DOM 3 XPath implementation for JavaScript objects.

npm install jsel
1 downloads in the last week
6 downloads in the last month


DOM 3 XPath implementation for JavaScript object data.

Written by Ali Chamas site.

Based on the npm module "xpath" originally written by Cameron McCormack (blog) with thanks to Yaron Naveh (blog).

Full documentation available at www.jselement.org.



Install locally with npm:

npm install jsel

Then require(..) module.

var jsel = require('jsel');
var dom = jsel(data);


To run jsel in the browser, just include the jsel.js file from the local jsel npm module installation into your web page.

<script src='jsel.js'></script>

You will receive a global jsel object. Use it like you would a require('jsel') object.

var dom = jsel(data);


Jsel allows you to apply XPath expressions against your JavaScript data objects, and return values from them.

Jsel supports the following XPath functions: last, position, count, match, replace, id, name, string, concat, starts-with, contains, substring-before, substring-after, substring, string-length, normalize-space, translate, boolean, not, true, false, lang, number, sum, floor, ceiling, round.

For example, say you had the following data.

var dom = jsel({
    title: 'abc',
    children: [
            foo: 'bar'
    subData: {
        foo: 555,
        foo2: 'bar2'

The following expressions would all be true.

dom.select('count(//*)') === 5;
dom.select('@title') === 'abc';
dom.select('//children/*[1]/@foo') === 'bar';
dom.select('count(//@foo)') === 2;
dom.select('//@foo[2]') === 555;
dom.select('count(//children//*)') === 2;
dom.select('//children/*[2]') === 'val';
dom.select('name(//children/*[2])') === 'string';
dom.select('name(*/*[2])') === 'subData';
dom.select('*/children/*[2]/text()') === 'val';
dom.selectAll('//@foo') === ['bar', 555];

Use dom.select(..) to return a single value, and dom.selectAll(..) to return a result set. You can return scalar results from XPath expressions, such as the count(..) function.

Jsel will return you the actual node values from your data. Think of it like turning any JavaScript object into a walkable DOM.


Why Schemas?

Jsel allows you to define a set of callbacks which resolve the following properties of each node in your data as required by the XPath engine:

  • nodeName - {string | null} the element name of the node
  • childNodes - {array | null} the children of the node
  • attributes - {object | null} a key/value object of the node's attributes
  • nodeValue - { | null}* the whole value of the node

These properties make up the structure of your data, according to XPath. You have total control of how to interpret these values from each node in your dom as they are being required. Jsel takes care of caching, so the values are only asked for once. This builds a virtual tree of walkable nodes for the XPath engine whenever you select or selectAll an expression, and will be cached for subsequent selections of that DOM.

Default Schema

Jsel has default schema built in that convert all string or number values to attributes, and all object or array values to named children. Therefore you don't even need to create a schema if you like, it all just works out of the box.

Defining Schemas

Defining your own schema allows you to completely design how your data is defined, and therefore searchable by XPath. For example, you could define a schema to handle CSV data, and pass a CSV string to jsel so you could find expressions like //line or //line[1]/field[2].

To provide your own custom schema pass an object to dom.schema(..) with all or some of the following keys (any keys not passed will use default schema handler).

    /** @returns {string|null} the element name of this current node */
    nodeName: function(node) {/*...*/},

    /** @returns {array|null} an array with the children of this node */
    childNodes: function(node) {/*...*/},

    /** @returns {object|null} an object with key values to represent the nodes attributes */
    attributes: function(node) {/*...*/},

    /** @returns {*|null} any value to represent the nodes value.
    * you can use text() in the expression to select the value */
    nodeValue: function(node) {/*...*/}

It's ok to return null from any function. That would be the same as not defining an adapter for that property and just means that there will no way to match that property in an expression.

It's even possible to create a temporary array or object structure for either childNodes and/or attributes and return that if you would like to simplify your expressions and aggregate your data.

Reuse and Share Schemas

The great thing about schemas is that once you've written one to support a data type, you can reuse and share that schema every time you access data of that type. You can distribute that schema so that other users can handle that data type too.

For example, you could write a schema to unify and normalise the ast tree of UglifyJS and esprima. Your schema (and mappings) would abstract both structures and you could use a common expression syntax to search both.

Writing Schemas for Uniform Data

If your data is uniform you could use the same logic for each node in your schema. Since each node will have a common structure, you can simplify how each property handler in the schema works for all nodes.

Writing Schemas for Variable Data

If your data has variable data schema, you might need to use conditional tests inside your schema such as if or switch statements. Since each node will vary, you'd need to test each node and handle each property accordingly.

User Nodes

You can create virtual nodes which only exist to help walk your data, these are called User Nodes. You create these nodes in your schema definitions to wrap parts of your data during expression evaluation, so that you can test for them during subsequent calls to the schema properties. You would usually do this in your childNodes property handler.

To create a User Node, just call jsel.Node(nodeName, value) and pass the node name and it's value. For example you could wrap a plain string and call it a data node so that you could test for that node later, and create more virtual types from that.


Jsel allows you to use regular expressions against your nodeName and attribute node values. This means you can use shorthand expressions, or map one expression to many nodeNames. This becomes very powerful when simplifying or abstracting your nodes.

To provide your own custom mappings, just pass an object to dom.map(..) with key value pairs of find and replace regular expressions. The key matches against your expression, the value matches against the node (nodeName or attribute).

For example, we can tell the dom that when we write "e" in our expressions, match it to "someLongElementName" nodes.

    'e': 'someLongElementName'

var result = dom.selectAll('//e');
// will return any nodes with name "someLongElementName"

We could also tell the dom that when we write "person", match it any node with "male" or "female" nodeName values.

    'person': 'male|female'

var result = dom.selectAll('//person');
// will return any nodes with nodeName "male" or "female"

We could even tell the dom that when we write "person" or "human", match it any node with "male" or "female" nodeName values.

    'person|human': 'male|female'

var result = dom.selectAll('//human');
// will return any nodes with nodeName "male" or "female"
// as would..
var result = dom.selectAll('//person');


Install mocha globally

npm install -g mocha

From the jsel module root, run mocha with spec reporter option

mocha -R spec

To run tests in the browser, open the tests/index.html file in your browser. They are the same tests.

npm loves you