Erector - User Guide

Make sure to check out the RDoc Documentation for more details on the API.

The Basics

The basic way to construct some HTML/XML with erector is to subclass Erector::Widget and implement a content method:

class Hello < Erector::Widget
  def content
    html do
      head do
        title "Hello"
      end
      body do
        text "Hello, "
        b "world!"
      end
    end
  end
end
<html>
  <head>
    <title>Hello</title>
  </head>
  <body>
  Hello, <b>world!</b>
  </body>
</html>

Once you have a widget class, you can instantiate it and then call its to_s method. If you want to pass in 'locals' (aka 'assigns'), then do so in the constructor's default hash. This will make instance variables of the same name, with Ruby's '@' sign.

class Email < Erector::Widget
  def content
    a @address, :href => "mailto:"
  end
end

>> Email.new(:address => "foo@example.com").to_s
=> "<a href="mailto:foo@example.com">foo@example.com</a>"

(If you want control over which locals are valid to be passed in to a widget, use the needs macro.)

Mixin

If all this widget stuff is too complicated, just do

include Erector::Mixin

and then call erector { } from anywhere in your code. It will make an inline widget for you, pass in the block, and call to_s on it. And if you pass any options to erector , like :prettyprint => true , it'll pass them along to to_s !

Examples:

erector { a "lols", :href => "http://icanhascheezburger.com/" }
=> "<a href=\"http://icanhascheezburger.com/\">lols</a>"

erector(:prettyprint => true) do
  ol do
    li "bacon"
    li "lettuce"
    li "tomato"
  end
end
=> "<ol>\n  <li>bacon</li>\n  <li>lettuce</li>\n  <li>tomato</li>\n</ol>\n" 

API Cheatsheet

code output
element('foo') <foo></foo>
empty_element('foo') <foo />
html <html></html>   …and likewise for all non-deprecated elements from the HTML 4.0.1 spec
b 'foo' <b>foo</b>
div { b 'foo' } <div><b>foo</b></div>
text 'foo' foo
text '&<>' &amp;&lt;&gt;   …all normal text is HTML escaped, which is what you generally want, especially if the text came from the user or a database
text raw('&<>') &<>   …raw text escapes being escaped
rawtext('&<>') &<>   …alias for text(raw())
text!('&<>') &<>   …another alias for text(raw())
div { text 'foo' } <div>foo</div>
div 'foo' <div>foo</div>
foo = 'bar'
div foo
<div>bar</div>
a(:href => 'foo.div') <a href="foo.div"></a>
a(:href => 'q?a&b') <a href="q?a&amp;b"></a>   …attributes are escaped like text is
a(:href => raw('&amp;')) <a href="&amp;"></a>   …raw strings are never escaped, even in attributes
a 'foo', :href => "bar" <a href="bar">foo</a>
text nbsp('Save Doc') Save&#160;Doc   …turns spaces into non-breaking spaces
text nbsp &#160;   …a single non-breaking space
text character(160) &#xa0;   …output a character given its unicode code point
text character(:right-arrow) &#x2192;   …output a character given its unicode name
instruct <?xml version="1.0" encoding="UTF-8"?>
comment 'foo' <!--foo-->
url 'http://example.com' <a href="http://example.com">http://example.com</a>
capture { div } <div></div>   …returns the block as a string, doesn't add it to the current output stream
div :class => ['a', 'b'] <div class="a b"></div>
javascript(
'if (x < y && x > z)
alert("don't stop");)
<script type="text/javascript">
// <![CDATA[
if (x < y && x > z) alert("don't stop");
// ]]>
</script>
jquery '$("p").wrap("<div></div>");' <script type="text/javascript">
// <![CDATA[
jQuery(document).ready(function($){
$("p").wrap("<div></div>");
});
// ]]>
</script>
join([widget1, widget2],
separator)
  …See examples/join.rb for more explanation

Lots more documentation is at the RDoc API pages especially for Erector::Widget so don't go saying we never wrote you nothin'.

Pretty-printing

Erector has the ability to insert newlines and indentation to make the generated HTML more readable. Newlines are inserted before and after certain tags.

To enable pretty-printing (insertion of newlines and indentation) of Erector's output, do one of the following:

Using Erector from Ruby on Rails

Your views are just ruby classes. Your controller can either call Rails' render :template method as usual, or directly instantiate the view class and call its content method.

For example:

app/controllers/welcome_controller.rb:
class WelcomeController < ApplicationController
  def index
    render :template => 'welcome/show'
  end
end
app/views/welcome/show.rb:
class Views::Welcome::Show < Erector::Widget
  def content
    html do
      head do
        title "Welcome page"
      end
      body do
        p "Hello, world"
      end
    end
  end
end

For Rails to find these .rb files during render :template , you must first either copy the erector source to vendor/plugins/erector , or add require 'erector' to config/environment.rb . You also should delete (or rename) any other view files with the same base name that might be getting in the way.

Currently there is only partial support for some standard Rails features like partials, layouts, assigns, and helpers. Check the erector Google Groups mailing list for status updates on these features.

Erector tool: Command-line conversion to and from HTML

To make Rails integration as smooth as possible, we've written a little tool that will help you erect your existing Rails app. The 'erector' tool will convert HTML or HTML/ERB into an Erector class. It ships as part of the Erector gem, so to try it out, install the gem, then run

erector app/views/foos/*.html.erb

or just

erector app/views

and then delete the original files when you're satisfied.

Here's a little command-line howto for erecting a scaffold Rails app:

rails foo
cd foo
script/generate scaffold post title:string body:text published:boolean

erector app/views/posts

mate app/views/posts
sleep 30 # this should be enough time for you to stop drooling

rm app/views/posts/*.erb
(echo ""; echo "require 'erector'") >> config/environment.rb
rake db:migrate
script/server
open http://localhost:3000/posts

On the erector-to-html side, pass in the --to-html option and some file names and it will render the erector widgets to appropriately-named HTML files. We're actually using erector to build this Erector documentation web site that you're reading right now. Check out the 'web' directory and the 'web' task in the Rakefile to see how it's done.

Layout Inheritance

Erector replaces the typical Rails layout mechanism with a more natural construct, the use of inheritance. Want a common layout? Just implement a layout superclass and inherit from it. Implement content in the superclass and implement template methods in its subclasses.

For example:

class Views::Layouts::Page < Erector::Widget
  def content
    html do
      head do
        title "MyApp - #{page_title}"
        css "myapp.css"
      end
      body do
        div :class => 'sidebar' do
          render_sidebar
        end
        div :class => 'body' do
          render_body
        end
        div :class => 'footer' do
          render_footer
        end
      end
    end
  end

  def render_sidebar
    a "MyApp Home", :href => "/"
  end

  def render_body
    p "This page intentionally left blank."
  end

  def render_footer
    p "Copyright (c) 2112, Rush Enterprises Inc."
  end
end
class Views::Faq::Index < Views::Layouts::Page
  def initialize
    super(:page_title => "FAQ")
  end

  def render_body
    p "Q: Why is the sky blue?"
    p "A: To get to the other side"
  end

  def render_sidebar
    super
    a "More FAQs", :href => "http://faqs.org"
  end
end

Notice how this mechanism allows you to...

all in a straightforward, easily understood paradigm (OO inheritance). (No more weird yielding to invisible, undocumented closures!)

To use this in Rails, declare layout nil in app/controllers/application.rb and then define your Page parent class as class Views::Layouts::Page in app/views/layouts as usual.

There's one trick you'll need to use this layout for non-erector view templates. Here's an example.

application.rb - The Erector layout superclass

class Views::Layouts::Application < Erector::Widget
  attr_accessor :content

  def content
    html do
      head { } # head content here
      # body content here
      body do
        text content
      end
    end
  end
end

application.mab - The markaby template (adjust for other appropriately templating technologies)

widget = Views::Layouts::Application.new(self)
widget.content = content_for_layout
self << widget.to_s

Here the abstract layout widget is used in a concrete fashion by the template-based layout. Normally, the content method would be implemented by subclassing widgets, but the layout template sets it directly and then calls to_s on the layout widget. This allows the same layout to be shared in a backward compatible way.

Inline Widgets

Instead of subclassing Erector::Widget and implementing a content method, you can pass a block to Erector.inline and get back a widget instance you can call to_s on. For example:

html = Erector.inline do
  p "Hello, world!"
end
html.to_s          #=> <p>Hello, world!</p>
This lets you define mini-widgets on the fly.

If you're in Rails, your inline block has access to Rails helpers if you pass a helpers object to to_s:

html = Erector.inline do
  image_tag("/foo")
end
html.to_s(:helpers => controller)          #=> <img alt="Foo" src="/foo" />

Note that inline widgets are usually redundant if you're already inside an Erector content method. You can just use a normal do block and the Erector methods will work as usual when called back from yield . Inline widgets get evaluated with instance_eval which may or may not be what you want. See the section on blocks in this user guide for more detail.

One extra bonus feature of inline widgets is that they can call methods defined on the parent class, even though they're out of scope. How do they do this? Through method_missing magic. (But isn't method_missing magic against the design goals of Erector? Yes, some would say so, and that's why we're reserving it for a special subclass and method. For Erector::Widget and subclasses, if you pass in a block, it's a plain old block with normal semantics.) But they can't directly access instance variables on the parent, so watch it.

Needs

Named parameters in Ruby are fun, but one frustrating aspect of the 'options hash' technique is that the code is less self-documenting and doesn't 'fail fast' if you pass in the wrong parameters, or fail to pass in the right ones. Even simple typos can lead to very annoying debugging problems.

To help this, we've added an optional feature by which your widget can declare that it needs a certain set of named parameters to be passed in. For example:

class Car < Erector::Widget
  needs :engine, :wheels => 4
  def content
    text "My #{@wheels} wheels go round and round; my #{@engine} goes vroom!"
  end
end
This widget will throw an exception if you fail to pass :engine => 'V-8' into its constructor. (Actually, it will work with any engine, but a V-8 is the baddest.)

See the rdoc for Widget#needs for more details. Note that as of version 0.7.0, using needs no longer automatically declares accessor methods.

Externals

Erector's got some nice tags, like script and style , that you can emit in the content method of your widget. But what if your widget needs something, say a JavaScript library, that should be included not in the main page, but inside the head section?

Externals are a way for your widget to announce to the world that it has an external dependency. It's then up to another widget to emit that dependency while it's rendering the head .

Here's an example:

class HotSauce < Erector::Widget
  external :css, "/css/tapatio.css"
  external :css, "/css/salsa_picante.css", :media => "print"
  external :js, "/lib/jquery.js"
  external :js, "/lib/picante.js"

  def content
    p :class => "tapatio" do
      text "esta salsa es muy picante!"
    end
  end
end
Then when Page emits the head it'll look like this:
<head>
  <meta content="text/html;charset=UTF-8" http-equiv="content-type" />
  <title>HotPage</title>
  <link href="/css/tapatio.css" media="all" rel="stylesheet" type="text/css" />
  <link href="/css/salsa_picante.css" media="print" rel="stylesheet" type="text/css" />
  <script src="/lib/jquery.js" type="text/javascript"></script>
  <script src="/lib/picante.js" type="text/javascript"></script>
</head>

It also collapses redundant externals, so if lots of your widgets declare the same thing (say, 'jquery.js'), it'll only get included once.

There's at least one drawback to this technique: since the externals get declared at class load time, then every external will be emitted on every page, even if that widget isn't on the page at all. This is not usually a problem, but it may lead to namespace collision, so be careful out there.

Page looks for the following externals:
:js included JavaScript file
:css included CSS stylesheet
:script inline JavaScript
:style inline CSS style
It might be a little difficult to remember the difference between :js and :script, and between :css and :style, so I'm thinking of maybe unifying them and looking at the content to determine whether it's inline or not. (Something simple like, if it includes a space, then it's inline.) Good idea? Let us know on the erector mailing list!

Blocks

Erector is all about blocks (otherwise known as closures). Unfortunately, there are some confusing aspects to working with blocks; this section aims to clarify the issues so if you find yourself stuck on an 'undefined method' or a nil instance variable, at least you'll have some context to help debug it.

There are basically three cases where you can pass a block to Erector:

1. To an element method

This is the normal case that provides the slick HTML DSL. In the following code:

class Person < Erector::Widget
  def content
    div do
      h3 @name
      p do
        b "Birthday: "
        span @birthday
      end
    end
  end
end

the blocks passed in to div and p are evaluated using normal yield semantics, and the @name and @birthday instance variables are evaluated in the context of the Person instance being rendered.

So far, so good.

2. To the constructor of an Erector::Widget

In this case you can build a widget "on the fly" and have it render whatever it wants, then call your block. This is useful for widgets like Form which want to wrap your HTML in some of their own tags.

class PersonActions < Erector::Widget
  needs :user
  def content
    div do
      widget(Form.new(:action => "/person/#{@user.id}", :method => "delete") do
        input :type => "submit", :value => "Remove #{@user.name}"
      end)
      widget(Form.new(:action => "/person/#{@user.id}/email", :method => "post") do
        b "Send message: "
        input :type => "text", :name => "message"
        input :type => "submit", :value => "Email #{@user.name}"
      end)
    end
  end
end

In this case, you will get two form elements, each of which has some boilerplate HTML for emitting the form element, emitting the hidden _method input tag in the case of the delete method, then calling back into your widget to emit the contents of the form. In this case, as above, the @user instance variable will be sought inside the calling widget (PersonActions) , not the called widget (Form) .

A quirk of this technique is that methods inside the block will be called on the calling widget, not the called widget. This doesn't cause any problems for element methods ( b and input above), but may be confusing if you want the block to be able to call methods on the target widget. In that case the caller can declare the block to take a parameter; this parameter will point to the nested widget instance.

widget(Form.new(:action => "/person/#{@user.id}", :method => "delete") do |f|
  span "This form's method is #{f.method}"
  input :type => "submit", :value => "Remove #{@user.name}"
end)

(As a variant of this case, note that the widget method can accept a widget class, hash and block, instead of an instance; in this case it will set the widget's block and this code:

widget Form, :action => "/person/#{@user.id}", :method => "delete" do
  input :type => "submit", :value => "Remove #{@user.name}"
end
will work the same as the version above.)

3. To the constructor of an Erector::InlineWidget

This is where things get hairy. Sometimes we want to construct a widget on the fly, but we're not inside a widget already. So any block we pass in will not have access to Erector methods. In this case we have a special subclass called Erector::InlineWidget which uses two magic tricks: instance_eval and method_missing to accomplish the following:

When using the mixin, you get an inline widget, so the above list of tricks applies.


One note for developers: when creating a widget like Form that needs to call back to its block, use the method call_block , which calls the block and passes in self as appropriate for both inline and normal widgets.