Skip to content
Blade - HTML Template Compiler, inspired by Jade & Haml
JavaScript HTML Shell
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Blade - HTML Template Compiler

Blade is a HTML Template Compiler, inspired by Jade & Haml, implemented in JavaScript, so it will run on your microwave oven.

It works like this...

  1. Write up your template in Blade (which is a Jade-like language)
  2. Use the Blade compiler to generate a Blade template (which is a JavaScript function)
  3. Pass variables into your generated template to produce HTML or XML

View a simple example

Never write HTML again. Please.


"Blade's blood is the key" :P Sorry... I had to...

This is NOT the Blade templating engine developed by Laravel. Laravel Blade was added in Sept, 2011; whereas, I did not begin development on Blade until May, 2012. Nevertheless, I still blame Laravel for choosing the same name and for creating any confusion.

Table of Contents

Why use Blade instead of Jade?

Here are the reasons Blade might be considered "better" than Jade:

  • Jade is an ornamental stone. Blade is a badass vampire hunter.
  • Client-side templates can be served to the browser, no problem. See Browser Usage and Blade Middleware for more info.
  • Meteor support - Blade works well with Meteor. See the documentation below.
  • Compatibility - The language syntax of Blade is very similar to Jade's. Jade is an awesome templating language, and if you are already familiar with it, getting started with Blade should take you very little time.
  • Smarter file includes Files compiled in Blade can be much smaller than Jade files when you are using file includes because file includes happen at runtime instead of at compile-time. If you re-use the same included file among multiple views, the included file does not need to be reloaded.
  • Blocks in Blade are awesome. We removed features from Jade like explicit template inheritance and then added features like blocks and parameterized blocks. You might find our idea of a block to be similar to Jade's, but just wait until you realize how much more flexible they are!
  • Just Functions, not mixins or partials. In Blade, there are no "mixins" or partial templates. There are only functions, and they work just like regular JavaScript functions that you've come to know and love. You can put your functions into separate files and include them into other templates, you can take advantage of the arguments Array-like Object, or whatever you want!
  • Other cool features For example, Blade provides a built-in syntax for taking content rendered by a function and loading it into a variable within your view template, allowing you to pass rendered HTML content to another function. This is just one of the many new features you can utilize when you make the switch to Blade.
            Jade            vs.             Blade

Jade    Blade

OK... it's admittedly not as funny as I thought it would be. But, I tried.


  • Write extremely readable short-hand HTML
  • Insert escaped and unescaped text and vanilla JavaScript code
  • Code and text are escaped by default for security/convenience
  • Functions (like Jade mixins)
  • Dynamic file includes
  • Regular blocks and Parameterized blocks (aids in supporting template inheritance)
  • True client-side template support with caching, etc.
  • Supports Express.JS
  • HTML Comments and block comments
  • Text filters
  • Nice error reporting to help you debug your broken templates
  • Command-line tool to compile/render templates (try blade --help)
  • Meteor smart package
  • Write DOM event handlers right into your views
  • Cool plugins (including Live UI)

Project Status

UPDATE: Meteor 0.5 support is now available, and many more features are coming soon! Thanks for your patience!

I'd say that Blade is stable. There are very few (if any) known issues, and I think that Blade is ready for production environments. I use Blade for many of my projects.

If you find a bug, please report it here. If you include the Blade code that failed along with the expected HTML output, that is always splendid.

By all means, please feel free to submit pull requests for new features, new tests, or whatever! For big changes, say ~100 lines of code, you might want to contact me first or submit an issue before getting started.


for Node (via npm): sudo npm install -g blade

Runtime for Browsers: wget

Minified runtime is about 6-7 KB, uncompressed.



Like Jade, a tag is simply a word. For example, the string html will render to <html></html>.

You can have 'id's:


which renders as <div id="awesome"></div>.

Any number of classes work, separated by a dot (.)


which renders as <div class="task-details container"></div>.

Tag attributes? Yep, they work pretty much like Jade, too. Put attributes in parenthesis, separate attributes with a comma, space, newline, or whatever.

a(href="/homepage", onclick="return false;") renders as:

<a href="/homepage" onclick="return false;"></a>

You can also have line feeds or weird whitespace between attributes, just like in Jade. Whatever. This works, for example:

        value="Your email here"

You can also put substitute an attribute value with vanilla JS code like this: input(type="text" name="contact-"+name value=value). For example, if you passed the object {name: "fred", value: "testing"} to your view, the above would render to: <input type="text" name="contact-fred" value="testing"/>

You cannot put whitespace, commas, newlines, or parentheses in the vanilla JavaScript code, though. Blade uses these characters to separate each attribute or to end the tag definition.

And, yes... the class attribute is handled with extra special care. Pass an array or string. Classes (delimited by ".") from before will be merged with the value of the class attribute.

For example:"another dude") renders as: <div id="foo" class="bar dummy another dude"></div>

Boolean attributes are allowed, as well. If the attribute value is boolean true, then the attribute is set; if the attribute value is boolean false, then the attribute is ignored entirely. For example:

input(type="text" checked=true) renders as: <input type="text" checked="checked"/>.

Or... you can write it HTML 5 style like this:

input(type="text" checked) which renders as: <input type="text" checked="checked"/>.

div, div, div can get annoying... so, we can omit the tag specifier if we specify an id or some classes:


renders as:

<div id="foo"></div><div class="bar"></div><div id="this" class="is cool"></div>

Blade just assumes anything without a tag name specifier is a <div> tag.

Also, tags without matching ending tags like <img/> render properly.


It works. You can indent with any number of spaces or with a single tab character. The only rule is to be consistent within a given file. Jade gives you a lot of weird indent flexibility. Blade, by design, does not.


renders as:

<html><head></head><body><div id="content"></div></body></html>

You can start a tag name with a bashslash to escape Blade keywords. Normally, include test would include a file, but \include test renders as:


This allows you to be flexible with tag names, so you are not restricted to rendering HTML, for example. You can render any XML document with Blade.


It works, too. Simply place content after the tag like this:

p This text is "escaped" by default. Kinda neat.

renders as:

<p>This text is &quot;escaped&quot; by default. Kinda neat.</p>

Want unescaped text? Large blocks of text? Done. Start a line of text with a |.

p! This will be <strong>unescaped</strong> text.
        How about a block? (this is "escaped", btw)
        Yep. It just works!

renders as:

<p>This will be <strong>unescaped</strong> text.
How about a block? (this is &quot;escaped&quot;, btw)
Yep. It just works!

Rules are:

  • Text is escaped by default.
  • Want unescaped text? Precede with a !
  • Precede with a = to evaluate and output some JavaScript.
  • Large text block? Use | and indent properly.
  • Unescaped text block? Use |! or even just ! works.
  • JavaScript code block? Use |= or even just = works.
  • Unescaped JavaScript code block? Yep. Use |!= or !=.
  • Newlines in text blocks are preserved.

Variable interpolation is supported for text blocks. Use #{var_name} notation, and anything between the curly braces is treated as vanilla JavaScript code.

For example, you can write:

(caution: indents are required on line 4 even though it is blank)

        I am just testing #{whatever + ", alright?"}


instead of writing the equivalent, but arguably less awesome...

        "I am just testing " + whatever + ", alright?" +
        "\n\n" +

Assuming a local variable whatever is passed to the template with value "Blade", both of the examples above will render to this:

<p>I am just testing Blade, alright?


Text filters

Need <br/> tags inserted? Use a built-in filter, perhaps?

        How about some text with some breaks?

        Yep! It works!

renders as:

<p>How about some text with some breaks?<br/><br/>Yep! It works!</p>

Built-in text filters include:

  • :nl2br - Escapes the content and converts newline characters to <br/>
  • :cdata - Surrounds text like this: <![CDATA[ ...text goes here... ]]> Text should not contain ]]>.
  • :markdown (must have markdown-js installed)
  • :md (alias for :markdown)
  • :javascript - Generates a <script> tag for your JavaScript code. If minify compiler option is set and UglifyJS is installed, your code is uglified automatically.
  • :js (alias for :javascript)
  • :coffeescript - Generates a <script> tag for the generated JavaScript. (must have coffee-script installed)
  • :cs (alias for :coffeescript)
  • :stylus - Generates a <style> tag for the generated CSS. If minify compiler option is set, your CSS is compressed automatically. (must have stylus installed)
  • :less - Generates a <style> tag for the generated CSS. (must have less installed)
  • :sass - Generates a <style> tag for the generated CSS. (must have sass installed)

Filters are essentially functions that accept a text string and return HTML. They cannot modify the AST directly. Also, you cannot inject JavaScript code into filters.

You can add custom filters at compile-time using the API.

Variable interpolation is supported for certain text filters, as well. If a text filter returns text in #{var_name} notation, then anything between the curly braces is replaced with vanilla JavaScript code. To avoid this behavior, text filters can either escape the #{stuff} with a backslash, or it can set its interpolation property to false.


Use dash (-) to indicate that JavaScript code follows, which will not output into the template. As before, use equals (=) to specify code output. A few examples, please?

Using dash (-):

    - if(task.completed)
        p You are done. Do more! >:O
    - else
        p Get to work, slave!

When inserting lines of code with -, curly braces or semicolons are inserted, as appropriate. In the example above, we have an if statement followed by an indented paragraph tag. In this case, Blade wraps the indented content with curly braces. If there is no indented content beneath the line of code, then a semicolon is appended instead.

Code that outputs (i.e. a code block or at the end of a tag). As mentioned before, it's just like a text block, except with an =.

#taskStatus= task.completed ? "Yay!" : "Awww... it's ok."
    | The task was due on
    |= task.dueDate

When using code that outputs, the default is to escape all text. To turn off escaping, just prepend a "!", as before:

    |!= some_var_containing_html

Missing "|" characters are okay, too. Just don't forget that stuff after the "=" needs to be valid JavaScript code!

    = "escape me" + " away & away"

renders <p>escape me away &amp; away</p>

Variable names to avoid

Blade, like other template engines, defines local variables within every single view. You should avoid using these names in your view templates whenever possible:

  • locals
  • cb
  • __ (that's two underscores)
  • Any of the compiler options (i.e. debug, minify, etc.)

Foreach block

The exact syntax of a foreach block is the word "foreach", followed by the variable name of the JavaScript Array or Cursor Object, optionally followed by "as" and an item alias. Finally, it is possible to follow the foreach block by an "else" block, which is only rendered if there were no items in the collection.

As a side note, a Cursor Object, as described above, is an Object with an observe() method, as described by cursor.observe(callbacks)

For example:

    foreach users as user
        li #{user.firstName} #{user.lastName} (#{user.age})
        li No users were found

Assuming that users is an Array, the above would produce the same as:

    - for(var i = 0; i < users.length; i++)
        - var user = users[i];
        li #{user.firstName} #{user.lastName} (#{user.age})
    - if(users.length == 0)
        li No users were found

The foreach block is preferred over the example above not only because of readability and brevity, but because it also provides Blade with the ability to better integrate with live page updating engines (like Meteor, Spark, Live UI, etc.). That is, if the live page update engine supports tracking reactive collections, the most efficient DOM operations may occur to update the view's results in-place, without re-rendering the entire Blade template.


Don't forget a doctype! Actually, you can, whatever...

Add a doctype using the doctype keyword or !!! like this:

!!! 5 means use HTML 5 doctype.

Use the list of built-in doctypes or pass your own like this:

doctype html PUBLIC "-//W3C//DTD XHTML Basic 1.1//EN"

which renders as <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML Basic 1.1//EN"><html></html>

Put the doctype at the top of your Blade files, please. Please refer to doctypes.js for the list of built-in doctypes.

You can modify the list of built-in doctypes through the API, if you insist.


Use // for a line comment. Use //- if you don't want the comment to be rendered. Block comments work, too.

//Comment example 1
//-Comment example 2
    p Block comments work, too

renders as:

<!--Comment example 1--><!--<div id="wow"></div><p>Block comments work, too</p>-->

Conditional comments work like this:

    //if lt IE 8

renders as:

<head><!--[if lt IE 8]><script src="/dear-microsoft-plz-stop-making-browsers-kthxbye.js"></script><![endif]--></head>

To comment out entire sections of Blade code, you can use non-rendering block comments with a text block.

        anything can go here... Blade code, JavaScript code, whatever...
        just make sure that the indenting is right.

or... even better... just use C-style block comments. Begin with /* to generate a non-rendering block comment, or begin with /** to generate a regular comment. End the comment with */. These comments are not parsed like // comments.

/* h1 Testing */
    h3 Notice that this chunk of Blade code is not parsed

renders as:

    h3 Notice that this chunk of Blade code is not parsed

Event Handlers

You can write event handlers right into your Blade templates. Here's an example:

form(method="post" action="/login")
    input(type="text" name="username")
            //javascript code goes here
            //this refers to this DOM element
            //e refers to the browser's event Object
    input(type="password" name="password")

The above code will automatically register the 'onchange' event handler with the corresponding input tags.

As shown in the example, your event handler may reference this (the DOM element that triggered the event) or e (the browser's event Object). Be aware that every browser's event Object might be slightly different, especially in legacy browsers. In addition, if you are rendering the template in the browser (i.e. using client-side templates), your event handler will have access to the view's locals.


Functions are reusable mini-templates. They are similar to 'mixins' in Jade.

Defining a function:

function textbox(name, value)
    input(type="text", name=name, value=value)

Calling a function and inserting into template structure:

    call textbox("firstName", "Blake")

Or... maybe just putting the generated HTML into a variable?

call textbox("firstName", "Blake") > text
//alternative syntax: call text = textbox("firstName", "Blake")

Both examples would render:

<form><input type="text" name="firstName" value="Blake"/></form>

You can also append content rendered by a function to a variable: call textbox("firstName", "Blake") >> text or... alternatively... call text += textbox("firstName", "Blake")

Note: when you define a block (see below) within a function, and you output the rendered content to a variable, the block will be destroyed immediately after the function call.

Yes, you can use arguments within your function, just like a "real" JavaScript function. In fact, functions are "real" JavaScript functions, so even closures work! Although, remember that functions have access to the variables in scope at the time the function was defined, not the variables in scope when the function is called.


- var x = 12;
function test(foo)
    - if(x)
    call test("Header")

would render: <div id="example"><h1>Header</h1><p>12</p></div>

Adding classes or an id to rendered function content

Yes, you can add a class name or id to the first element rendered by a function:

function dialog(msg)
        = msg
call dialog("Blade is awesome")

which would render as <div id="foobar" class="dialog foo bar">Blade is awesome</div>.

Although, if you try it with something like this, you get an error because the first child rendered by the function is not a tag.

function dialog(msg)
    = msg
call dialog("Blade is awesome")
//compiler might generate an error, or it might just ignore the id and classes

Dynamic file includes

include "file.blade"

This will insert "file.blade" right into the current view at runtime, as if the contents of the included file were copied right into the current view.

If you don't know the name of the file to be included until runtime, that's no problem. The include statement can also be followed by the name of a JavaScript variable containing the filename to be included. These are called dynamic filename includes.

- var filename = "file.blade"
include filename

CAUTION: When using dynamic filename includes in the browser, be sure that you have properly loaded all views that might be included into the browser's cache before executing the view containing the dynamic filename include. See the implementation details for a more detailed explanation.

If you do not specifiy a file extension, .blade will be appended to your string internally.

You may also place an include inside of a function or block.

Finally, you can specify which local variables should be passed to the included view template by using the exposing keyword. By default, Blade will pass the parent's local variables to the included template; however, when using the exposing keyword, you can specify exactly which variables are to be exposed to the included template.

For example:

- header = "Header: 1, 2, 3"
- text = "This is some text: 1, 2, 3"
- for(var i = 0; i < 10; i++)
    include "foobar" exposing i, text

In the example above, variables i and text are exposed to "foobar.blade"; the header variable will not be accessible from "foobar.blade".


Blocks allow you to mark places in your template with code that may or may not be rendered later.

You can do a lot with blocks, including template inheritance, etc. They behave quite differently from Jade.

There are two types of blocks: regular blocks and parameterized blocks.

Regular blocks

Regular blocks are defined using the "block" keyword followed by a block name. Then, you optionally put indented block content below. Like this:

block regular_block
    h1 Hello
    p This is a test

Assuming nothing else happens to the block, it will be rendered as <h1>Hello</h1><p>This is a test</p> as expected. Empty blocks are also permitted. A simple, empty block looks like this: block block_name

Of course, the purpose of declaring/defining a block is to possibly modify it later. You can modify a block using three different commands:

  • Use the append keyword to append to the matching block.
  • Use the prepend keyword to prepend to the matching block.
  • Use the replace keyword to replace the matching block.


append regular_block
    p This is also a test

Replacing a block

Replacing a block is somewhat confusing, so I will explain further. If you replace a block, you are not changing the location of the defined block; you are only replacing the content of the block at its pre-defined location. If you want to change the location of a block, simply re-define a new block (see below).

In addition, when you replace a block, all previously appended and prepended content is lost. The behavior is usually desired, but it can sometimes be a source of confusion.

If you replace a parameterized block (described below) with a regular block, you cannot call "render" on that block.

You can replace a regular block with a parameterized block (described below). This will also clear the contents of the block, as expected.

Parameterized blocks

The other type of block is called a parameterized block, and it looks like this:

block param_block_yo(headerText, text)
    h1= headerText
    p= text

Parameterized blocks do not render automatically because they require parameters. Therefore, assuming nothing else happens to the block, the block will not be rendered at all.

To render a block, use the "render" keyword like this:

render param_block_yo("Some header text", 'Some "paragraph" text')

Now, assuming nothing else happens to the block, the block will be rendered as:

<h1>Some header text</h1><p>Some &quot;paragraph&quot; text</p>

You can render as many times as you wish, and by default, the rendered content will be appended to the block. You can also prepend the rendered content to the block or replace the contents of the block with rendered content. Here are the variations:

  • render param_block_yo("Some header text", 'Some "paragraph" text')
  • render append param_block_yo("Some header text", 'Some "paragraph" text') (same as above)
  • render prepend param_block_yo("Some header text", 'Some "paragraph" text')
  • render replace param_block_yo("Some header text", 'Some "paragraph" text')

Parameterized blocks are really cool because regular "append", "prepend", and "replace" all work, too. Just remember that order matters.

Another example:

    block header(pageTitle)
        title= pageTitle
    h1 Hello
    render header("Page Title")
    append header
    render header("Page Title")
    prepend header

Will output:

    <title>Page Title</title>
    <script type="text/javascript"></script>
    <title>Page Title</title>

What happens if I define the same block more than once?

You can re-define a block that has already been defined with another "block" statement. This completely destroys the previously defined block. Previously executed "append", "prepend", "replace", and "render" blocks do not affect the re-defined block.

In summary...

  • Use the block keyword to mark where the block will go (block definition).
  • Use the render keyword to render the matching "parameterized" block. Do not use this on a regular block.
  • Use the append keyword to append to the matching block.
  • Use the prepend keyword to prepend to the matching block.
  • Use the replace keyword to replace the matching block.

You may not render, append to, prepend to, or replace undefined blocks. If you do so, an error message will occur.

When you define a block within a function, and you output the function's rendered content to a variable, the defined block will be destroyed immediately after the function call.

Template Inheritance

There is no extends keyword. Just use blocks and includes:


        block title(pageTitle)
        block body


include "layout.blade"
render title("Homepage")
replace block body
    h1 Hello, World

If you render layout.blade, you get: <html><head></head><body></body></html>, but if you render homepage.blade, you get:

        <h1>Hello, World</h1>


Isolates are only useful when using a live page update engine. Creating an isolate ensures that if data dependencies relating only to that isolate are updated, then only the part of the template within isolate will be re-rendered. All other parts of the Blade template will not be re-rendered.

Example (using Meteor):

- console.log("Rendering header...")
h1 This is a header
    p The current user is: #{Session.get("user")}

In the example above, "Rendering header..." will be printed to the console when the whole template is rendered, but if the reactive variable is updated using Session.set("user", ...), only the isolate will be re-rendered. In this case, nothing will print to the console.

Note: As with Blade functions, any blocks defined within an isolate will be deleted and unaccessible outside the isolate block.

See Reactivity isolation on the Meteor documentation for more details.


Chunks are deprecated as of Blade 3.0. You should use isolates instead.

Chunks are simply functions that return HTML. They behave a bit differently than conventional Blade functions.

Functions are called with call statements, and their rendered content is injected right into a template. You can also capture the HTML they render by outputting to a variable, as described above. Chunks, on the other hand, always return HTML, and they cannot be called using call statements. The only way to render a chunk is to call it via your code (see example below).

One reason you might define a chunk is to pass it to Meteor's Meteor.ui.chunk function; however, chunks can be used for other purposes, as well. You can also use chunks to work with Meteor.ui.listChunk. As of Meteor 0.4.0, the Meteor functions above no longer exist.


chunk header(text)
    h1= text

!= __.chunk.header("Hello")

The above example defines a named chunk header with one parameter. Then, the chunk is called by calling the __.chunk.header function. When defining a chunk, parameters are optional, and if you omit the name, the chunk is simply named last.

Another example:

    h1 Hello!
    != __.chunk.last()

renders as <div><h1>Hello!</h1></div>

If you override the templateNamespace compiler option, you will need to replace all instances of the double underscore (__) variable with the templateNamespace variable.


var blade = require('blade');

blade.compile(string, [options,] cb)

Compiles a Blade template from a string.

  • string is a string of Blade
  • options include:
    • filename - the filename being compiled (required when using includes or the cache option)
    • cache - if true, the compiled template will be cached (defaults to false)
    • debug - outputs debugging information to the console (defaults to false)
    • minify - if true, Blade generates a minified template without debugging information (defaults to true if cache option is set; false, otherwise) If UglifyJS is installed, Blade may automatically compress or prettify the template depending on whether minify is true or false.
    • includeSource - if true, Blade inserts the Blade source file directly into the compiled template, which can further improve error reporting, although the size of the template is increased significantly. (defaults to true if and only if process.env.NODE_ENV is "development" and minify is false; defaults to false, otherwise)
    • doctypes - an Object to specify additional doctypes or overwrite any built-in ones. This object is merged with blade.Compiler.doctypes
    • selfClosingTags - Array of self-closing tags to be used instead of blade.Compiler.selfClosingTags
    • filters - an Object to specify additional filters or overwrite any built-in ones. This object is merged with blade.Compiler.filters
    • templateNamespace - the name of the reserved variable in the view (defaults to two underscores: __). Other reserved names are listed here
    • basedir - the base directory where Blade templates are located. This option is primarily used by the Blade middleware to allow the Blade runtime to properly load file includes.
  • cb is a function of the form: cb(err, tmpl) where err contains any parse or compile errors and tmpl is the compiled template. If an error occurs, err may contain the following properties:
    • message - The error message
    • expected - If the error is a 'SyntaxError', this is an array of expected tokens
    • found - If the error is a 'SyntaxError', this is the token that was found
    • filename - The filename where the error occurred
    • offset - The offset in the string where the error occurred
    • line - The line # where the error occurred
    • column - The column # where the error occurred

Note: if there is a problem with the Blade compiler, or more likely, if there is a syntax error with the JavaScript code in your template, Node.js will not provide any line number or other information about the error. See issue #40 for more details.

You can render a compiled template by calling the function: tmpl(locals, cb)

  • locals are the local variables to be passed to the view template
  • cb is a function of the form function(err, html) where err contains any runtime errors and html contains the rendered HTML.

In addition, a compiled template has these properties and methods:

  • template - a function that also renders the template but accepts 3 parameters: tmpl.template(locals, runtime, cb). This simply allows you to use a custom runtime environment, if you choose to do so.
  • filename - the filename of the compiled template (if provided)
  • dependencies - an array of files that might be included by this template at runtime, relative to the path of this template
  • unknownDependencies - if true, this template uses dynamic filename includes and may include any file at any time.
  • reldir - the path to this template, relative to the base/root path
  • toString() - a function that converts the view template function into a string of JavaScript code. If you need a client-side template for example, you can use this function. UglifyJS is now used if you have it installed.

blade.compileFile(filename, [options,] cb)

Asynchronously compile a Blade template from a filename on the filesystem.

  • filename is the filename
  • options - same as blade.compile above, except filename option is always overwritten with the filename specified. There is also a synchronous option that will tell Blade to read and compile the file synchronously instead of asynchronously.
  • cb - same as blade.compile above

blade.renderFile(filename, options, cb)

Convenience function to compile a template and render it.

  • filename is the filename
  • options - same as blade.compileFile above. This object is also passed to the view, so it should also contain your view's local variables. A few reserved local variables are removed before passing the locals to the view.
  • cb - a function of the form function(err, html)

blade.middleware(sourcePath, options)

Express middleware for serving compiled client-side templates to the browser. For example, if you visit the URL "/views/homepage.blade" on your server, you can compile the view stored at sourcePath + "/homepage.blade"

  • sourcePath - the path on the server where your views are stored
  • options include:
    • mount - the URL path where you can request compiled views (defaults to "/views/")
    • runtimeMount - the URL path where the minified Blade runtime is served to the browser (defaults to "/blade/blade.js"). Use null to disable this functionality.
    • pluginsMount - the URL path where Blade plugins will be served to the browser (defaults to "/blade/plugins/"). Use null to disable this functionality.
    • compileOptions - options passed to blade.compile(). Defaults to:
    'cache': process.env.NODE_ENV == "production",
    'minify': process.env.NODE_ENV == "production",
    'includeSource': process.env.NODE_ENV == "development"


The compiler itself. It has some useful methods and properties.


Just generates the parse tree for the string. For debugging purposes only.

Example using the API:

var blade = require('blade');
blade.compile("string of blade", options, function(err, tmpl) {
    tmpl(locals, function(err, html) {

Here is a sample Express application that uses Blade for server-side and client-side templates:

var express = require('express'),
    blade = require('blade');
var app = express.createServer();
app.use(blade.middleware(__dirname + '/views') ); //for client-side templates
app.use(express.static(__dirname + "/public") ); //maybe we have some static files
app.set('views', __dirname + '/views'); //tells Express where our views are stored
app.set('view engine', 'blade'); //Yes! Blade works with Express out of the box!
app.get('/', function(req, res, next) {

Browser Usage

The Blade runtime should work on every browser, and since Blade provides an Express middleware for serving compiled templates to the browser (see above), rendering Blade templates in the browser is a breeze.

Once you have the middleware setup, you can now serve your compiled Blade views to the client. Simply include the /blade/blade.js file in your <script> tags, and then call blade.Runtime.loadTemplate.

blade.Runtime.loadTemplate(filename, cb)

  • filename - the filename of the view you wish to retrieve, relative to the sourcePath you setup in the Blade middleware.
  • cb - your callback of the form cb(err, tmpl) where tmpl is your compiled Blade template. Call the template like this: tmpl(locals, function(err, html) {...});

Your template will be stored in blade.cachedViews and will be cached until the user reloads the page or navigates to another page.

Yes, included files work, too. Like magic.

Example client-side JavaScript:

blade.Runtime.loadTemplate("homepage.blade", function(err, tmpl) {
    tmpl({'users': ['John', 'Joe']}, function(err, html) {
        console.log(html); //YAY! We have rendered HTML

Additionally, you can set blade.Runtime.options to control how the templates are loaded:

  • blade.Runtime.options.mount - the URL path where you can request compiled views (defaults to "/views/")
  • blade.Runtime.options.loadTimeout - the maximum number of milliseconds to wait before loadTemplate throws an error (defaults to 15 seconds).

As a side note, you can override the blade.Runtime.loadTemplate function with your own implementation.

Simple Example

The following Blade document ...

!!! 5
        title Blade
                - for(var i in nav)
                        a(href=nav[i])= i
            h1 Blade is cool

... compiles to this JavaScript function ...

function tmpl(locals,cb,__){__=__||[],__.r=__.r||blade.Runtime,__.func||(__.func={},__.blocks={},__.chunk={}),__.locals=locals||{};with(__.locals){__.push("<!DOCTYPE html>","<html",">","<head",">","<title",">","Blade","</title>","</head>","<body",">","<div",' id="nav"',">","<ul",">");for(var i in nav)__.push("<li",">","<a"),__.r.attrs({href:{v:nav[i],e:1}},__),__.push(">",__.r.escape(i),"</a>","</li>");__.push("</ul>","</div>","<div",' id="content"',' class="center"',">","<h1",">","Blade is cool","</h1>","</div>","</body>","</html>")}||__.r.done(__),cb(null,__.join(""),__)}

... now you call the function like this...

    'nav': {
        'Home': '/',
        'About Us': '/about',
        'Contact': '/contact'
}, function(err, html) {
    if(err) throw err;

... and you get this:

<!DOCTYPE html>
        <div id="nav">
                <li><a href="/">Home</a></li>
                <li><a href="/about">About Us</a></li>
                <li><a href="/contact">Contact</a></li>
        <div id="content" class="center">
            <h1>Blade is cool</h1>


Live UI

As of Blade 3.0, the Live UI plugin has been moved to another repository.

More information about the Live UI plugin is coming soon...

~~Blade provides a Live UI plugin that allows Blade to support live binding. Live binding provides automatic two-way synchronization between your models and views on a given web page. That is, when data in your Model is updated, the rendered Blade views on the client's browser are automatically updated with the new content, and similarly, when a Blade view is rendered in the browser, the Blade event handlers can update data in the model.

Complete documentation for the Live UI plugin (including several examples) can be found on the Live UI Plugin wiki page.

Eventually, the Live UI plugin might live in a separate repository and work for any templating language. More information coming soon...


This plugin is a prerequisite for the Live UI plugin if you plan on using Live UI in Internet Explorer 8.~~

Meteor Support

Blade also provides a Meteor smart package under the meteor directory. At the time of this writing, Blade is not a part of the Meteor core smart package list. The easiest thing to do right now is to symlink that directory into your Meteor packages directory like this:

ln -s /path/to/.../blade/meteor /path/to/.../meteor/packages/blade

Of course, the actual path where Blade and Meteor are installed on your system may vary. You need to replace the above command with the correct paths, as appropriate.

Then, execute meteor add blade in your Meteor project directory.

An Atmosphere smart package is also available.

To install Blade using Atmosphere, simply install Meteorite, navigate to your Meteor project directory, and type mrt add blade. Then, don't forget to run your project using mrt instead of meteor.

More documentation and examples for Meteor + Blade can be found on this wiki page.

Implementation Details


The Blade parser is built using PEG.js. Thanks to the PEG.js team for making this project much easier than I had anticipated! To modify the parser, simply change ./lib/parser/blade-grammer.pegjs, and the new parser will be automatically built the next time you run tests.

Running tests

To install all devDependencies, just do: npm link or install manually. To run tests, ensure devDependencies are installed, then run: npm test

Compiler-runtime relationship

Also, I'd like to mention here that the Blade compiler and Blade runtime are rather closely coupled. Unfortunately, that means that templates compiled with an older Blade compiler might not be compatible with a newer runtime and vice versa. To avoid issues, be sure that your Blade templates were compiled with the compiler of the same version as the runtime on which they will run. If you think this is too inconvenient, please feel free to complain, but I probably will ignore you. :)

File Includes

Included Blade templates MUST be loaded synchronously, and if this is not possible, an error will be thrown. Obviously, when rendering views on the server, this is not a problem since Node provides synchronous file system calls; however, on the client, it is only possible to include a file synchronously when the file is already in the browser's cache. When the name of the file to be included is known at compile-time (i.e. you are not using a dynamic filename include), the compiler will notify the Blade middleware of a particular view's dependencies. This allows the client-side template loader to also load and cache any dependent views in advance, preventing any issues from occurring. Nevertheless, when dynamic filename includes are used, the compiler has no way of determining which views will be included at runtime, and if a dynamically included view is not loaded into the browser's cache when the include statement is reached, the included view must be be loaded asynchronously and, as such, an error will be thrown.

Loading and compiling files synchronously may temporarily reduce your application's responsiveness, but because compiled views are often cached, this is not really much of an issue.

Event Handlers

Event handlers in Blade work by injecting the event handler function as an HTML comment directly before the bound element. Then, the appropriate event attribute (i.e. onclick, onchange, etc.) on the element is set to call blade.Runtime.trigger. The trigger function basically grabs the HTML comment, passes the contents through eval(), and binds the event handler directly to the element. This means that the event handlers work on templates rendered on the browser or on the server. Everything gets wired up the first time that the event occurs on the browser.

The Blade runtime also keeps track of any event handlers bound to a specific element by assigning each element an 'id' attribute, if necessary. When the view has finished rendering, the Blade runtime will pass a bunch of information (blocks, functions, or event handlers that were defined, etc.) to the 3rd (undocumented) argument of the render callback function. If you are rendering Blade templates on the browser, you can access the list of event handlers and bind the defined event handler directly to the element by looking up the element by its 'id' instead of letting the trigger function do its magic. The advantage of binding direclty to the defined event handler is that (thanks to closures) you can still reference the locals that were passed to your view and modify them, as needed... directly from your event handler. This allows your view code to automatically synchronize with your model, providing one-way view-to-model synchronization capabilties. Very cool! For examples of this and for more information, check out the Live UI plugin.


See the Benchmark wiki page for more information.


See the LICENSE.txt file.

Something went wrong with that request. Please try again.