Fork me on GitHub

Extending Pebble

Extending Pebble


Pebble was designed to be flexible and accomodate the requirements of any project. You can add your own tags, functions, operators, filters, tests, and global variables. The majority of these are quite trivial to implement.

Begin by creating a class that implements Extension. For your own convenience, I recommend extending AbstractExtension if you can. After implementing the required methods, register your extension with the PebbleEngine before compiling any templates:

PebbleEngine engine = new PebbleEngine.Builder().extension(new CustomExtension()).build();


To create custom filters, implement the getFilters() method of your extension which will return a map of filter names and their corresponding implementations. A filter implementation must implement the Filter interface. The Filter interface requires two methods to be implemented, getArgumentNames() and apply(). The getArgumentNames() method returns a list of Strings that define both the order and names of expected arguments.

The apply method is the actual filter implementation. Here's the parameters definition.

Parameter nameDescription
inputthe data to be filtered
argsthe map of arguments the user may have provided
selfAn instance of PebbleTemplate which can be used to retrieve the template name for example
contextAn instance of EvaluationContext which can be used to retrieve the locale for example
lineNumberUseful when throwing exception to provide line number

Because Pebble is dynamically typed, you will have to downcast the arguments to the expected type. Here is an example of how the upper filter might be implemented:

public class UpperFilter implements Filter {

	public List<String> getArgumentNames() {
		return null;

	public Object apply(Object input, Map<String, Object> args, PebbleTemplate self, EvaluationContext context, int lineNumber){
		if(input == null){
			return null;
		if (input instanceof String) {
			return ((String) input).toUpperCase(context.getLocale());
		} else {
			return input.toString().toUpperCase(context.getLocale());



Adding custom tests is very similar to custom filters. Implement the getTests() method within your extension which will return a map of test names and their corresponding implementations. A test implementation will implement the Test interface. The Test interface is exactly like the Filter interface except the apply method returns a boolean instead of an arbitrary object of any type.

Here is an example of how the even test might be implemented:

public class EvenTest implements Test {

	public List<String> getArgumentNames() {
		return null;

	public boolean apply(Object input, Map<String, Object> args, PebbleTemplate self, EvaluationContext context, int lineNumber){
		if (input == null) {
			throw new PebbleException(null, "Can not pass null value to \"even\" test.", lineNumber, self.getName());
		if (input instanceof Integer) {
			return ((Integer) input) % 2 == 0;
		} else {
			return ((Long) input) % 2 == 0;



Adding functions is also very similar to custom filters. First and foremost, it's important to understand the different intentions behind a function and a filter because it can often be ambiguous which one should be implemented. A filter is intended to modify existing content where a function is moreso intended to produce new content.

To add functions, implement the getFunctions() method within your extension which will return a map of function names and their corresponding implementations. A function implementation will implement the Function interface. The Function interface is very similar to the Filter and Test interfaces.

Here is an example of how a fictional fibonacciString function might be implemented:

public class FibonnaciStringFunction implements Function {

	public List<String> getArgumentNames() {
		List<String> names = new ArrayList<>();
		return names;

	public Object execute(Map<String, Object> args, PebbleTemplate self, EvaluationContext context, int lineNumber) {
		Integer length = (Integer)args.get("length");
		Integer prev1 = 0;
		Integer prev2 = 1;

		StringBuilder result = new StringBuilder();


		for(int i = 2; i < length; i++){
			Integer next = prev1 + prev2;
			prev1 = prev2;
			prev2 = next;
		return result.toString();


Positional and Named Arguments

For filters, tests, and functions it is required that you implement the getArgumentNames method even if it returns null. Returning a list of strings will allow the end user to call your filter/test/function using named arguments. Using the above fictional fibonacci function as an example, a user can invoke it in two different ways:

{{ fibonacci(10) }}
{{ fibonacci(length=10) }}

If the end user excludes the names and only uses positional arguments, the argument values will still end up be mapped to the proper names when it's time to invoke the function's execute method. Your function implementation doesn't have to worry whether the user used positional or named arguments. It is important though that if the filter/function/test expects more than one argument, then the developer must communicate to the user the expected order of arguments in the chance that the user wants to invoke it without using names.

Some functions such as the built in min and max functions accept an unlimited amount of arguments. For this to happen, your function must not accept any named arguments (i.e. your getArgumentNames method will return null or empty) and your `execute`` method will simply iterate over the values of the user provided argument map while ignoring the keys of that map (Pebble will use arbitrary keys if there are no names to map to).

Global Variables

Adding global variables, which are variables that are accessbile to all templates, is very trivial. In your custom extension, implement the getGlobalVariables() method which returns a Map<String,Object>. The contents of this map will be merged into the context you provide to each template at the time of rendering.


Operators are more complex to implement than filters or tests. To add custom operators, implement the getBinaryOperators() or the getUnaryOperators() method in your extension, or both. These methods return a list of BinaryOperator or UnaryOperator objects, respectively.

Binary operators require the following information:

  • Precedence: an integer relative to other operators which defines the order of operations.
  • Symbol: a String representing the actual operator. This is typically a single character but doesn't have to be.
  • Expression Class: A class that extends BinaryExpression. This class will perform the actual operator implementation.
  • Associativity: Either left or right depending on how the operator is used.

A unary operator is much the same except it's expression class must extend UnaryExpression and there is no associativity.

The precedence values for existing core operators are as followed:

  • or: 10
  • and: 15
  • is: 20
  • is not: 20
  • ==: 30
  • !=: 30
  • >: 30
  • <: 30
  • >=: 30
  • <=: 30
  • +: 40
  • -: 40
  • not: 50 (Unary)
  • *: 60
  • /: 60
  • %: 60
  • |: 100
  • +: 500 (Unary)
  • -: 500 (Unary)

The following is an example of how the addition operator (+) might have been implemented:

public class AdditionOperator implements BinaryOperator {

	public int getPrecedence(){
		return 30;

	public String getSymbol(){
		return "+";

    public BinaryExpression<?> createInstance() {
        return new AddExpression();

    public BinaryOperatorType getType() {
        return BinaryOperatorType.NORMAL;

	public Associativity getAssociativity(){
		return Associativity.LEFT;


Alongside each operator class you will also need to implement a corresponding BinaryExpression class which actually implements the operator. The above example references a fictional AdditionExpression class which might look like the following:

public class AdditionExpression extends BinaryExpression<Object> {

	public Object evaluate(PebbleTemplateImpl self, EvaluationContext context){
		Integer left = (Integer)getLeftExpression().evaluate(self, context);
		Integer right = (Integer)getRightExpression().evaluate(self, context);

		return left + right;


In the above example you will notice that children of BinaryExpression have access to two other expressions, leftExpression, and rightExpression; these are the operands of your operator. Please note that in the above example both operands are casted to Integers but in reality you can't always make that assumption; the true addition expression is much more complex to handle different types of operands (Integers, Longs, Doubles, etc).


Creating new tags is one of the most powerful abilities of Pebble. Your extension should start by implementing the getTokenParsers() method. A TokenParser is responsible for converting all necessary tokens to appropriate RenderableNodes. A token is a significant and irreducible group of characters found in a template (such as an operator, whitespace, variable name, delimiter, etc) and a RenderableNode is a Pebble class that is responsible for generating output.

Let us look at an example of a TokenParser:

public class SetTokenParser implements TokenParser {

	public String getTag(){
		return "set";

	public RenderableNode parse(Token token, Parser parser) {
		TokenStream stream = parser.getStream();
		int lineNumber = token.getLineNumber();

		// skip the "set" token;

		// use the built in expression parser to parse the variable name
		String name = parser.getExpressionParser().parseNewVariableName();

		stream.expect(Token.Type.PUNCTUATION, "=");

		// use the built in expression parser to parse the variable value
		Expression<?> value = parser.getExpressionParser().parseExpression();

		// expect to see "%}"

		// NodeSet is composed of a name and a value
		return new SetNode(lineNumber, name, value);


The getTag() method must return the name of the tag. Pebble's main parser will use this name to determine when to delegate responsibility to your custom TokenParser. This example is parsing the set tag.

The parse method is invoked whenever the primary parser encounters a set token. This method should return one RenderableNode instance which when rendered during the template evaluation, will write output to the provided Writer object. If the RenderableNode contains children nodes, it should invoke the render method of those nodes as well.

The best way to learn all the details of parsing is to look at some of the tools used, as well as some examples. Here is a list of classes I suggest reading:

  • TokenParser
  • Parser
  • SetTokenParser
  • ForTokenParser
  • IfNode
  • SetNode

Attribute resolver (v3 only)

To create a new attribute resolver, implement the getAttributeResolver() method of your extension which will return a list of attribute resolvers to run. A attribute resolver implementation must implement the AttributeResolver interface. The AttributeResolver interface requires one method to be implemented, resolve().

The custom attribute resolver will be executed before all default pebble attribute resolvers. It replaces the DynamicAttributeProvider interface

public class DefaultAttributeResolver implements AttributeResolver {

  public ResolvedAttribute resolve(Object instance,
                                   Object attributeNameValue,
                                   Object[] argumentValues,
                                   boolean isStrictVariables,
                                   String filename,
                                   int lineNumber) {
    if (instance instanceof CustomObject) {
      return "customValue";
    return null;