Temple is an abstraction and a framework for compiling templates to pure Ruby. It's all about making it easier to experiment, implement and optimize template languages. If you're interested in implementing your own template language, or anything else related to the internals of a template engine: You've come to the right place.

Have a look around, and if you're still wondering: Ask on the mailing list and we'll try to do our best. In fact, it doesn't have to be related to Temple at all. As long as it has something to do with template languages, we're interested: http://groups.google.com/group/guardians-of-the-temple.

• Source: http://github.com/judofyr/temple
• Bugs: http://github.com/judofyr/temple/issues
• List: http://groups.google.com/group/guardians-of-the-temple
• API documentation:
  • Latest Gem: http://rubydoc.info/gems/temple/frames
  • GitHub master: http://rubydoc.info/github/judofyr/temple/master/frames
• Abstractions: http://github.com/judofyr/temple/blob/master/EXPRESSIONS.md

Temple is built on a theory that every template consists of three elements:

• Static text
• Dynamic text (pieces of Ruby which are evaluated and sent to the client)
• Codes (pieces of Ruby which are evaluated and not sent to the client, but might change the control flow).

The goal of a template engine is to take the template and eventually compile it into the core abstraction:

 [:multi,
   [:static, "Hello "],
   [:dynamic, "@user.name"],
   [:static, "!\n"],
   [:code, "if @user.birthday == Date.today"],
   [:static, "Happy birthday!"],
   [:code, "end"]]

Then you can apply some optimizations, feed it to Temple and it generates fast Ruby code for you:

 _buf = []
 _buf << ("Hello #{@user.name}!\n")
 if @user.birthday == Date.today
   _buf << "Happy birthday!"
 end
 _buf.join

In Temple, an Sexp is simply an array (or a subclass) where the first element is the type and the rest are the arguments. The type must be a symbol and it's recommended to only use strings, symbols, arrays and numbers as arguments.

Temple uses Sexps to represent templates because it's a simple and straightforward data structure, which can easily be written by hand and manipulated by computers.

Some examples:

 [:static, "Hello World!"]

 [:multi,
   [:static, "Hello "],
   [:dynamic, "@world"]]

 [:html, :tag, "em", [:html, :attrs], [:static, "Hey hey"]]

NOTE: SexpProcessor, a library written by Ryan Davis, includes a Sexp class. While you can use this class (since it's a subclass of Array), it's not what Temple mean by "Sexp".

The idea behind Temple is that abstractions are good, and it's better to have too many than too few. While you should always end up with the core abstraction, you shouldn't stress about it. Take one step at a time, and only do one thing at every step.

So what's an abstraction? An abstraction is when you introduce a new types:

 # Instead of:
 [:static, "<strong>Use the force</strong>"]

 # You use:
 [:html, :tag, "strong", [:html, :attrs], [:static, "Use the force"]]

First of all, it means that several template engines can share code. Instead of having two engines which goes all the way to generating HTML, you have two smaller engines which only compiles to the HTML abstraction together with something that compiles the HTML abstraction to the core abstraction.

Often you also introduce abstractions because there's more than one way to do it. There's not a single way to generate HTML. Should it be indented? If so, with tabs or spaces? Or should it remove as much whitespace as possible? Single or double quotes in attributes? Escape all weird UTF-8 characters?

With an abstraction you can easily introduce a completely new HTML compiler, and whatever is below doesn't have to care about it at all. They just continue to use the HTML abstraction. Maybe you even want to write your compiler in another language? Sexps are easily serialized and if you don't mind working across processes, it's not a problem at all.

All abstractions used by Temple are documented in EXPRESSIONS.md.

A compiler is simply an object which responds a method called #call which takes one argument and returns a value. It's illegal for a compiler to mutate the argument, and it should be possible to use the same instance several times (although not by several threads at the same time).

While a compiler can be any object, you very often want to structure it as a class. Temple then assumes the initializer takes an optional option hash:

 class MyCompiler
   def initialize(options = {})
     @options = options
   end

   def call(exp)
     # do stuff
   end
 end

In Temple, a parser is also a compiler, because a compiler is just something that takes some input and produces some output. A parser is then something that takes a string and returns an Sexp.

It's important to remember that the parser should be dumb. No optimization, no guesses. It should produce an Sexp that is as close to the source as possible. You should invent your own abstraction. Maybe you even want to separate the parsers into several parts and introduce several abstractions on the way?

A filter is a compiler which take an Sexp and returns an Sexp. It might turn convert it one step closer to the core-abstraction, it might create a new abstraction, or it might just optimize in the current abstraction. Ultimately, it's still just a compiler which takes an Sexp and returns an Sexp.

For instance, Temple ships with {Temple::Filters::DynamicInliner} and {Temple::Filters::StaticMerger} which are general optimization filters which works on the core abstraction.

An HTML compiler would be a filter, since it would take an Sexp in the HTML abstraction and compile it down to the core abstraction.

A generator is a compiler which takes an Sexp and returns a string which is valid Ruby code.

Most of the time you would just use {Temple::Generators::ArrayBuffer} or any of the other generators in {Temple::Generators}, but nothing stops you from writing your own.

In fact, one of the great things about Temple is that if you write a new generator which turns out to be a lot faster then the others, it's going to make every single engine based on Temple faster! So if you have any ideas, please share them - it's highly appreciated.

When you have a chain of a parsers, some filters and a generator you can finally create your engine. Temple provides {Temple::Engine} which makes this very easy:

 class MyEngine < Temple::Engine
   # First run MyParser
   use MyParser

   # Then a custom filter
   use MyFilter

   # Then some general optimizations filters
   filter :MultiFlattener
   filter :StaticMerger
   filter :DynamicInliner

   # Finally the generator
   generator :ArrayBuffer
 end

 engine = MyEngine.new(strict: "For MyParser")
 engine.call(something)

You've ran the template through the parser, some filters and in the end a generator. What happens next?

Temple provides helpers to create template classes for Tilt and Rails.

 require 'tilt'

 # Create template class MyTemplate and register your file extension
 MyTemplate = Temple::Templates::Tilt(MyEngine, register_as: 'ext')

 Tilt.new('example.ext').render     # => Render a file
 MyTemplate.new { "String" }.render # => Render a string

You need at least Ruby 1.9.3 to work with Temple. Temple is published as a Ruby Gem which can be installed as following:

 $ gem install temple

• Slim
• Hamlit
• Faml
• Sal
• Temple-Mustache (Example implementation)
• Temple ERB example implementation (Temple::ERB::Template)
• WLang

Thanks to _why for creating an excellent template engine (Markaby) which is quite slow. That's how I started experimenting with template engines in the first place.

I also owe Ryan Davis a lot for his excellent projects ParserTree, RubyParser, Ruby2Ruby and SexpProcessor. Temple is heavily inspired by how these tools work.