Cover by Alan Gates

Stay ahead with the world's most comprehensive technology and business learning platform.

With Safari, you learn the way you learn best. Get unlimited access to videos, live online training, learning paths, books, tutorials, and more.

Start Free Trial

No credit card required

O'Reilly logo

User Defined Functions

Much of the power of Pig lies in its ability to let users combine irs operators with their own or others’ code via UDFs. Up through version 0.7, all UDFs must be written in Java and are implemented as Java classes.[13] This makes it very easy to add new UDFs to Pig by writing a Java class and telling Pig about your JAR file.

As of version 0.8, UDFs can also be written in Python. Pig uses Jython to execute Python UDFs, so they must be compatible with Python 2.5 and cannot use Python 3 features.

Pig itself comes packaged with some UDFs. Prior to version 0.8, this was a very limited set, including only the standard SQL aggregate functions and a few others. In 0.8, a large number of standard string-processing, math, and complex-type UDFs were added. For a complete list and description of built-in UDFs, see Built-in UDFs.

Piggybank is a collection of user-contributed UDFs that is packaged and released along with Pig. Piggybank UDFs are not included in the Pig JAR, and thus you have to register them manually in your script. See Piggybank for more information.

Of course you can also write your own UDFs or use those written by other users. For details of how to write your own, see Chapter 10. Finally, you can use some static Java functions as UDFs as well.

Registering UDFs

When you use a UDF that is not already built into Pig, you have to tell Pig where to look for that UDF. This is done via the register command. For example, let’s say you want to use the Reverse UDF provided in Piggybank (for information on where to find the Piggybank JAR, see Piggybank):

register 'your_path_to_piggybank/piggybank.jar';
divs      = load 'NYSE_dividends' as (exchange:chararray, symbol:chararray,
                date:chararray, dividends:float);
backwards = foreach divs generate

This example tells Pig that it needs to include code from your_path_to_piggybank/piggybank.jar when it produces a JAR to send to Hadoop. Pig opens all of the registered JARs, takes out the files, and places them in the JAR that it sends to Hadoop to run your jobs.

In this example, we have to give Pig the full package and class name of the UDF. This verbosity can be alleviated in two ways. The first option is to use the define command (see define and UDFs). The second option is to include a set of paths on the command line for Pig to search when looking for UDFs. So if instead of invoking Pig as pig register.pig we change our invocation to pig -Dudf.import.list=org.apache.pig.piggybank.evaluation.string register.pig, we can change our script to:

register 'your_path_to_piggybank/piggybank.jar';
divs      = load 'NYSE_dividends' as (exchange:chararray, symbol:chararray,
                date:chararray, dividends:float);
backwards = foreach divs generate Reverse(symbol);

Using yet another property, we can get rid of the register command as well. If we add -Dpig.additional.jars=/usr/local/pig/piggybank/piggybank.jar to our command line, the register command is no longer necessary.

In many cases it is better to deal with registration and definition issues explicitly in the script via the register and define commands than use these properties. Otherwise, everyone who runs your script has to know how to configure the command line. However, in some situations your scripts will always use the same set of JARs and always look in the same places for them. For instance, you might have a set of JARs used by everyone in your company. In this case, placing these properties in a shared properties file and using that with your Pig scripts will make sharing those UDFs easier and assure that everyone is using the correct versions of them.

In 0.8 and later versions, the register command can also take HDFS paths. If your JARs are stored in HDFS, you could then say register 'hdfs://user/jar/acme.jar';. Starting in 0.9, register accepts globs. So if all of the JARs you need are stored in one directory, you could include them all with register '/usr/local/share/pig/udfs/*.jar'.

Registering Python UDFs

register is also used to locate resources for Python UDFs that you use in your Pig Latin scripts. In this case you do not register a JAR, but rather a Python script that contains your UDF. The Python script must be in your current directory. Using the examples provided in the example code, copying udfs/python/ to the data directory looks like this:

register '' using jython as bballudfs;
players  = load 'baseball' as (name:chararray, team:chararray,
                pos:bag{t:(p:chararray)}, bat:map[]);
nonnull  = filter players by bat#'slugging_percentage' is not null and
                bat#'on_base_percentage' is not null;
calcprod = foreach nonnull generate name, bballudfs.production(

The important differences here are the using jython and as bballudfs portions of the register statement. using jython tells Pig that this UDF is written in Python, not Java, and it should use Jython to compile that UDF. Pig does not know where on your system the Jython interpreter is, so you must include jython.jar in your classpath when invoking Pig. This can be done by setting the PIG_CLASSPATH environment variable.

as bballudfs defines a namespace that UDFs from this file are placed in. All UDFs from this file must now be invoked as bballudfs.udfname. Each Python file you load should be given a separate namespace. This avoids naming collisions when you register two Python scripts with duplicate function names.

One caveat: Pig does not trace dependencies inside your Python scripts and send the needed Python modules to your Hadoop cluster. You are required to make sure the modules you need reside on the task nodes in your cluster and that the PYTHONPATH environment variable is set on those nodes such that your UDFs will be able to find them for import. This issue has been fixed after 0.9, but as of this writing is not yet released.

define and UDFs

As was alluded to earlier, define can be used to provide an alias so that you do not have to use full package names for your Java UDFs. It can also be used to provide constructor arguments to your UDFs. define also is used in defining streaming commands, but this section covers only its UDF-related features. For information on using define with streaming, see stream. The following provides an example of using define to provide an alias for org.apache.pig.piggybank.evaluation.string.Reverse:

register 'your_path_to_piggybank/piggybank.jar';
define reverse org.apache.pig.piggybank.evaluation.string.Reverse();
divs      = load 'NYSE_dividends' as (exchange:chararray, symbol:chararray,
                date:chararray, dividends:float);
backwards = foreach divs generate reverse(symbol);

Eval and filter functions can also take one or more strings as constructor arguments. If you are using a UDF that takes constructor arguments, define is the place to provide those arguments. For example, consider a method CurrencyConverter that takes two constructor arguments, the first indicating which currency you are converting from and the second which currency you are converting to:

register 'acme.jar';
define convert'dollar', 'euro');
divs      = load 'NYSE_dividends' as (exchange:chararray, symbol:chararray,
                date:chararray, dividends:float);
backwards = foreach divs generate convert(dividends);

Calling Static Java Functions

Java has a rich collection of utilities and libraries. Because Pig is implemented in Java, some of these functions can be exposed to Pig users. Starting in version 0.8, Pig offers invoker methods that allow you to treat certain static Java functions as if they were Pig UDFs.

Any public static Java function that takes no arguments or some combination of int, long, float, double, String, or arrays thereof,[14] and returns int, long, float, double, or String can be invoked in this way.

Because Pig Latin does not support overloading on return types, there is an invoker for each return type: InvokeForInt, InvokeForLong, InvokeForFloat, InvokeForDouble, and InvokeForString. You must pick the appropriate invoker for the type you wish to return. This method takes two constructor arguments. The first is the full package, classname, and method name. The second is a space-separated list of parameters the Java function expects. Only the types of the parameters are given. If the parameter is an array, [] (square brackets) are appended to the type name. If the method takes no parameters, the second constructor argument is omitted.

For example, if you wanted to use Java’s Integer class to translate decimal values to hexadecimal values, you could do:

define hex InvokeForString('java.lang.Integer.toHexString', 'int');
divs  = load 'NYSE_daily' as (exchange, symbol, date, open, high, low,
            close, volume, adj_close);
nonnull = filter divs by volume is not null;
inhex = foreach nonnull generate symbol, hex((int)volume);

If your method takes an array of types, Pig will expect to pass it a bag where each tuple has a single field of that type. So if you had a Java method com.yourcompany.Stats.stdev that took an array of doubles, you could use it like this:

define stdev InvokeForDouble('com.acme.Stats.stdev', 'double[]');
A = load 'input' as (id: int, dp:double);
B = group A by id;
C = foreach B generate group, stdev(A.dp);


Invokers do not use the Accumulator or Algebraic interfaces, and are thus likely to be much slower and to use much more memory than UDFs written specifically for Pig. This means that before you pass an array argument to an invoked method, you should think carefully about whether those inefficiencies are acceptable. For more information on these interfaces, see Accumulator Interface and Algebraic Interface.

Invoking Java functions in this way does have a small cost because reflection is used to find and invoke the methods.

Invoker functions throw Java an IllegalArgumentException when they are passed null input. You should place a filter before the invocation to prevent this.

[13] This is why UDF names are case-sensitive in Pig.

[14] For int, long, float, and double, invoker methods can call Java functions that take the scalar types but not the associated Java classes (so int but not Integer, etc.).

With Safari, you learn the way you learn best. Get unlimited access to videos, live online training, learning paths, books, interactive tutorials, and more.

Start Free Trial

No credit card required