How do I compile a Hive UDF

问题:

I am trying to compile this UDF:

package com.dataminelab.hive.udf;
import org.apache.hadoop.hive.ql.exec.UDF;
import org.apache.hadoop.io.Text;
import java.security.*;

/**
 * Calculate md5 of the string
*/
public final class Md5 extends UDF {

    public Text evaluate(final Text s) {
        if (s == null) {
            return null;
        }
        try {
            MessageDigest md = MessageDigest.getInstance("MD5");
            md.update(s.toString().getBytes());
            byte[] md5hash = md.digest();
            StringBuilder builder = new StringBuilder();
            for (byte b : md5hash) {
            builder.append(Integer.toString((b & 0xff) + 0x100, 16).substring(1));
            }
            return new Text(builder.toString());
            } catch (NoSuchAlgorithmException nsae) {
            System.out.println("Cannot find digest algorithm");
            System.exit(1);
        }
        return null;
    }
}

Trying to compile with:

javac Md5.java

But I get:

Md5.java:2: package org.apache.hadoop.hive.ql.exec does not exist
import org.apache.hadoop.hive.ql.exec.UDF;
                                     ^
Md5.java:3: package org.apache.hadoop.io does not exist
import org.apache.hadoop.io.Text;

I assume these are in a jar file somewhere but I'm not sure where hadoop install them to so I can't add them to my classpath. Does anyone know the default location or how to find out?

回答1:

The following works for me, but I suspect the details will vary depending on your installation and what your source code does:

export CLASSPATH=/usr/lib/hive/lib/hive-exec-0.9.0.15.jar:/usr/lib/hadoop/hadoop-core.jar

回答2:

Have you included jars of the form $HIVE_HOME/lib/hive-serde-*.jar and $HIVE_HOME/lib/hive-exec-*.jar in your classpath?

回答3:

add hive-0.4.1 jar file to lib