Skip to content
Generate JavaScript classes/code from JSON Schema
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
bindings Add built-in table renderer Sep 9, 2014
comparison Fixed RegExp bug Sep 9, 2014
source Fixed RegExp bug Sep 9, 2014
tests Fixed RegExp bug Sep 9, 2014
.gitignore Add tests Jun 17, 2014
.gitmodules Performs all type-checking Jun 17, 2014
Gruntfile.js Fixed RegExp bug Sep 9, 2014 Fixed RegExp bug Sep 9, 2014
index.html Fixed RegExp bug Sep 9, 2014
json-model.min.js Fixed RegExp bug Sep 9, 2014 Fixed RegExp bug Sep 9, 2014
package.json Refactor comparison into browser-compatible module Sep 9, 2014

JSON Models

This package handles JSON data and associated JSON Schemas. This includes fast schema validation/assignment, and a wrapper class that adds events plus HTML bindings for UI/display (HTML on server-side, DOM in browser with shared code).


On Node/CommonJS, use the 'json-model' package:

var JsonModel = require('json-model');

In the browser, it registers itself as the JsonModel global object.

Getting a validator

To get a validator for a given schema:

var validator = JsonModel.validator(schema);

var result = validator(data);
console.log(result.errors); // List of errors
console.log(result.schemas); // Map from JSON Pointer paths --> schema URLs
console.log(result.links); // Map: path --> links
console.log(result.missing); // Map: path --> missing schemas

If some schemas need to be fetched, then the validator will not be completely functional at first. You can supply a callback function to be notified when the validator is ready (which also supplies the same validator as a result):

JsonModel.validator(schema, function (error, validator) {

Setting the request function

This module has the ability to fetch schemas/data, but it needs you to supply an appropriate request function:

JsonModel.setRequestFunction(function (params, callback) {
	/* do whatever */
	callback(error, jsonData, headers);

The arguments to the callback are error, jsonData (the fetched data, JSON-decoded), and headers (an object representing the headers). headers may be omitted (e.g. when loading from a file). Non-JSON responses do not need to be supported.

Creating/opening data

You can create a JsonModel wrapper directly:

var model = JsonModel.create(jsonData, url, schemas, callback);

Everything except the initial value (jsonData) is optional, but if you supply schemas you must supply url as well (although it may be null). If callback is provided, it will be called (with two arguments error and model) after all relevant schemas have loaded. The model argument will be the same as the return value of create().

You can also open remote data: (hintSchemas is an optional set of schemas to use if the remote resource doesn't supply its own)'', hintSchemas, function (error, model) {...});

In both cases, the callback is only called when all the schemas have been loaded.

The schemas/hintSchemas arguments can be strings (URIs), objects (anonymous schemas), or arrays of strings/objects.

Model methods

The following methods are also available on wrapper objects:

Events and inspection

  • model.on(event, callback),[event, [listener]], model.once(event, callback), model.emit(event, ...) - event methods. addListener()/etc. variants are also present
  • model.errors([pathSpec], [includeFetchErrors]) - returns current validation errors. If the includeFetchErrors flag is set, then missing schemas (that encountered an error during fetching) are included as errors
  • model.path([pathSpec]) - a child model
  • model.pointer() - the JSON Pointer of this model relative to the document root

Generic value methods

  • model.jsonType() - the current basic type of the data (null/boolean/string/number/object/array)
  • model.get([pathSpec]) - gets the value from the model.
  • model.set([pathSpec], value) - gets the value from the model.
  • model.getHtml([pathSpec]) - gets the value, HTML-encoded
  • model.schemas([pathSpec]) - gets the schemas (URLs if known, or the schema itself for anonymous schemas)
  • model.hasSchema(url) - whether

Array methods

  • model.length() - array length
  • model.item(index) - a child item
  • model.items(callback) - iterate over the child items (function callback(itemModel, index) {...})
  • - maps the array value to a new array

Object methods

  • model.keys() - array length
  • model.prop(key) - a child property
  • model.props(callback) - iterate over the child properties (function callback(propModel, key) {...})
  • model.props(keys, callback) - iterate over a particular set of child properties, in order
  • model.mapProps(callback) - maps the object value to a new object
  • model.mapProps(keys, callback) - maps the object value to an array (corresponding to a particular set of keys)

For any method that takes an (optional) first pathSpec argument, this may either be a JSON Pointer (e.g. "/foo/bar"), or a property/index (e.g. "foo" or 5). If it is missing then the immediate value is returned.

Other utilities

  • - whether the supplied object is a JsonModel wrapper or not
  • JsonModel.schemasFetched() - whether all schemas have been fetched for the moment
  • JsonModel.whenSchemasFetched(callback) - callback is executed when all schemas have been fetched
  • JsonModel.extend(newMethods) - adds methods to the model prototype

Fast validation/assignment (when re-using schemas)

Schemas are compiled into validators (generating custom JS code), which has an up-front overhead but leads to much faster validation upon re-use.

Speed table

Here's a table of measured times for various validation setups (using the JSON Schema Test Suite) on Node:

SetupTime (ms)Relative speedTest scoreRepeats
json-model@0.2.24 (precompiled)0.41100%9769
json-model@0.2.24 (compile and validate)60.1158.2100%62
tv4 (validateResult)27.873.194.7%134
tv4 (validateMultiple)
json-model@old (sanity check)0.41100%9583

As you can see, the first time you compile a validator it is definitely slower than tv4. However, if you re-use that compiled validator then it is faster than tv4 by an order of magnitude. If you're going to be validating against the same schema multiple times, then this will probably end up faster.

Schema assignment

The result object you get back from a validator includes a schema property, which is a map from JSON Pointer paths to schema URIs:

	"valid": true,
	"errors": [],
	"schemas": {
		"": [""],
		"/foo": [""],
		"/foo/0": [""]
You can’t perform that action at this time.