travel/admin/node_modules/css-tree/docs/traversal.md

8.0 KiB
Raw Blame History

AST traversal

AST traversal API is provided by walk() method.

var csstree = require('css-tree');
var ast = csstree.parse('.a { color: red; }');

csstree.walk(ast, function(node) {
    console.log(node.type);
});
// StyleSheet
// Rule
// SelectorList
// Selector
// ClassSelector
// Block
// Declaration
// Value
// Identifier

The facts you should know about walk() internals:

  • Method uses structure field value of every node type to define the way how to iterate the nodes:
    • A function-iterator is generating for every node type.
    • Node's properties iterates in the order it defined in structure (reverse option can invert an order).
    • Properties that are not defined in structure are ignoring (doesn't interate).
    • An exception is possible when a tree is not following to expected structure (it may happen if AST was built outside the CSSTree parser or transformed in a wrong way). In case you are not sure about correctness of a tree structure, you can use try/catch or check the tree with csstree.lexer.validateStructure(ast) before iterate it.
  • Only children fields can contain a list of nodes. A list of nodes should be represented as a List instances. But for certain cases, children can be an array. Since List provides a similar to Array API, traversal can work in most cases, but without any guarantee. Therefore usings arrays in AST is not recomended, use them on your own risk.

walk(ast, options)

Method visits each node of passed tree in a natural way and calls a handler for each one. It takes two arguments: a root node (ast) and an object (options). In simple case, it can take a function (handler) instead of options (walk(ast, fn) is equivalent to walk(ast, { enter: fn })).

Options:

enter

Type: function or undefined
Default: undefined

Handler on node entrance, i.e. before any nested node is processed.

var csstree = require('css-tree');
var ast = csstree.parse('.a { color: red; }');

csstree.walk(ast, {
    enter: function(node) {
        console.log(node.type);
    }
});
// StyleSheet
// Rule
// SelectorList
// Selector
// ClassSelector
// Block
// Declaration
// Value
// Identifier

In case options has a single enter field, it can replaced for the handler passed as a value for enter, i.e. walk(ast, { enter: fn })walk(ast, fn).

Handler receives a three arguments:

  • node the AST node a walker entering to
  • item node wrapper, that contains references to prev and next nodes in a list, and data reference for the node
  • list is a reference for the list; it's useful for list operations like remove() or insert()

NOTE: If children is an array, the last two arguments are index and array, like for Array#forEach() or Array#map() methods.

const csstree = require('css-tree');
const ast = csstree.parse(`
    .a { foo: 1; bar: 2; }
    .b { bar: 3; baz: 4; }
`);

// remove declarations with `bar` property from the tree
csstree.walk(ast, function(node, item, list) {
    if (node.type === 'Declaration' && node.property === 'bar' && list) {
        // remove a declaration from a list it
        list.remove(item);
    }
});

console.log(csstree.generate(ast));
// .a{foo:1}.b{baz:4}

NOTE:

  • item and list are not defined for nodes that are not in a list. Even Declaration can be outside of any list in case it is a root of tree or a part of @supports prelude, e.g. @supports (bar: 123) { ... }. Therefore, it's recomended to check item or list are defined before using of it (those values both are defined or both are undefined, so it's enough to test one of them)
  • Only List instances are safe for tree transformations such as node removal. In case you perform such operations, you can ensure that all children in a tree is a List instances by calling csstree.fromPlainObject(ast) before traversal.
  • It's better to use visit option when possible to reach better performance

Context (this) for a handler is an object with a references to the closest ancestor nodes:

  • root refers to AST root node (actually it's a node passed to walk() method)
  • stylesheet refers to StyleSheet node, usually it's a root node
  • atrule refers to closest Atrule node if any
  • atrulePrelude refers to AtrulePrelude node if any
  • rule refers to closest Rule node if any
  • selector refers to SelectorList node if any
  • block - refers to closest Block node if any
  • declaration refers to Declaration node if any
  • function refers to closest Function, PseudoClassSelector or PseudoElementSelector node if current node inside one of them
const csstree = require('css-tree');
const ast = csstree.parse(`
    @import url(import.css);
    .foo { background: url('foo.jpg'); }
    .bar { background-image: url(bar.png); }
`);

// collect all urls in declarations
const urls = [];
csstree.walk(ast, function(node) {
    if (this.declaration !== null && node.type === 'Url') {
        const value = node.value;

        if (value.type === 'Raw') {
            urls.push(value.value);
        } else {
            urls.push(value.value.substr(1, value.value.length - 2));
        }
    }
});

console.log(urls);
// [ 'foo.jpg', 'bar.png' ]

leave

Type: function or undefined
Default: undefined

The same as enter handler but invokes on node exit, i.e. after all nested nodes are processed.

var csstree = require('css-tree');
var ast = csstree.parse('.a { color: red; }');

csstree.walk(ast, {
    leave: function(node) {
        console.log(node.type);
    }
});
// ClassSelector
// Selector
// SelectorList
// Identifier
// Value
// Declaration
// Block
// Rule
// StyleSheet

visit

Type: string or null
Default: null

Invokes a handler for a specified node type only.

var csstree = require('css-tree');
var ast = csstree.parse('.a { color: red; } .b { color: green; }');

csstree.walk(ast, {
    visit: 'ClassSelector',
    enter: function(node) {
        console.log(node.name);
    }
});

// example above is equal to
csstree.walk(ast, {
    enter: function(node) {
        if (node.type === 'ClassSelector') {
            console.log(node.name);
        }
    }
});

The traveral for some node types can performs faster (10-15 times depending on the CSS structure), because some subtrees may to be skipped since they can't contain a node of specified type (e.g. Rule can't be used inside of Declaration, so declaration's subtree can be exclude from traversal path). Fast traversal is supported for node types:

  • Atrule
  • Rule
  • Declaration

NOTE: When fast traversal is applied, some nodes may not be reached in case of an incorrect location in the tree. That's may happen if AST was built outside the CSSTree parser or transformed in a wrong way. If you need to be 100% sure that every node of type will be visited (even in wrong position), don't use visit option and test node type by your own.

reverse

Type: boolean
Default: false

Inverts a natural order of traversal of nodes. To achieve this, the following actions are performed:

  • children nodes are iterated in reverse order (from last to first)
  • properties are iterated in reverse order (according to structure definition of node)
  • enter and leave handlers are swapped
var assert = require('assert');
var csstree = require('css-tree');
var ast = csstree.parse('.a { color: red; }');

var natural = [];
csstree.walk(ast, {
    enter: function(node) {
        natural.push('enter ' + node.type);
    },
    leave: function(node) {
        natural.push('leave ' + node.type);
    }
});

var reverse = [];
csstree.walk(ast, {
    reverse: true,
    enter: function(node) {
        reverse.push('enter ' + node.type);
    },
    leave: function(node) {
        reverse.push('leave ' + node.type);
    }
});

// will be truthy
assert.deepEqual(
    reverse,
    natural.reverse()
);