Skip to content

Dust Tutorial

Seth Kinast edited this page May 27, 2015 · 136 revisions

Table of Contents

The Wiki is deprecated in favor of Try the new tutorial at

Dust Tutorial

  • Richard Ragan - PayPal ( Author )
  • Veena Basavaraj - LinkedIn ( Edits )

Why JavaScript Templating?

There are a number of advantages provided by implementing templates in JavaScript with the Dust template rendering solution. They include:

  • Dust templates are compiled to JavaScript for speed of execution
  • Since JavaScript is cacheable, Dust templates can be sent to a browser once and reused which is not the case with server-side generated JSP pages
  • In addition to being cacheable, the Dust templates compiled to JavaScript can be served from a CDN for faster loading
  • No need for UI developers to start a Tomcat server to render a page served with JSON-- much faster cycle time, less issues
  • Continuing acceleration of JavaScript performance in browsers continues to make client-side work faster
  • Write once and can run the same on both the client (browser) and server using JS engine such as V8/rhino

Show me some Dust

A quick sample is often the best way to get a general sense of something. In that vein, here is a simple Dust template and it's JSON data below it




 "title": "Famous People", 
 "names" : [{ "name": "Larry" },{ "name": "Curly" },{ "name": "Moe" }]

This will output:

Famous People 

Dust templates output plain old text and processes dust tags -- {xxxxxx} being a dust tag format. The tag structure is similar to html in general form but using braces instead of <>, e.g. {name /}, {name}body inside tag{/name} and with parameters {name param="val1" param2="val",... }.

The simplest form is just

and is called a key. It references a value from the JSON data named "name". In our example, you saw the key {title} which produced an output value of "Famous People". The other tag form we saw in the example was
. This is called a section.
  • If the property is an array, it finds the "names" property from the JSON model and iterates over 1 to n times where n is the number of array elements. In our example, we looped over the three "name" values in the JSON model using the {name} key in the section body.
  • If the property exists and is a non-empty scalar, the section outputs the value of the scalar provided it is referenced in the body
  • If the property exists and is a valid object, the section outputs properties in the object if referenced by keys of the form {object.propName}
  • If the property does not exist or has an empty array, nothing from the body of the section is emitted.
As you go through this tutorial, you might get the urge to experiment. You can easily do this at

Just put your template in the upper left box, the data model in the upper right and click in the bottom right box to trigger running the template against the data. The lower left box shows the compiled JavaScript.

More on Dust Output and Dust Filters

Things worth knowing:

  • If there is no value found for a key, nothing is output
  • Template whitespace is largely discarded. See section on whitespace control for details if you want a different behavior. Take special care if you have a JavaScript code block and have comments of the form // message. When all the newlines are removed, this will comment out the following statement. Use the /* message */ form instead.
  • {! Comment syntax !} is how you write comments
  • {` Preserve all new lines, whitespace, and braces `} - Available with 2.3.3 release
  • All output values are escaped to avoid Cross Site Scripting (XSS) unless you use filters:
    • Hello {name|s} suppresses auto-escaping
    • Hello {name|h} force HTML escaping
    • Hello {name|j} force JavaScript escaping
    • Hello {name|u} encodes with JS native encodeURI
    • Hello {name|uc} encodes with JS native encodeURIComponent
    • Hello {name|js} stringify JSON literal
    • Hello {name|jp} parse JSON string to object
  • Filters can be chained – Hello {name|s|h}
  • Special characters can be escaped if you need to output them: {~n} – newline, {~r} –CR, {~lb} - left bracket, {~rb} - right bracket, {~s} - space

Dust Data and Referencing

Dust gets its data values from the JavaScript object literal used to render the template. Object literals contain three main types of data: scalars (name: xxx}, arrays: ( ["aa",] ) and objects ( name: {firstName: "Jane", lastName: "Doe" } ). The data can also be a function but we are not considering that case here.

Referencing a scalar value is done with a simple key, e.g. {name}. Individual array elements can be referenced by subscripting, e.g. array[3]. Object properties are referenced using paths, e.g name.firstName. Of course, a path reference can be to a scalar or an array which could then be subscripted.


A section is a Dust tag of the form {#names}...{/names}. It is the way you loop over data in Dust. What happens is the current context (more on context shortly) is set to the "names" part of your JSON data. If names is an array, the body wrapped by the section tag is executed for each element of the array. If the element is not an array, the body will just be executed once. If the "names" element does not exist or is empty, the body will be skipped.

During the execution of the section iteration, two variables are defined: $idx - the index of the current iteration starting with zero and $len - the number of elements in the data being iterated.

Note: $idx and $len work for arrays of primitive types as of the 1.1 release.

Sections and Context

So if I have two instances of name: value in my JSON, how does Dust decide which one to use to render {name}?

Dust has a concept of context to provide rules around how a value is found in the JSON model. When you use a section reference like {#name}....{/name} Dust sets it's context to the portion of the JSON model identified by name. When you first start rendering, the context is set to the outermost level of the JSON object. Thus the {#names} section positions the context to the block of JSON in the names: [...] part, which happens to be an array. Therefore, the {name} key in the section body is matched against the one in the context "names".

Let's explore how context works with a more complex JSON model.

   "name": "root",
   "anotherName": "root2",

As we learned earlier, if you have {#A}{name}{/A} the current context is A: and everything under it (i.e. it includes the B: stuff). The key for {name} will output "Albert" because that is the direct value of name in the context of "A".

So how does it work if you have {#A}{anotherName}{/A}? You will get "root2" as the output. That's because "anotherName" could not be found directly under "A" so Dust tries to walk up to the parent of "A" (which is the root context in our case) and finds "anotherName" hence using its value. In general, a simple key reference will look first in the current context and, if not found, search all higher levels up to the root looking for the name. It will not search downward into things like "B" that are nested within "A".


Suppose our context is the root and say we want to work with the data "only" under "B". Like in JavaScript itself, you can use a dotted notation called a "path" to do this. For example, {} will output "Bob".

Simple key references like {} are sometimes not enough. You might need to use a section to iterate over a sub-part of the JSON model. Remember when you use {#xxxx} for a section, you also establish a new context at that point. For example, in the case of {#A.B}{name}{/A.B}, This will output "Bob" because our context has been set to B within A. Path notation only allows you to reference a path visible within the current context.

You CANNOT reference a value using paths when that pathed value is outside your current context. Lets look at an example to make this point clear.

  name in B={name} name in A= {} 

The above will output "name in B=Bob name in A=" showing that is not accessible inside the context A.B. What goes on is that dust looks for the initial part of the path in the current context, e.g. A, and gives up when it cannot find it.

There is a way out of this behavior. Follow along with the below dust fragment. {#A}, is a non-pathed reference so it is allowed to search upward and find "A". Then {#A} sets a new context to "A" allowing us to reference the "name" value under "A". When the closing tag {/A} is reached, the context reverts to {#A.B}, Yes, the context acts like a stack.

IMPORTANT: While you cannot use a dotted path notation to reference ancestor/parents JSON from the current context, you can use a non-pathed section reference to adjust your context to a higher point. For example,

{#A.B} name in B={name} 
   name in A: {name} 

Another way to reference a value outside your current context is to pass the value into the section as an inline parameter (we will talk more about parameters soon). Here is an example of how to access within the {#A.B} context using a parameter on the {#A.B} section

  name in B={name} name in A: {param} 

New path behavior available as of dust 2.0.0 release

A change to the limitation of dust paths not searching outside the current context was made in the 2.0.0 release. Note that there is a very slight possible incompatibility which we will discuss in a bit.

So what does the change do? If you look at the example above (using the same data):

  name in B={name} name in A={} 

with this path change it will output "name in B=Bob name in A=Albert". The new rules for resolving path references are as follows:

  • If the path is {x.y.z} and x is not in the current context, begin looking in outer contexts for x. If found, look from there for y.z. If found, return the value. If not found, give up and return no value. Do not look further.
  • If the path is {x.y.z} and x is not found in any context, then look for the path in globals similar to the way simple key references look in globals as a final search location. The difference is that paths can now be found in globals.
Some things remain unchanged from the original dust:
  • If the path starts with a period, e.g. {.x.y.z}, the search remains restricted to the current context.
  • If the path is {x.y.z} and the entire path can be found in the current context, the value of the path will be used.
  • If an explicit context (covered in next section) is established by {name:context}, then only the context determined by name and that provided by :context are visible. You cannot escape out of these two contexts and reference anything further up the stack. This is like the original dust.
This change eliminates the need for using a {#x} sort of hack to reach a path that is in an outer context.

For those wondering about compatibility implications, they should be slight. The two main cases are:

  • A path reference that used to return no value, might start returning a value if the path can be found in an outer context of the data. The fix is to prefix the path with a period to constrain the search to the current context.
  • The code might test for the existence of data using a path to determine whether to output some conditional text. Presence or absence of the path in the data used to control this behavior. Now if the path is omitted but it can be found in outer context, the output that used to be omitted will start appearing. Like the previous case, the different output relies on the path actually appearing somewhere in an outer context. As in the first case, adding a leading period will constrain the search to just the current context restoring the old behavior.

Explicit context setting

Normally the visibility of data from the JSON model is controlled by your current context set by the # tag, or by inline parameters, plus the ability to access values by the key reference {name} and to reset the current context based on a #section reference to outer block using {#outerBlock}.

There is another way to control and limit visibility for a block of code. The notation

{#name:name2}.... {/name}
will do that.

Specifically it does the folllowing:

  • Hides all nested context levels above "name"
  • Puts "name2" data as the parent context and name as the current context
This prevents {key} references from accessing any data not in the name or name2 contexts. No further reaching up can happen even with simple key forms like {name}. This scope limitation might be useful for data hiding from components. Another use for it could be to make only the current context and it's peer available.

Given a data model where A and B are peers and we need to iterate over A and also reference data from B, without explicit context setting we would have trouble doing this.

          ["Albert", "Alan"]
However, the following:
  {#A:A2} {#names}{.} - {type} {/names} {/A}

will output "Albert - Student Alan - Student since both A and A2 are on the context stack even though A2 would not normally be there.

Sections with parameters

Since we just dropped a teaser about parameters, let's look at them. Section tags allow you to pass parameters into the section for subsequent use. Parameter values can be simple string constants or the name of a value from the data model. For example, using the same data model as earlier:

  {#A.B foo="Hi" bar=" Good to see you"}
      {foo} {name} {bar}

This will output "Hi Bob Good to see you

As we saw earlier, values from the data model can also be passed. Consider

  {#A.B bar=anotherName}
      {foo} {name} {bar}

This will output "Albert Bob root2". It's important to understand the context at the point the parameter values are established. With above, is evaluated before the context is moved to A.B, thus is accessible.

However, if the parameter values are interpolated into strings, they are evaluated in the context of the section using them. Therefore, the following will just output "Bob root2" because {} is not accessible from the {#A.B} context.

  {#A.B foo="{}" bar="{anotherName}" }
      {foo} {name} {bar}

While you can specify an object as a parameter, e.g.

  {#A.B foo=A }

you cannot do anything useful with it since the {} reference is going to look for foo in the current context but that context is the element of the current iteration of the section #A.B (in this case just the name: "Bob", value). Therefore, "foo" won't be found. The foo parameter is on the context stack but one level higher than the current element iteration so unreachable by a path reference.

When deciding on parameter names, try to be unique. Inline parameters will not override the current context if a property of the same name exists. Let's look at an example:

{#A name="Not Albert"} 
  name is {name}.

will output "name is Albert" since preference goes to data in the current context followed by inline parameters then up the context tree.

If we want to be sure we get the value in the parameter we can make it unique.

{#A paramName="Not Albert"} 
  name is {paramName} and {} is still Bob.
will output "name is Not Albert and Bob is still Bob"

Parameter rules

Since parameters are on your mind, let's discuss the three forms parameters can take.

  • param=name
  • param="xxx"
  • param="{yyy}"
In the first form, name is obtained from the context. It can be a path. In the second form, the value looks like a string and is, in fact, a string value. In the third form the value for the name yyy is obtained from the context and interpolated into the string resulting in a string value.

One more subtle nuance is when the value of the parameter is determined. With the first form param=name, value of name is obtained from the context before the section (or partial or helper) is processed/invoked. Ditto for the second form which is a string constant. The third form is the tricky one. Dust only evaluates the interpolated string within the section/partial/helper. Mostly this does not matter but if a new value for yyy is added to the context stack via another param of the same section/partial/helper (i.e. {#A.B param="{yyy}" yyy="baz"}), that value will be found and used rather than the one known at the point of call. Be careful naming parameters the same as data you are referencing from the context.

Logic in Templates

Templates with logic versus "logic-less" templates is a hotly debated point among template language designer and users. Dust straddles the divide by adopting a "less logic" stance. We all know that real business logic does not belong in the presentation layer, but what about simple presentation-oriented things like coloring alternate rows in table or marking the selected option in a <select> dropdown? It seems equally wrong to ask the controller/business logic code to compute these down to simple booleans in order to reduce the logic in the presentation template. This route just lead to polluting the business layer code with presentation-oriented logic.

Dust provides some simple logic mechanisms and trusts you to be sensible in minimizing the logic in your templates to only deal with presentation-oriented decisions. That said, let's take a look at the ways Dust let's you have logic.

There are two other special section notations that provide conditional testing:

  • {?name} body {/name}
    not only tests the *existence* of name in the current context, but also evaluates the value of name in the JSON model. If name is "true" (see below for what true means), the body is processed.
  • {^name} body {/name}
    not only tests the *non-existence* of name in the current context, but also evaluates the value of name in the JSON model. If name is not true (see below for what true means), the body is processed.
Note that the value of name is evaluated as follows:
  • "" or ' ' will evaluate to false, boolean false will evaluate to false as well, null, or undefined will evaluate to false.
  • Numeric 0 evaluates to true, so does, string "0", string "null", string "undefined" and string "false".
  • Empty array -> [] is evaluated to false and empty object -> {} and non-empty object are evaluated to true.
Here is an example of doing something special when the array is empty.


  <li>{name}, {age}{~n}</li>
  <p>You have no friends!</p>


   friends: [
      { name: "Moe", age: 37 },
      { name: "Larry", age: 39 },
      { name: "Curly", age: 35 }

This renders html as expected:

  <li>Moe, 37</li>
  <li>Larry, 39</li>
  <li>Curly, 35</li>

If we change the friends array to be empty, the {:else} block is triggered

  friends: [ ]

In the original dust, it does not trigger the {:else} block. Our version fixed it, to keep # and ? consistent

Take special care if you are trying to pass a boolean parameter. param=true and param=false do not pass true/false as you might expect. They are treated as references to variables named "true" and "false". Unlike JavaScript, they are not reserved names. Note that they are not reserved in JSON either so you can have a property named true or false. So you might think to pass 0 and 1 to your boolean-like parameter. That won't work either. dust's boolean testing ( {?xxx} is more of an existence test than a boolean test. Therefore, with param=1 and param=0 both value exists and so are considered true. Your best bet is to pass 1 and "", e.g. param=1 or param="". You could also leave off param="" if you are sure the name is not elsewhere in your JSON data and accessible.


A Dust template named "xxx" is authored in a file named xxx.dust. You can have multiple .dust files and reference one Dust template as part of another one. This is the basis for "components or reusable templates for tasks like a common header and footer on multiple pages. Note that the .dust file extension is used here in examples but .tl is also commonly seen. Since it only matters to the build process you can use whatever extension works for you.

Let's peek under the covers to see how the Dust template rendering knows about a template. As we said earlier, Dust templates are compiled to JavaScript. Part of that compiled result is a call to dust.register(name, functionThatImplementsCompiledTemplate). The register call associates a template name with the function to run that template. So consider this example of how partials might be used:

  {>header /}
   ... template for the body of the page...
   {>footer  /}

As long as the JavaScript for the header.dust and footer.dust templates is loaded and registered prior to executing this template, it will run the header template, then its own body view and finally the footer template.

The partial reference syntax {>name /} also supports paths so you can have a template at a path like "shared/header.dust} and reference it as {>"shared/header" /}. This allows partials to be organized into library-like structures using folders.

Like sections, partials accept parameters so you can build reusable components that are parameterizable easily. This gives you the same foundation for building libraries as other languages. By passing all the data into the partial using parameters, you isolate the partial from any dependence on the context when it is invoked. So you might have things like {>header mode="classic" /} to control the header behavior.

Just like in sections, inline parameters will not override the current context if a property of the same name exists. For example, if the current context already has {name: "Albert"} adding name as a parameter will not override the value when used inside the partial foo.

{>foo name="will not override Albert"/} 

For dust users of versions prior to 2.0.0, if you use parameters to pass an object like:

  homeAddress: {
    street: "1 Main St",
    city: "Anytown"

{>displayAddress address=homeAddress /}

then you will not be able to reference {address.street} or {} in the body of the partial. These get treated as a path reference and the params are higher in the context stack at the point of reference so cannot be found. You need to code such things as:

  {street} {city}

From dust 2.0.0 on, you can write the more natural

{address.street} {}

Dynamic Partials for Logic

Note that you can also use dynamic partials, that conditionally select the partial to render based on the value in the JSON.

  {>"flowViews/flowView{flowName}" /}

This sort of usage might suit a case where you have a multi-page flow and the controller could pass "page1", "page2",... in the data model to dynamically choose which partial to use to implement the view.


Helpers can be found in a separate repo at: In order to use the dust helpers described below, you need to load the dustjs-linkedin library first and then load the dustjs-helpers library. If you are using node.js, then the following require statements will ensure you have the needed dust code available.


Logic Helpers

{@select key="xxx"} + @eq, @ne, @lt, @lte, @gt, @gte, @default, Available in Dust 1.0

Select provides a key value that can be tested within its scope to output desired values. It mimics the switch/case statement. Here are some examples:

    {@select key=\"{foo}\"}
         {@eq value=\"bar\"}foobar{/eq}
         {@eq value=\"baz\"}foobaz{/eq}
         {@default} - default Text{/default}

     {@select key=foo}
         {@gte value=5}foobar{/gte}

Each test condition is executed and if true, the body is output and all subsequent condtions are skipped. If no test condition has been met and a @default is encountered, it will be executed and the select process terminates.

The @eq (for example) can be used without a {@select}.The most common pattern of usage would be for an HTML <select>/<option> list to mark the selected element with a "selected" attribute. The code for that looks like this where {#options} is an array of options from the data model. Here the key is directly on the eq rather than on the select helper.

          <select name="courses">
              <option value="{value}"{@eq key=value value=courseName}  selected="true"{/eq} >{label}</option>

Similarly, {@ne}, {@lt}, {@gt}, {@lte}, {@gte} can be used standalone and allow nesting. The following is a valid example

          {@eq key="CS201" value=courseName}
           {@eq key="CS101" value=prereq}
            print it is CS201 course and has CS 101 as prereq 

Note that all of {@eq}, {@ne}, {@lt}, {@gt}, {@lte}, {@gte} support an else block so you can output an alternative result if the test is false.

          {@eq key="CS201" value=courseName}
            You are enrolled in CS201
            You are not enrolled in CS201

==== {@select key="xxx"} used to provide "or" functionality -- new in release 1.6.0 The addition of {@any} and {@none} helpers broadened the scope of what {@select} can be used for. {@any} must be contained in an {@select block}. It will render its body if any truth test within the {@select} block is true. Multiple {@any} tags can appear and all will be rendered if any truth test in the select is true. Muliple {@none} tags can appear anywhere in the {@select} block and all of them will be rendered if none of the truth values are true.


{! foo is 1 or 2 or 3 !}
{@select key=foo}
  {@eq value="1"/}
  {@eq value="2"/}
  {@eq value="3"/}
  {@any}Foo is 1, 2, or 3{/any}
  {@none}Foo is not 1, 2, or 3{/none}

{! Output something specific as well !}
{@select key=foo}
  {@any}Congratulations! You got: {/any}
  {@eq value="1"}First Place{/eq}
  {@eq value="2"}Second Place{/eq}
  {@eq value="3"}Third Place{/eq}
  {@none}Better luck next time!{/none}

{! If any truth test passes, all {@any} blocks will be evaluated !}
{@select key=foo}
  {@any}<span class="place">{/any}
  {@eq value=1}First Place{/eq}
  {@eq value=2}Second Place{/eq}
  {@eq value=3}Third Place{/eq}

{@math} - math helper, Available in Dust 1.1 release

The math helper provides simple computational capabilities. Operations supported are: add, subtract, multiply, divide, mod,abs, floor, and ceil. The general syntax is:

{@math key="operand1" method="mathOpName" operand="operand2" /}
The helper computes a result using the key, method, and operand values. Some examples will clarify:
{@math key="16" method="add" operand="4"/}  - Result will be 20
{@math key="16.5" method="floor"/} - Result will be 16
{@math key="16.5" method="ceil"/} - Result will be 17
{@math key="-8" method="abs"/} - Result will be 8
{@math key="{$idx}" method="mod" operand="2"/} - Return 0 or 1 according to $idx value

@math with bodies

Sometimes you need to choose something to output based on the result of a math helper computation. For example, if the table row number is odd, you want to give it a gray background.

<tr class="
 {@math key=$idx method="mod" operand=2}
    {@eq value=0}

The above evaluates the mod with the given key and operand i.e $idx % 2 and then checks if the output is 0, and prints the block inside the @eq helper, if not the else block. Be careful to use numeric values for tests and not strings, e.g. {eq value="0"} will never be true.

Another example

{@math key="13" method="add" operand="12"}
 {@gt value=123}
  13 + 12 > 123
  Math is fun

Using the nested @eq @lt etc. syntax allows you to output values like a select/case similar to the select helper.

{@if cond="condition"} - if helper, Removed in 1.6.0 release

There are a few cases where a simple true/false or exists/non-exists or single eq or lt or gt test won't suffice. For those, there is the if helper.

Some examples

 {@if cond="{x} < {y} && {b} == {c} && '{e}'.length || '{f}'.length"}
  <div> x is less than y and b == c and either e or f exists in the output </div> 
 {@if cond="({x} < {y}) || ({x} < 3)"} <div> x<y or x<3 {/if}
 {@if cond="{x} < {y} && {b} == {c} && '{e}'.length || '{f}'.length "}
   <div>  x is less than y and b == c and either e or f exists in the output </div> 
   <div> x is >= y </div>

Caveat #1: In the above example, if there is a possibility of undefined or false value for the {x} or {y} in the JSON, the correct syntax would be to check it exists and then check for {x} > {y}. This is a known limitation since, {x} returns nothing when the value of x is undefined or false and thus results in invalid js condition in the if helper

 {@if cond="'{x}'.length && '{y}.length && {x} < {y} && {b} == {c} && '{e}'.length > 0 || '{f}'.length > 0 "}
  <div> x is less than y and b == c and either e or f exists in the output </div> 

Caveat #2: The if helper internally uses javascript eval, for complex expression evaluation. Excessive usage of if may lead to sub-optimal performance with rendering, since eval is known to be slow. It can also be unsafe.

Release 1.6.0 added safer alternatives to do most of what his helper provided.

Other Helpers

Dust provides a mechanism to extend the capabilities of the templating solution. Currently there is a small set of helpers that come with the release:

{@sep} - Separator helper, Available in Dust V1.0 release

When outputting lists of things, you often need to do something different for the last iteration. Consider the case

My friends are: 

As written this will produce Hurley,Kate,Sawyer,Desmond, leading to the "dangling comma problem". This can be fixed by using the {@sep} helper tag as follows:

My friends are: 

The {@sep} helper tag will output it's body content unless this is the final iteration of the containing loop.

{@idx} - Index helper, Removed in 1.6.0 release

The idx helper tag provides a way to get the index of the current iteration. The need for this has been eliminated by the introduction of {$idx}.

The {@first} and {@last} helpers were added in the 1.6.0 release. If it is the first iteration of a section loop, {@first} will render its body. If it is the final iteration, {@list} will render its body.

For example,

My friends are: 
 <option value="id_{@idx}{.}{/idx}">{name}</option>

Here we are using idx to generate a unique id for each option tag in a dropdown. Therefore, we would have "id_0, id_1,... for id values. Within the idx helper {.} references the current iteration count.

{@size key="xxx" } - size helper, Available in Dust V1.1 release

The size helper computes the size of the keyparameter. The size computed depends on the type of the subject parameter as follows:

  • Array - number of elements, [1,2,3,4] has size=4
  • String - length of the string, "abcdef" has size=6
  • Object - Number of properties in the object, {a:4, b:8, c:15, d:16} has size=4
  • Number - Value of the number, 23 has size 23 and 3.14 has size 3.14
  • Undefined, 0, empty string - zero
  • Any other value - length after conversion to string

{@contextDump key="current|full" to="output|console"/} - contextDump helper, Available in Dust V1.1 release

The contextDump helper outputs the current context portion of the JSON data model to the output stream. This can help with debugging if you suspect the context data is not as expected or you aren't sure what the current context is. If you want to change the defaults of key="current" and to="output", use the parameters. Remove this tag when done debugging.

Blocks and Inline Partials

An important need in developing a multi-page web application is to have common elements of the pages defined just once and shared by all pages (Don't Repeat Yourself). Dust provides this with the concept of blocks. Consider a common case where several pages share a header and footer but have different body content.

Blocks in the base template can contain default content and a child template can override that content. A block tag has the form {+name}default Content{/name}. In the following example, the base template has three blocks: pageHeader, bodyContent, and pageFooter. The pageHeader and pageFooter have default content that is shown if the child template does not override them.

Base template

<div class="page">
  <div class="bodyContent">
  <div class="footer">
       <a href="/contactUs">Contact Us</a>

Now that we have defined a base template with named blocks pageHeader, bodyContent, pageFooter , let's look at how a child template can use it to supply body content and override the pageFooter. First, you insert the base template as a partial. Then you use one or more "inline partials" defining the values for the named blocks in the template.

Child template

{! First, insert the base template as a partial !}

{! Then populate the base template named blocks. Supply the desired bodyContent and pageFooter !}
<p>These are your current settings:</p>
       <a href="/contactUs">About Us</a> |
       <a href="/contactUs">Contact Us</a>

Note that inline partials like {<name}xxx{/name}, define "name" globally within the template. While this might be useful, remember the pains caused by global variables in JavaScript and use these with the knowledge that others can stomp on your chosen name inadvertently.

Dust under the covers

Here we take a look at what goes on to actually run a dust template behind the scenes. Typically, your framework or environment is taking care of this for you. If not, then you need to know how to do it.

Compiling a Dust template

Assuming you have the dust compiler in your JS environment, you can compile a Dust template source file to a JavaScript form. The following will compile our very simple template "Hello {name}!" to a JavaScript string that will also register it as a template named "intro".

var compiled = dust.compile("Hello {name}!", "intro");

If you include the "compiled" string as part of a script block of JS that you load, then the "intro" template will be defined and registered. If you want to do it immediately then do:

to reach the same state.

Running a Dust Template

Assuming you have the "intro" template we previously compiled and loaded, you can run it using the code below. The first argument to dust.render is the registered template name, the second argument is the JSON model, the third argument is a callback function where "out" contains the rendered output from running the template (and err has any error information).

dust.render("intro", {name: "Fred"}, function(err, out) {

Debugging a Dust Template

In addition to the contextDump helper, dust, as of 2.1.0 has a debug option that provides a lot of error checking on your templates that are being run. Reporting logs to the console.

To activate debugging, set dust.isDebug = true; A separate variable controls the level of debug logging: dust.debugLevel = 'DEBUG' | 'INFO' | 'WARN' | 'ERROR' The default setting is 'INFO';

Setting your own context before render

Dust normally has a context stack reflecting your template's use of sections to access the JSON model. However, at the top level of the context stack, you can setup other data you want Dust to find. These are referred to as globals. As of dust 2.0.0, both key and path references will search in globals if not match was found earlier. Prior to dust 2.0.0, only key searches looked in globals.

Using globals, you can provide Javascript functions for general use by the template. First use dust.makeBase to create a base context of data, or in this example, a function to return an output value; then for the second argument to dust.render, use base.push(yourRegularJsonModel).

// Set up a base context with a global helper function
var base = dust.makeBase({
  sayHello: function() { return "Hello!" }

// Push to the base context at render time
dust.render("index", base.push({foo: "bar"}), function(err, out) {

Controlling whitespace suppression

Dust 2.5.0: Simply set the config flag.

   dust.config.whitespace = true;

Dust pre-2.5.0: You can manually clobber the whitespace optimizer.

   dust.optimizers.format = function(ctx, node) { return node };

Warning: pre-2.5.0, this will break Dust when you compile very large (over 150KB or so) templates.

Determining dust version

As of 2.6.0 release, dust.version has the release number

Writing a dust helper

Dust helpers are javascript functions registered with the dust.helpers object as a property. Thus the general form of a helper is:

    dust.helpers.myHelper = function(chunk, context, bodies, params) {
      code of the helper

As far as the parameters go:

  • chunk is the currently accumulating output of the template render process. You will most likely contribute additional output as part of your helper.
  • context is the current context stack (e.g that which changes when you do things like {#list}
  • bodies holds any body sections nested within the helper. For example, the {:else} body.
  • params is an object that holds all the parameters used when calling the custom helper
Here is a sample custom helper that implements a substring capability.

{@substr str="xxx" begin="x" end="y" len="z" /}

  • begin is optional, zero if omitted
  • end and len are choices with len taking priority if the user supplies both.
  • If len is present, str.substr(begin,len) is used for the result.
  • If end is present, str.slice(begin,end) is used for the result
  • If end and len are both missing then you get the whole string back (e.g. str.substr(0).
The annotated code to implement it is:
    dust.helpers.substr = function (chunk, ctx, bodies, params) {
    // Get the values of all the parameters. The tap function takes care of resolving any variable references
    // used in parameters (e.g. param="{name}"
    var str = dust.helpers.tap(params.str, chunk, ctx),
        begin = dust.helpers.tap(params.begin, chunk, ctx),
        end = dust.helpers.tap(params.end, chunk, ctx),
        len = dust.helpers.tap(params.len, chunk, ctx);
    begin = begin || 0; // Default begin to zero if omitted
    // Use JavaScript substr if len is supplied.
    // Helpers need to return some value using chunk. Here we write the substring into chunk.
    // If you have nothing to output, just return chunk.write("");
    if (!(typeof(len) === 'undefined')) {
        return chunk.write(str.substr(begin,len));
    if (!(typeof(end) === 'undefined')) {
        return chunk.write(str.slice(begin,end));
    return chunk.write(str);

Parameters can come in many forms,e.g. param=2, param={a}, param="{a}". Some can be accessed and used directly from the params parameter to the helper. Others require evaluating a function to obtain the final value. You can avoid all the bother around this by using dust.helpers.tap(, chunk, ctx) which returns you the final value of the parameter.

If you need to work with the body of the helper, then the following will get it for you.

    var body = bodies.block;

To evaluate the body, you call it with body(chunk, context). There are other parameters if you intend to emulate a looping structure like a section letting you define $idx and $len.

context is the dust context stack. Normally you will just use the dust get method when retrieving values from the context stack. If you need a deeper knowledge, take a look at the code for Context.prototype.get. Prior to dust 2.1.0, getPath was used for references to paths. With dust 2.1.0, that capability has been incorporated into get.

Generally, you should always return the output of your helper as a chunk using the write method on it to add your helper's generated output to the accumulation in the chunk. Dust chains the result of your helper to further actions which expect to be able to add to the chunk.

External Support for dust usage

Loading dust in the browser

If you want to use the dust libraries in the browser, your best performance is to load them from a CDN. There are a couple of CDNs that host dust libraries. Here are some links:

jdelivr supports multiple versions so this URL will let you choose one for dust and the helpers:!dustjs!dustjs.helpers also hosts dust but, as of now, does not seem to host the dust helpers.

As of dust 2.6.0, AMD loading in the browser is supported. See for details.

Standalone dustc compiler

A standalone dust compiler for use with builds is available with the 2.6.0 release.

Command: dustc [options] [path1]

Example: dustc --output=compiled.js tmpl/**/*.dust'

Command options are:

    name: ['n', 'name to use when rendering the template (multiple inputs automatically use their paths)', 'string']
    output: ['o', 'concatenate all output to this file', 'path']
    split: ['s', 'create one output file per input file instead of concatenating them to --output' ]
    pwd: [false, 'generate template names starting from this directory', 'string']
    whitespace: ['w', 'preserve whitespace in templates', ]
    amd: ['a', 'register templates as AMD modules' ]
Clone this wiki locally