    return total;
  }

}

2）UDAF（ http://svn.apache.org/repos/asf/hive/trunk/contrib/src/java/org/apache/hadoop/hive/contrib/udaf/）

package org.apache.hadoop.hive.contrib.udaf.example;

import org.apache.hadoop.hive.ql.exec.UDAF;
import org.apache.hadoop.hive.ql.exec.UDAFEvaluator;

/**
 * This is a simple UDAF that calculates average.
 * 
 * It should be very easy to follow and can be used as an example for writing
 * new UDAFs.
 * 
 * Note that Hive internally uses a different mechanism (called GenericUDAF) to
 * implement built-in aggregation functions, which are harder to program but
 * more efficient.
 * 
 */
public final class UDAFExampleAvg extends UDAF {

  /**
   * The internal state of an aggregation for average.
   * 
   * Note that this is only needed if the internal state cannot be represented
   * by a primitive.
   * 
   * The internal state can also contains fields with types like
   * ArrayList<String> and HashMap<String,Double> if needed.
   */
  public static class UDAFAvgState {
    private long mCount;
    private double mSum;
  }

  /**
   * The actual class for doing the aggregation. Hive will automatically look
   * for all internal classes of the UDAF that implements UDAFEvaluator.
   */
  public static class UDAFExampleAvgEvaluator implements UDAFEvaluator {

    UDAFAvgState state;

    public UDAFExampleAvgEvaluator() {
      super();
      state = new UDAFAvgState();
      init();
    }

    /**
     * Reset the state of the aggregation.
     */
    public void init() {
      state.mSum = 0;
      state.mCount = 0;
    }

    /**
     * Iterate through one row of original data.
     * 
     * The number and type of arguments need to the same as we call this UDAF
     * from Hive command line.
     * 
     * This function should always return true.
     */
    public boolean iterate(Double o) {
      if (o != null) {
        state.mSum += o;
        state.mCount++;
      }
      return true;
    }

    /**
     * Terminate a partial aggregation and return the state. If the state is a
     * primitive, just return primitive Java classes like Integer or String.
     */
    public UDAFAvgState terminatePartial() {
      // This is SQL standard - average of zero items should be null.
      return state.mCount == 0 ? null : state;
    }

    /**
     * Merge with a partial aggregation.
     * 
     * This function should always have a single argument which has the same
     * type as the return value of terminatePartial().
     */
    public boolean merge(UDAFAvgState o) {
      if (o != null) {
        state.mSum += o.mSum;
        state.mCount += o.mCount;
      }
      return true;
    }

    /**
     * Terminates the aggregation and return the final result.
     */
    public Double terminate() {
      // This is SQL standard - average of zero items should be null.
      return state.mCount == 0 ? null : Double.valueOf(state.mSum
          / state.mCount);
    }
  }

  private UDAFExampleAvg() {
    // prevent instantiation
  }

}

3）UDTF（ http://svn.apache.org/repos/asf/hive/trunk/contrib/src/java/org/apache/hadoop/hive/contrib/udtf/）

package org.apache.hadoop.hive.contrib.udtf.example;

import java.util.ArrayList;
import java.util.List;

import org.apache.hadoop.hive.ql.exec.Description;
import org.apache.hadoop.hive.ql.exec.UDFArgumentException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDTF;
import org.apache.hadoop.hive.serde2.objectinspector.ListObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;

/**
 * GenericUDTFExplode2.
 *
 */
@Description(name = "explode2",
    value = "_FUNC_(a) - like explode, but outputs two identical columns (for testing purposes)")
public class GenericUDTFExplode2 extends GenericUDTF {

  ListObjectInspector listOI = null;

  @Override
  public void close() throws HiveException {
  }

  @Override
  public StructObjectInspector initialize(ObjectInspector[] args)
      throws UDFArgumentException {

    if (args.length != 1) {
      throw new UDFArgumentException("explode() takes only one argument");
    }

    if (args[0].getCategory() != ObjectInspector.Category.LIST) {
      throw new UDFArgumentException("explode() takes an array as a parameter");
    }
    listOI = (ListObjectInspector) args[0];

    ArrayList<String> fieldNames = new ArrayList<String>();
    ArrayList<ObjectInspector> fieldOIs = new ArrayList<ObjectInspector>();
    fieldNames.add("col1");
    fieldNames.add("col2");
    fieldOIs.add(listOI.getListElementObjectInspector());
    fieldOIs.add(listOI.getListElementObjectInspector());
    return ObjectInspectorFactory.getStandardStructObjectInspector(fieldNames,
        fieldOIs);
  }

  Object forwardObj[] = new Object[2];

  @Override
  public void process(Object[] o) throws HiveException {

    List<?> list = listOI.getList(o[0]);
    for (Object r : list) {
      forwardObj[0] = r;
      forwardObj[1] = r;
      forward(forwardObj);
    }
  }

  @Override
  public String toString() {
    return "explode";
  }
}

Ref:  https://cwiki.apache.org/confluence/display/Hive/LanguageManual+UDF