parseley

parseley

Parser for CSS selectors, based on nearley.

---

Goals / features

• Convert CSS selector strings into objects that are easy to work with;

• Serialize back if needed;

• Get specificity for free;

• Code is easy to understand and maintain.

Non-goals

• Top performance;

• Extra permissivity;

• To-the-letter CSS spec implementation.

These are great but only as long as it doesn't come in conflict with actual goals.

Install

> npm i parseley

Usage example

const parseley = require('parseley');
const util = require('util');

const str = 'div#id1 > .class1[attr1]';

const ast = parseley.parse1(str);
console.log(util.inspect(ast, { breakLength: 45, depth: null }));

const serialized = parseley.serialize(ast);
console.log(`Serialized: '${serialized}'`);

Example output

{ type: 'compound',
  list:
   [ { type: 'class',
       name: 'class1',
       specificity: [ 0, 1, 0 ] },
     { type: 'attrPresence',
       name: 'attr1',
       namespace: null,
       specificity: [ 0, 1, 0 ] },
     { type: 'combinator',
       combinator: '>',
       left:
        { type: 'compound',
          list:
           [ { type: 'tag',
               name: 'div',
               namespace: null,
               specificity: [ 0, 0, 1 ] },
             { type: 'id',
               name: 'id1',
               specificity: [ 1, 0, 0 ] } ],
          specificity: [ 1, 0, 1 ] } } ],
  specificity: [ 1, 2, 1 ] }
Serialized: 'div#id1>.class1[attr1]'

Documentation

• Functions
• AST types
• Snapshots

Motivation and inspiration

Package	Hits	Misses
parsel	Sensible AST; specificity calculation; cool name	Not friendly to node.js; based on regex
css-what and css-select	The idea to process complex selectors in right-to-left order	css-select is a solution for a different problem compared to what I needed; css-what produces only a list of tokens
scalpel	Introduced me to nearley parsing toolkit	AST it produces is very far from what I can use
css-selector-parser	Configurable and lightweight	Again, AST is far from my needs

Input reference

https://www.w3.org/TR/selectors-4/#grammar

https://www.w3.org/TR/css-syntax-3/#token-diagrams

Terminology used in this project is more or less consistent to the spec, with some exceptions made for clarity. The term "type" is way too overloaded in particular, the term "tag" is used where appropriate instead.

Any pseudo elements are left for possible future implementation. I have no immediate need for them and they require some careful consideration.

Output AST

Consistency: overall AST shape is always the same. This makes client code simpler, at least for a certain processing tasks.

For example, always use compound selectors, even when there is only one simple selector inside.

Comma-separated selectors might not be needed for every use case. So there are two functions - one can parse commas and always returns the top-level list regardless of the comma presence in a particular selector, and the other can't parse commas and returns a compound selector AST directly.

Complex selectors are represented in the way that makes the left side to be an another condition on the right side element. This was made with the right-to-left processing direction in mind. One consequence of this is that there is no such thing as a "complex selector" node in the AST hierachy, but there are "combinator" nodes attached to "compound selector" nodes.

All AST nodes have their specificity computed (except the top-level list of comma-separated selectors where it doesn't really make sense).

Changelog

Available here: CHANGELOG.md.

Roadmap

• add pseudo-classes and pseudo-elements support.

Share your use cases in issues so I can get a better idea where to move.

Changelog

Version 0.7.0

• Switched to TypeScript;
• Added type definitions for AST;
• Hybrid package (ESM, CommonJS);
• Renamed sort() to normalize() in order to better reflect what it does;
• Replaced compareArrays() with compareSpecificity() and compareSelectors() - more sensible API;
• Generated documentation.