tree-walk is a JavaScript library providing useful functions for traversing,
inspecting, and transforming arbitrary tree structures. It's based on the
walk
module that I wrote for Underscore-contrib.
The most basic operation on a tree is to iterate through all its nodes, which
is provided by preorder
and postorder
. They can be used in much the same
way as Underscore's 'each' function. For example, take a simple tree:
var tree = {
'name': { 'first': 'Bucky', 'last': 'Fuller' },
'occupations': ['designer', 'inventor']
};
We can do a preorder traversal of the tree:
var walk = require('tree-walk');
walk.preorder(tree, function(value, key, parent) {
console.log(key + ': ' + value);
});
which produces the following output:
undefined: [object Object]
name: [object Object]
first: Bucky
last: Fuller
occupations: designer,inventor
0: designer
1: inventor
A preorder traversal visits the nodes in the tree in a top-down fashion: first
the root node is visited, then all of its child nodes are recursively visited.
postorder
does the opposite, calling the visitor function for a node
only after visiting all of its child nodes.
This module provides versions of most of the
Underscore collection functions, with
some small differences that make them better suited for operating on trees. For
example, you can use filter
to get a list of all the strings in a tree:
var walk = require('tree-walk');
walk.filter(walk.preorder, _.isString);
Like many other functions in this module, the argument to filter
is a function
indicating in what order the nodes should be visited. Currently, only
preorder
and postorder
are supported.
Sometimes, you have a tree structure that can't be naïvely traversed. A good
example of this is a DOM tree: because each element has a reference to its
parent, a naïve walk would encounter circular references. To handle such cases,
you can create a custom walker by invoking walk
as a function, and passing
it a function which returns the descendants of a given node. E.g.:
var walk = require('tree-walk');
var domWalker = walk(function(el) {
return el.children;
});
The resulting object has the same functions as walk
, but parameterized
to use the custom walking behavior:
var buttons = domWalker.filter(walk.preorder, function(el) {
return el.tagName === 'BUTTON';
});
However, it's not actually necessary to create custom walkers for DOM nodes -- walk handles DOM nodes specially by default.
A parse tree is tree that represents the syntactic structure of a formal
language. For example, the arithmetic expression 1 + (4 + 2) * 7
might have the
following parse tree:
var tree = {
'type': 'Addition',
'left': { 'type': 'Value', 'value': 1 },
'right': {
'type': 'Multiplication',
'left': {
'type': 'Addition',
'left': { 'type': 'Value', 'value': 4 },
'right': { 'type': 'Value', 'value': 2 }
},
'right': { 'type': 'Value', 'value': 7 }
}
};
We can create a custom walker for this parse tree:
var walk = require('tree-walk');
var parseTreeWalker = walk(function(node) {
return _.pick(node, 'left', 'right');
});
Using the find
function, we could find the first occurrence of the addition
operator. It uses a pre-order traversal of the tree, so the following code
will produce the root node (tree
):
parseTreeWalker.find(tree, function(node) {
return node.type === 'Addition';
});
We could use the reduce
function to evaluate the arithmetic expression
represented by the tree. The following code will produce 43
:
parseTreeWalker.reduce(tree, function(memo, node) {
if (node.type === 'Value') return node.value;
if (node.type === 'Addition') return memo.left + memo.right;
if (node.type === 'Multiplication') return memo.left * memo.right;
});
When the visitor function is called on a node, the memo
argument contains
the results of calling reduce
on each of the node's subtrees. To evaluate a
node, we just need to add or multiply the results of the left and right
subtrees of the node.