Twig Internals

Twig is very extensible and you can hack it. Keep in mind that you should probably try to create an extension before hacking the core, as most features and enhancements can be handled with extensions. This chapter is also useful for people who want to understand how Twig works under the hood.

How does Twig work?

The rendering of a Twig template can be summarized into four key steps:

• Load the template: If the template is already compiled, load it and go to the evaluation step, otherwise:
  • First, the lexer tokenizes the template source code into small pieces for easier processing;
  • Then, the parser converts the token stream into a meaningful tree of nodes (the Abstract Syntax Tree);
  • Finally, the compiler transforms the AST into PHP code.
• Evaluate the template: It means calling the display() method of the compiled template and passing it the context.

The Lexer

The lexer tokenizes a template source code into a token stream (each token is an instance of \Twig\Token, and the stream is an instance of \Twig\TokenStream). The default lexer recognizes 13 different token types:

• \Twig\Token::BLOCK_START_TYPE, \Twig\Token::BLOCK_END_TYPE: Delimiters for blocks ({% %})
• \Twig\Token::VAR_START_TYPE, \Twig\Token::VAR_END_TYPE: Delimiters for variables ({{ }})
• \Twig\Token::TEXT_TYPE: A text outside an expression;
• \Twig\Token::NAME_TYPE: A name in an expression;
• \Twig\Token::NUMBER_TYPE: A number in an expression;
• \Twig\Token::STRING_TYPE: A string in an expression;
• \Twig\Token::OPERATOR_TYPE: An operator;
• \Twig\Token::PUNCTUATION_TYPE: A punctuation sign;
• \Twig\Token::INTERPOLATION_START_TYPE, \Twig\Token::INTERPOLATION_END_TYPE: Delimiters for string interpolation;
• \Twig\Token::EOF_TYPE: Ends of template.

You can manually convert a source code into a token stream by calling the tokenize() method of an environment:

$stream = $twig->tokenize(new \Twig\Source($source, $identifier));

As the stream has a __toString() method, you can have a textual representation of it by echoing the object:

echo $stream."\n";

Here is the output for the Hello {{ name }} template:

TEXT_TYPE(Hello )
VAR_START_TYPE()
NAME_TYPE(name)
VAR_END_TYPE()
EOF_TYPE()

$twig->setLexer($lexer);

The Parser

The parser converts the token stream into an AST (Abstract Syntax Tree), or a node tree (an instance of \Twig\Node\ModuleNode). The core extension defines the basic nodes like: for, if, … and the expression nodes.

You can manually convert a token stream into a node tree by calling the parse() method of an environment:

$nodes = $twig->parse($stream);

Echoing the node object gives you a nice representation of the tree:

echo $nodes."\n";

Here is the output for the Hello {{ name }} template:

\Twig\Node\ModuleNode(
  \Twig\Node\TextNode(Hello )
  \Twig\Node\PrintNode(
    \Twig\Node\Expression\NameExpression(name)
  )
)

$twig->setParser($parser);

The Compiler

The last step is done by the compiler. It takes a node tree as an input and generates PHP code usable for runtime execution of the template.

You can manually compile a node tree to PHP code with the compile() method of an environment:

$php = $twig->compile($nodes);

The generated template for a Hello {{ name }} template reads as follows (the actual output can differ depending on the version of Twig you are using):

/* Hello {{ name }} */
class __TwigTemplate_1121b6f109fe93ebe8c6e22e3712bceb extends Template
{
    protected function doDisplay(array $context, array $blocks = [])
    {
        // line 1
        echo "Hello ";
        echo twig_escape_filter($this->env, (isset($context["name"]) ? $context["name"] : null), "html", null, true);
    }

    // some more code
}

$twig->setCompiler($compiler);