@hyperjump/browser

Hyperjump Browser

The hyperjump browser is an experimental generic hypermedia client. It aims to provide a uniform interface for working with hypermedia enabled media types. When you use a web browser, you don't interact with HTML, you interact with the UI that the HTML represents. The hyperjump browser aims to do the same except with data. It abstracts away the hypermedia so you can work data as if it's just plain JSON data without having to leave the browser.

The hyperjump browser allows you to plug in support for different media types, but it comes with support for application/reference+json. This media type is based on the JSON Reference I-D with some additions and improvements.

Installation

npm install @hyperjump/browser --save

Contributing

Tests

Run the tests

npm test

Run the tests with a continuous test runner

npm test -- --watch

Usage

The following is a quick set of examples with little explanation. See the JSON Reference section below for the theory behind JSON Reference.

GET http://json-reference.hyperjump.io/example1 HTTP/1.1
Accept: application/reference+json

HTTP/1.1 200 OK
Content-Type: application/reference+json

{
  "foo": "bar",
  "aaa": { "$ref": "#/foo" },
  "ccc": { "$ref": "#/aaa" },
  "ddd": {
    "111": 111,
    "222": { "$ref": "#/aaa/bbb" }
  },
  "eee": [333, { "$ref": "#/ddd/111" }],
  "fff": {
    "$id": "http://json-reference.hyperjump.io/example2",
    "abc": 123
  }
}

const JRef = require("@hyperjump/browser/json-reference");

(async () => {
  // Get a document by absolute URL
  const doc = await JRef.get("http://json-reference.hyperjump.io/example1", JRef.nil);

  // Get a document with a relative URL using another document as the context
  const aaa = await JRef.get("/aaa", doc);

  // Get the value of a document
  JRef.value(aaa); // => "bar"

  // Get the JSON Pointer for the document
  JRef.pointer(aaa); // => "/aaa"

  // Map over a document whose value is an array
  const eee = JRef.get("#/eee", doc);
  const types = await JRef.map((item) => JRef.value(item) * 2, eee); // => [666, 222];

  // Get the key/value pairs of a document whose value is an object
  const ddd = JRef.get("#/ddd", doc);
  await JRef.entries(ddd); // => [
                           //      ["111", JRef.get("#/ddd/111", doc)],
                           //      ["222", JRef.get("#/ddd/222", doc)]
                           //    ]

  // Apply operations as a pipeline that works with promises
  const doubleEee = JRef.pipeline([
    JRef.get("#/eee"),
    JRef.map((items) => JRef.value(items) * 2)
  ]);
  await doubleEee(doc); // => [666, 222]
}());

JSON Reference

History

JSON Reference is best known for its role in JSON Schema. Although it had an author in common with JSON Schema, JSON Reference started as an independent, standalone specification. Both JSON Schema and JSON Reference were abandoned by their authors before reaching RFC status. In 2016, a new group picked up the JSON Schema specification and eventually folded JSON Reference into JSON Schema.

With this implementation, I use JSON Reference draft-03 from the original authors as a starting point and evolve the concept from there. Therefore, the $ref and $id in this implementation IS NOT the same $ref and $id used in recent drafts of JSON Schema.

Documentation

To understand how this implementation works, you need to think about it like a document in a browser. Like HTML in a web browser, a JSON Reference document is identified by a URL and relative URLs within the document are resolved against that URL.

An HTTP message with Content-Type: application/reference+json should be interpreted as a JSON Reference document. This content is a JSON object that can be parsed with any RFC-7150 compliant JSON parser. The URL fragment used to identify the document should be interpreted as a JSON Pointer (RFC-6901).

Value

The "value" of a JSON Reference document is the result of applying the JSON Pointer in the URL fragment to the JSON message body. In the following example, the URL is http://json-reference.hyperjump.io/example#/foo, which means the fragment is /foo, and the "value" is "bar".

Request:

GET http://json-reference.hyperjump.io/example#/foo HTTP/1.1
Accept: application/reference+json

Response:

HTTP/1.1 200 OK
Content-Type: application/reference+json

{
  "foo": "bar"
}

$ref

In a JSON Reference document, the $ref property defines a reference to another document or a different part of the current document. The value of the $ref property is a string that defines a relative or absolute URL as specified by RFC-3986.

When the "value" is an object with a $ref property, it should follow the reference like following a link. In the following example the fragment points /aaa, which is a reference that points to /foo, and thus the "value" is "bar".

Request:

GET http://json-reference.hyperjump.io/example#/aaa HTTP/1.1
Accept: application/reference+json

Response:

HTTP/1.1 200 OK
Content-Type: application/reference+json

{
  "foo": "bar",
  "aaa": { "$ref": "#/foo" }
}

A $ref is a document boundary that JSON Pointers should not cross. $refs should not be followed in order to resolve the fragment's JSON Pointer.

$id

In a JSON Reference document, the $id property is a string that defines an absolute URL that identifies a document within the parent document. It's the inlined version of a $ref. This is a little like the HTTP/2 server push feature. It's sending additional documents with the request because we know the client is just going to request those documents next.

In the example below, the "value" of the document is 111.

Request:

GET http://json-reference.hyperjump.io/example#/foo HTTP/1.1
Accept: application/reference+json

Response:

HTTP/1.1 200 OK
Content-Type: application/reference+json

{
  "foo": {
    "$id": "http://json-reference.hyperjump.io/example2#/aaa",
    "aaa": 111
  }
}

An $id is a document boundary that JSON Pointers should not cross. A JSON Reference's fragment JSON Pointer should not point to a separate document inlined with $id.

Limitations

The problem with inlining $refs with $id is that we don't get the HTTP headers that describe important things like caching. An optional $headers keyword is being considered.