Getting started with JSON Hyper-Schema
JSON Hyper-Schema is close to becoming an official spec, so let’s see what it’s all about.
JSON Hyper-Schema is a format for describing an API. You can automate a lot with a document that describes your API.
- API SDKs
- Documentation
- Tools such as API sandboxes or Postman collections to simplify the development process
- Contract tests to ensure your API doesn’t change unexpectedly
- Mock API servers to test your design before building the API
OpenAPI is another API description format, but whereas OpenAPI usually remains content with server-based uses, clients can use JSON Hyper-Schema directly. A client can use this schema to build their API requests including URLs, query parameters, and the request body. When an API changes, they simply have to update the schema and all clients will adapt accordingly. The client doesn’t have to write a single line of code.
In this article I’m going to show some JSON Hyper-Schema examples and walk you through what makes them tick. This information should give you a foundation to continue learning about JSON Hyper-Schema. In future articles I will expand on that foundation and help you describe your entire API.
But before we can discuss JSON Hyper-Schema, you need a basic understanding of JSON schema.
JSON Schema
JSON Schema describes JSON data. It’s like a database schema for JSON and can be used to validate a JSON instance before it’s sent to an API.
Here is an example schema for a blog post
{
"type": "object",
"properties": {
"id": {
"type": "number"
},
"title": {
"type": "string"
},
"urlSlug": {
"type": "string"
},
"body": {
"type": "string"
}
},
"required": ["id"]
}
And here is JSON data that matches that schema
{
"id": 5,
"title": "JSON Hyper-Schema",
"urlSlug": "json-hyper-schema",
"body": "My long post about JSON Hyper-Schema..."
}
Let’s go through each property in the schema
type
type
describes the data type such as string, object or number. You can find the whole list here. Each type has its own set of validation keywords that make up the rest of the schema. We’ll describe two validation keywords below.
properties
properties
is validation keyword for JSON objects. This properties object defines each valid property along with an additional level of validation schema. If one of the properties is an object, you can continue to nest that representation as deep as necessary.
required
The required
property is another validation keyword specific to objects. The value of required
should be an array of strings, where each string is a key in the properties
object. JSON data is not valid against this schema if any of the properties in the required
array are missing.
So using the example above, the following JSON data is valid because the title
and body
are optional
{
"id": 5
}
And this JSON data is not valid because id
is required
{
"title": "JSON Hyper-Schema"
}
Once your API consumers have a JSON schema, they can use one of the many JSON Schema libraries to validate their JSON data.
JSON Hyper-Schema
Now that we have a foundation of JSON Schema, let’s add the “Hyper” part. JSON Schema was built with expansion in mind, via a system called vocabularies. The spec defines this well, so here’s a quote.
A JSON Schema vocabulary is a set of keywords defined for a particular purpose.
JSON Schema core defines the general vocabulary. In the JSON Schema example above I described part of the validation vocabulary. In the following section I will describe part of the hypermedia vocabulary, which include everything needed to describe an API request.
Let’s start with an example. Here’s a JSON Hyper-Schema document for a blog post.
{
"type": "object",
"properties": {
"id": {
"type": "number"
},
"title": {
"type": "string"
},
"urlSlug": {
"type": "string"
},
"post": {
"type": "string"
}
},
"required": ["id"],
"base": "http://api.dashron.com/",
"links": [{
"rel": "self",
"href": "posts/{id}",
"templateRequired": ["id"]
}]
}
This hyper-schema starts with the blog post schema from earlier and adds two new new properties. These new properties are base
and links
.
base
base
is similar to the HTML <base> tag. The value should be a URI, which becomes the root of any relative URI in the JSON instance. For example, a base of http://dashron.com/api/
would be combined with the relative URI of users
and become http://dashron.com/api/users
.
links
links
is an array of Link Description Objects (LDOs) as described below.
Link Description Object (LDO)
An LDO contains all the information necessary to describe the actions available to your client. Here’s the LDO from the previous example.
{
"rel": "self",
"href": "posts/{id}",
"templateRequired": ["id"]
}
You might be expecting to see things like method
or requestBody
for HTTP requests, but you won’t find them. JSON Hyper-Schema doesn’t limit itself to any one protocol. JSON fans can use JSON Hyper-Schema to describe a file system as easily as an HTTP API.
Let’s dig into each property of the LDO above.
rel
rel
is a string that describes how the LDO relates to the JSON. The LDO in this example uses the relation self
. Because this schema describes a blog post, the self
relation describes how you access that blog post.
There are many different predefined relations you can use to describe your LDOs. Start by checking if your relation is defined in this list. If you can’t find what you need, JSON Hyper-Schema recommends you use a URI as your relation. If you need your relation to be human readable, check out the tag URI scheme for more flexibility.
Before we move on it’s important to point out that no two LDOs should share a relation. The relation is the defining information of an LDO and should be unique across every LDO.
href
href
is a URI template. URI templates let you describe the structure of your URLs.
"href": "posts/{id}"
JSON Hyper-Schema builds on top of URI templates by explaining how to populate the URI template variables.
Populating URI templates with JSON data
In the href
example earlier, JSON Hyper-Schema assumes the {id}
variable lines up with the id
property in your JSON schema. So our blog post has an id
of 5 and our final href
would be /posts/5
.
When all the variables in the template are populated, you check to see if the URI is a relative or absolute URI. Absolute URIs start with a URI scheme, such as http://
and should be considered complete URIs. Relative URIs do not start with a scheme, and should be resolved according to RFC 3986. In this case because our base URI ends with a /
and our relative URI does not start with a /
, we can simply append the relative URI onto the base
, giving us https://api.dashron.com/posts/5
.
If your URI variables don’t line up with your JSON properties you can use templatePointers
. This object allows you to map URI template variables to a part of your JSON instance. In the following example the pointer indicates that the id
variable should use the urlSlug
property.
"href": "/posts/{id}",
"templatePointers": {
"id": "/urlSlug"
}
A template pointer is a absolute or relative JSON pointer. JSON Pointers allow you to refer directly to sections of a JSON document, but are out of the scope of this article. I recommend learning more when you have the time!
But what if urlSlug
is not in your JSON? We won’t have this problem with id
because it is a required field, but urlSlug
is optional. When a property is missing JSON schema assumes the value is an empty string. In the above example our fully resolved url would become https://api.dashron.com/posts/
, which isn’t the URL we are describing. To avoid this problem use templateRequired
.
"templateRequired": ["id"]
templateRequired
is an array of strings, where each string is a mandatory URI template variable. If the JSON Hyper-Schema library can not fill the required URI template variables, the LDO is considered invalid and must be ignored by the client.
Populating URI templates with outside input
Instead of populating the URI from the JSON you can ask for outside input such as user submitted data. To do this, add the hrefSchema
property to define a validation schema for any outside variables.
Using the blog post example again, let’s say we want the mobile app to prompt the user for a blog id. Your hrefSchema
would look like this.
"href": "/posts/{id}",
"hrefSchema": {
"id": {
"type": "number"
}
}
In this example I copied the id
schema from earlier in the document. It’s hard to maintain duplicate data, so I’m going to introduce the final concept of this article. To ensure we have a single source of truth for our id
schema we can use schema references.
Schema References
Schema References are JSON objects that allow you to point to a schema. The schema could be this schema, an entirely different schema, or a portion of a schema. These objects will always contain a single property, $ref
with a URI value.
Let’s start with the simplest example, a reference to a different schema. The following schema reference points outside of our schema to another document. That document is identified by the URI https://schemas.dashron.com/users
.
{"$ref": "https://schemas.dashron.com/users"}
Let’s assume that the current schema is identified by https://schemas.dashron.com/posts
. We can reference the id
property of the current schema by including a JSON pointer in the fragment portion of the URI.
{"$ref": "https://schemas.dashron.com/posts#/properties/id"}
If you plan on referring to the current schema, you can leave out everything but the fragment portion of the URI. The above schema can be simplified to the following example.
{"$ref": "#/properties/id"}
And if you want to refer directly to the entire current schema, you can leave off the JSON pointer.
{"$ref": "#"}
This self-referencing schema reference is the most common schema reference.
Now that I’ve covered schema references, let’s clean up that hrefSchema
example. The example below points to the id
property of the current schema instead of duplicating it.
"href": "/posts/{id}",
"hrefSchema": {"$ref": "#/properties/id"}
What’s next?
With this information, you can start writing schemas for all the JSON in your API, along with their hypermedia links. In the next article I cover request and response bodies.
In the meanwhile try building a JSON Hyper-Schema for one of your API endpoints, and validating your JSON with one of the many JSON Schema libraries.
This is part one of a three-parter. Check out part two.