Template Engines
Jul 5, 2014
See also


jmustache-1.13 (Aug 8, 2016)
Oct 24, 2010
Jan 12, 2024
Michael Bayne (samskivert)
Martin Ellis (martiell)
Jeroen van Erp (hierynomus)
Robby Morgan (robbytx)
Tom Insam (tominsam)
Mickael Jeanroy (mjeanroy)
David Hoover (deadmoose)
Serkan Kaba (skaba)
Georgios Valotasios (valotas)
Michael Bayne (samskivert)
Henrik Huttunen (egaga)
Daniel Bruckner (bruckner)
John Montgomery (johnsensible)
Sean Scanlon (sps)
Whymarrh Whitby (whymarrh)
Jacob Lewallen (jlewallen)
Alexander Schrab (meros)
Show all (24)24
Source code

This is a Java implementation of the Mustache template language.


  • Zero dependencies. You can include this single tiny library in your project and start making use of templates.

  • Usability on a variety of target platforms. This implementation makes very limited demands on the JVM in which it runs and as a result is usable on Android, or on other limited JVMs. It is even possible to avoid the use of reflection and provide all of your data as a series of nested maps.

  • Proguard and JarJar friendly. Though the library will reflectively access your data (if you desire it), the library makes no other internal use of reflection or by name instantiation of classes. So you can embed it using Proguard or JarJar without any annoying surprises.

  • Minimal API footprint. There are really only two methods you need to know about: compile and execute. You can even chain them together in cases where performance is of no consequence.

Its existence justified by the above motivations, this implementation then strives to provide additional benefits:

  • It is available via Maven Central, see below for details.
  • It is reasonably performant. Templates are parsed separately from execution. A template will specialize its variables on (class of context, name) pairs so that if a variable is first resolved to be (for example) a field of the context object, that will be attempted directly on subsequent template invocations, and the slower full resolution will only be tried if accessing the variable as a field fails.

Get It

JMustache is available via Maven Central and can thus be easily added to your Maven, Ivy, etc. projects by adding a dependency on com.samskivert:jmustache:1.15. Or download the pre-built jar file.


In addition to the usage section below, the following documentation may be useful:


Using JMustache is very simple. Supply your template as a String or a Reader and get back a Template that you can execute on any context:

String text = "One, two, {{three}}. Three sir!";
Template tmpl = Mustache.compiler().compile(text);
Map<String, String> data = new HashMap<String, String>();
data.put("three", "five");
// result: "One, two, five. Three sir!"

Use Reader and Writer if you're doing something more serious:

void executeTemplate (Reader template, Writer out, Map<String, String> data) {
   Mustache.compiler().compile(template).execute(data, out);

The execution context can be any Java object. Variables will be resolved via the following mechanisms:

  • If the context is a MustacheCustomContext, MustacheCustomContext.get will be used.
  • If the context is a Map, Map.get will be used.
  • If a non-void method with the same name as the variable exists, it will be called.
  • If a non-void method named (for variable foo) getFoo exists, it will be called.
  • If a field with the same name as the variable exists, its contents will be used.


class Person {
    public final String name;
    public Person (String name, int age) { = name;
        _age = age;
    public int getAge () {
        return _age;
    protected int _age;

String tmpl = "{{#persons}}{{name}}: {{age}}\n{{/persons}}";
Mustache.compiler().compile(tmpl).execute(new Object() {
    Object persons = Arrays.asList(new Person("Elvis", 75), new Person("Madonna", 52));

// result:
// Elvis: 75
// Madonna: 52

As you can see from the example, the fields (and methods) need not be public. The persons field in the anonymous class created to act as a context is accessible. Note that the use of non-public fields will not work in a sandboxed security environment.

Sections behave as you would expect:

  • Boolean values enable or disable the section.
  • Array, Iterator, or Iterable values repeatedly execute the section with each element used as the context for each iteration. Empty collections result in zero instances of the section being included in the template.
  • An unresolvable or null value is treated as false. This behavior can be changed by using strictSections(). See Default Values for more details.
  • Any other object results in a single execution of the section with that object as a context.

See the code in for concrete examples. See also the Mustache documentation for details on the template syntax.


If you wish to make use of partials (e.g. {{>subtmpl}}) you must provide a Mustache.TemplateLoader to the compiler when creating it. For example:

final File templateDir = ...;
Mustache.Compiler c = Mustache.compiler().withLoader(new Mustache.TemplateLoader() {
    public Reader getTemplate (String name) {
        return new FileReader(new File(templateDir, name));
String tmpl = "...{{>subtmpl}}...";

The above snippet will load new File(templateDir, "subtmpl") when compiling the template.


JMustache implements lambdas by passing you a Template.Fragment instance which you can use to execute the fragment of the template that was passed to the lambda. You can decorate the results of the fragment execution, as shown in the standard Mustache documentation on lambdas:

String tmpl = "{{#bold}}{{name}} is awesome.{{/bold}}";
Mustache.compiler().compile(tmpl).execute(new Object() {
   String name = "Willy";
   Mustache.Lambda bold = new Mustache.Lambda() {
        public void execute (Template.Fragment frag, Writer out) throws IOException {
// result:
<b>Willy is awesome.</b>

You can also obtain the results of the fragment execution to do things like internationalization or caching:

Object ctx = new Object() {
    Mustache.Lambda i18n = new Mustache.Lambda() {
         public void execute (Template.Fragment frag, Writer out) throws IOException {
             String key = frag.execute();
             String text = // look up key in i18n system
// template might look something like:

There is also limited support for decompiling (unexecuting) the template and obtaining the original Mustache template text contained in the section. See the documentation for Template.Fragment for details on the limitations.

Default Values

By default, an exception will be thrown any time a variable cannot be resolved, or resolves to null (except for sections, see below). You can change this behavior in two ways. If you want to provide a value for use in all such circumstances, use defaultValue():

String tmpl = "{{exists}} {{nullValued}} {{doesNotExist}}?";
Mustache.compiler().defaultValue("what").compile(tmpl).execute(new Object() {
    String exists = "Say";
    String nullValued = null;
    // String doesNotExist
// result:
Say what what?

If you only wish to provide a default value for variables that resolve to null, and wish to preserve exceptions in cases where variables cannot be resolved, use nullValue():

String tmpl = "{{exists}} {{nullValued}} {{doesNotExist}}?";
Mustache.compiler().nullValue("what").compile(tmpl).execute(new Object() {
    String exists = "Say";
    String nullValued = null;
    // String doesNotExist
// throws MustacheException when executing the template because doesNotExist cannot be resolved

When using a Map as a context, nullValue() will only be used when the map contains a mapping to null. If the map lacks a mapping for a given variable, then it is considered unresolvable and throws an exception.

Map<String,String> map = new HashMap<String,String>();
map.put("exists", "Say");
map.put("nullValued", null);
// no mapping exists for "doesNotExist"
String tmpl = "{{exists}} {{nullValued}} {{doesNotExist}}?";
// throws MustacheException when executing the template because doesNotExist cannot be resolved

Do not use both defaultValue and nullValue in your compiler configuration. Each one overrides the other, so whichever one you call last is the behavior you will get. But even if you accidentally do the right thing, you have confusing code, so don't call both, use one or the other.


Sections are not affected by the nullValue() or defaultValue() settings. Their behavior is governed by a separate configuration: strictSections().

By default, a section that is not resolvable or which resolves to null will be omitted (and conversely, an inverse section that is not resolvable or resolves to null will be included). If you use strictSections(true), sections that refer to an unresolvable value will always throw an exception. Sections that refer to a resolvable but null value never throw an exception, regardless of the strictSections() setting.


JMustache extends the basic Mustache template language with some additional functionality. These additional features are enumerated below:

Not escaping HTML by default

You can change the default HTML escaping behavior when obtaining a compiler:

Mustache.compiler().escapeHTML(false).compile("{{foo}}").execute(new Object() {
    String foo = "<bar>";
// result: <bar>
// not: &lt;bar&gt;

User-defined object formatting

By default, JMustache uses String.valueOf to convert objects to strings when rendering a template. You can customize this formatting by implementing the Mustache.Formatter interface:

Mustache.compiler().withFormatter(new Mustache.Formatter() {
    public String format (Object value) {
      if (value instanceof Date) return _fmt.format((Date)value);
      else return String.valueOf(value);
    protected DateFormat _fmt = new SimpleDateFormat("yyyy/MM/dd");
}).compile("{{msg}}: {{today}}").execute(new Object() {
    String msg = "Date";
    Date today = new Date();
// result: Date: 2013/01/08

User-defined escaping rules

You can change the escaping behavior when obtaining a compiler, to support file formats other than HTML and plain text.

If you only need to replace fixed strings in the text, you can use Escapers.simple:

String[][] escapes = {{ "[", "[[" }, { "]", "]]" }};
  compile("{{foo}}").execute(new Object() {
      String foo = "[bar]";
// result: [[bar]]

Or you can implement the Mustache.Escaper interface directly for more control over the escaping process.

Special variables


You can use the special variable this to refer to the context object itself instead of one of its members. This is particularly useful when iterating over lists.

Mustache.compiler().compile("{{this}}").execute("hello"); // returns: hello
Mustache.compiler().compile("{{#names}}{{this}}{{/names}}").execute(new Object() {
    List<String> names () { return Arrays.asList("Tom", "Dick", "Harry"); }
// result: TomDickHarry

Note that you can also use the special variable . to mean the same thing.

Mustache.compiler().compile("{{.}}").execute("hello"); // returns: hello
Mustache.compiler().compile("{{#names}}{{.}}{{/names}}").execute(new Object() {
    List<String> names () { return Arrays.asList("Tom", "Dick", "Harry"); }
// result: TomDickHarry

. is apparently supported by other Mustache implementations, though it does not appear in the official documentation.

-first and -last

You can use the special variables -first and -last to perform special processing for list elements. -first resolves to true when inside a section that is processing the first of a list of elements. It resolves to false at all other times. -last resolves to true when inside a section that is processing the last of a list of elements. It resolves to false at all other times.

One will often make use of these special variables in an inverted section, as follows:

String tmpl = "{{#things}}{{^-first}}, {{/-first}}{{this}}{{/things}}";
Mustache.compiler().compile(tmpl).execute(new Object() {
    List<String> things = Arrays.asList("one", "two", "three");
// result: one, two, three

Note that the values of -first and -last refer only to the inner-most enclosing section. If you are processing a section within a section, there is no way to find out whether you are in the first or last iteration of an outer section.


The -index special variable contains 1 for the first iteration through a section, 2 for the second, 3 for the third and so forth. It contains 0 at all other times. Note that it also contains 0 for a section that is populated by a singleton value rather than a list.

String tmpl = "My favorite things:\n{{#things}}{{-index}}. {{this}}\n{{/things}}";
Mustache.compiler().compile(tmpl).execute(new Object() {
    List<String> things = Arrays.asList("Peanut butter", "Pen spinning", "Handstands");
// result:
// My favorite things:
// 1. Peanut butter
// 2. Pen spinning
// 3. Handstands

Compound variables

In addition to resolving simple variables using the context, you can use compound variables to extract data from sub-objects of the current context. For example:

Mustache.compiler().compile("Hello {{field.who}}!").execute(new Object() {
    public Object field = new Object() {
        public String who () { return "world"; }
// result: Hello world!

By taking advantage of reflection and bean-property-style lookups, you can do kooky things:

Mustache.compiler().compile("Hello {{}}!").execute(new Object());
// result: Hello java.lang.Object!

Note that compound variables are essentially short-hand for using singleton sections. The above examples could also be represented as:

Hello {{#field}}{{who}}{{/field}}!
Hello {{#class}}{{name}}{{/class}}!

Note also that one semantic difference exists between nested singleton sections and compound variables: after resolving the object for the first component of the compound variable, parent contexts will not be searched when resolving subcomponents.

Newline trimming

If the opening or closing section tag are the only thing on a line, any surrounding whitespace and the line terminator following the tag are trimmed. This allows for civilized templates, like:

Favorite foods:
  <li>{{first_name}} {{last_name}} likes {{favorite_food}}.</li>

which produces output like:

Favorite foods:
  <li>Elvis Presley likes peanut butter.</li>
  <li>Mahatma Gandhi likes aloo dum.</li>

rather than:

Favorite foods:

  <li>Elvis Presley likes peanut butter.</li>

  <li>Mahatma Gandhi likes aloo dum.</li>


which would be produced without the newline trimming.

Nested Contexts

If a variable is not found in a nested context, it is resolved in the next outer context. This allows usage like the following:

String template = "{{outer}}:\n{{#inner}}{{outer}}.{{this}}\n{{/inner}}";
Mustache.compiler().compile(template).execute(new Object() {
    String outer = "foo";
    List<String> inner = Arrays.asList("bar", "baz", "bif");
// results:
// foo:
// foo.baz
// foo.bif

Note that if a variable is defined in an inner context, it shadows the same name in the outer context. There is presently no way to access the variable from the outer context.

Invertible Lambdas

For some applications, it may be useful for lambdas to be executed for an inverse section rather than having the section omitted altogether. This allows for proper conditional substitution when statically translating templates into other languages or contexts:

String template = "{{#condition}}result if true{{/condition}}\n" +
  "{{^condition}}result if false{{/condition}}";
Mustache.compiler().compile(template).execute(new Object() {
    Mustache.InvertibleLambda condition = new Mustache.InvertibleLambda() {
        public void execute (Template.Fragment frag, Writer out) throws IOException {
            // this method is executed when the lambda is referenced in a normal section
            out.write("if (condition) {console.log(\"");
        public void executeInverse (Template.Fragment frag, Writer out) throws IOException {
            // this method is executed when the lambda is referenced in an inverse section
            out.write("if (!condition) {console.log(\"");
        private String toJavaScriptLiteral (String execute) {
            // note: this is NOT a complete implementation of JavaScript string literal escaping
            return execute.replaceAll("\\\\", "\\\\\\\\").replaceAll("\"", "\\\\\"");
// results:
// if (condition) {console.log("result if true")}
// if (!condition) {console.log("result if false")}

Of course, you are not limited strictly to conditional substitution -- you can use an InvertibleLambda whenever you need a single function with two modes of operation.

Standards Mode

The more intrusive of these extensions, specifically the searching of parent contexts and the use of compound variables, can be disabled when creating a compiler, like so:

Map<String,String> ctx = new HashMap<String,String>();
ctx.put("", "baz");
// result: baz

Thread Safety

JMustache is internally thread safe with the following caveats:

  • Compilation: compiling templates calls out to a variety of helper classes: Mustache.Formatter, Mustache.Escaper, Mustache.TemplateLoader, Mustache.Collector. The default implementations of these classes are thread-safe, but if you supply custom instances, then you have to ensure that your custom instances are thread-safe.

  • Execution: executing templates can call out to some helper classes: Mustache.Lambda, Mustache.VariableFetcher. The default implementations of these classes are thread-safe, but if you supply custom instances, then you have to ensure that your custom instances are thread-safe.

  • Context data: if you mutate the context data passed to template execution while the template is being executed, then you subject yourself to race conditions. It is in theory possible to use a thread-safe map (ConcurrentHashMap or Collections.synchronizedMap) for your context data, which would allow you to mutate the data while templates were being rendered based on that data, but you're playing with fire by doing that. I don't recommend it. If your data is supplied as POJOs where fields or methods are called via reflection to populate your templates, volatile fields and synchronized methods could similarly be used to support simultaneous reading and mutating, but again you could easily make a mistake that introduces race conditions or cause weirdness when executing your templates. The safest approach when rendering the same template via simultaneous threads is to pass immutable/unchanging data as the context for each execution.

  • VariableFetcher cache: template execution uses one internal cache to store resolved VariableFetcher instances (because resolving a variable fetcher is expensive). This cache is thread-safe by virtue of using a ConcurrentHashMap. It's possible for a bit of extra work to be done if two threads resolve the same variable at the same time, but they won't conflict with one another, they'll simply both resolve the variable instead of one resolving the variable and the other using the cached resolution.

So the executive summary is: as long as all helper classes you supply are thread-safe (or you use the defaults), it is safe to share a Mustache.Compiler instance across threads to compile templates. If you pass immutable data to your templates when executing, it is safe to have multiple threads simultaneously execute a single Template instance.


In the name of simplicity, some features of Mustache were omitted or simplified:

  • {{= =}} only supports one or two character delimiters. This is just because I'm lazy and it simplifies the parser.