melody-parser

This parser is an extensible parser for the Twig template language. It takes source code as input and produces an abstract syntax tree (AST) as output.

Usage

The easiest way to use this parser is the parse function. Simply pass it a string of Twig or Melody code:

const { parse } = require('melody-parser');
 
const code = '{% spaceless %} This is some Twig code {% endspaceless %}';
const abstractSyntaxTree = parse(code);

The parse function is a convenient façade function that does the instantiation of Lexer, Tokenizer, and Parser, and passes forward any options that may be provided. Example with options:

const abstractSyntaxTree = parse(code, {
    ignoreComments: false,
    ignoreHtmlComments: false,
    decodeEntities: true,
});

There is also a third parameter called extensions. It collects the third parameter and all parameters that come after it, so you can pass as many extensions as you want:

import { extension as coreExtensions } from 'melody-extension-core';
import customExtension from 'melody-extension-custom';
 
const abstractSyntaxTree = parse(
    code,
    {
        decodeEntities: true,
    },
    coreExtension,
    customExtension
);

If you have no options to pass, you can also omit them, and start passing extensions from the 2nd parameter on:

import { extension as coreExtensions } from 'melody-extension-core';
import customExtension from 'melody-extension-custom';
 
const abstractSyntaxTree = parse(code, coreExtension, customExtension);

Options

allowUnknownTags (false)

By default, the Melody parser throws an error when it encounters an unknown tag. This makes sense if you are preparing the generation of runnable code, because you don't know what to generate for unknown tags.

However, if you only want to parse a Twig file, unknown tags are fine. In this case, set allowUnknownTags to true. The parser will now generate GenericTwigTag nodes in the AST for Twig tags it does not know. For example, imagine this custom exit tag:

{% exit 404 %}

If we allow Melody to parse it generically, it will return the following AST node (some properties stripped and reordered for clarity):

Object {
  "type": "GenericTwigTag",
  "tagName": "exit",
  "parts": Array [
    Object {
      "type": "NumericLiteral",
      "value": 404,
    },
  ],
  "sections": Array []
}

We can see the node type GenericTwigTag, and the tagName property, which has the value exit. Let's look at the other two properties:

• parts is about what is going on inside the current Twig tag. In this example, there is only one more part, "404", which has been parsed as a NumericLiteral.
• sections is only relevant for "multiTags", i.e., sequences of Twig tags that belong together.

As an example of a multi tag, let's look at the nav tag from Craft CMS:

{% nav items as item %}
    <li>{{ item.name }}</li>
{% endnav %}

Here, nav and endnav belong together. It's desirable to have this pair represented as one AST node, not two. The result looks like this (again, optimized and shortened for better readability):

Object {
  "type": "GenericTwigTag",
  "tagName": "nav",
  "parts": Array [
    Object {
      "name": "items",
      "type": "Identifier",
    },
    Object {
      "name": "as",
      "type": "Identifier",
    },
    Object {
      "name": "item",
      "type": "Identifier",
    },
  ],
  "sections": Array [
    Object {
      "type": "SequenceExpression",
      "expressions": Array [
        Object {
          "name": "li",
          "type": "Element",
          "attributes": Array [],
          "children": Array [
            Object {
              // item.name representation
            },
          ],
        },
      ],
    },
    Object {
      "type": "GenericTwigTag",
      "tagName": "endnav",
      "parts": Array [],
      "sections": Array []
    },
  ],
}

Some things to note about this example:

• The nav tag node has 3 entries in its parts array: items, as, and item, which are all Identifier nodes. Of course, semantically, only two of them (items and item) are real identifiers, whereas as is a keyword. However, since Melody does not know this tag, it does not distinguish between the two.
• The nav tag has 2 entries in its sections array: A SequenceExpression representing everything between {% nav %} and {% endnav %}, and a GenericTwigTag representing the {% endnav %} tag. This shows us that everything up to and including the closing {% endnav %} is part of the nav tag. The GenericTwigTag representing the {% endnav %} tag is not an independent, top-level AST node, but part of the AST node representing the {% nav %} tag.

However, in order to get this outcome, we have to tell Melody that {% nav %} and {% endnav %} belong together. This is done through the multiTags option.

multiTags ({})

In the section on allowUnknownTags, we saw that some tags belong together, like {% nav %} and {% endnav %}. In order to make this known to Melody, we use the multiTags option:

{
    "multiTags": {
        "nav": ["endnav"],
        "switch": ["case", "default", "endswitch"]
    }
}

We can see that multiTags is an object. Its keys are tag names. These tag names must come first in a sequence of tags, e.g., {% nav %} always comes before {% endnav %}, therefore we must not configure multiTags like this, with the order reverted:

{
    "multiTags": {
        "endnav": ["nav"] // DON'T DO THIS!!!
    }
}

The values of the multiTags object are arrays containing the other tag names that can occur in the tag sequence started by the first tag name. Here, it's important that the tag name closing the sequence comes last. For example, when configuring for the switch tag in Craft CMS, endswitch has to come last:

{
    "multiTags": {
        "switch": ["case", "default", "endswitch"]
    }
}

Other than that, the order of the other tag names (here, case and default) is not important. Moreover, you can add as many multiTags entries as you want.

Note: If multiTags is non-empty, allowUnknownTags will automatically be set to true.

ignoreComments (true)

If set to true, Twig comments will not be part of the resulting abstract syntax tree (AST). Defaults to true.

ignoreHtmlComments (true)

If set to true, HTML comments will not be part of the resulting abstract syntax tree (AST). Defaults to true.

preserveSourceLiterally (false)

This option can be useful for pretty printing scenarios, hence it is turned off by default. If set to true, transformations like decoding of HTML entities or escape sequences will not be applied.

decodeEntities (true, deprecated, use preserveSourceLiterally instead)

Character references/entities like ‎ will be decoded if this is set to true (default). Otherwise, the string will be taken over verbatim to the AST.

ignoreDeclarations (true)

Declarations like <!DOCTYPE html> are not added to the parse tree if set to true.