json-decoder allows TypeScript and JavaScript projects to adorn class declarations with JSON decoding decorators to support automatic decoding and marshalling of JSON objects to full prototype objects.
- Installation
- Reflect Metadata
- Simple Decoding
- Property Aliases
- Type Marshalling
- Custom Constructors
- JSON Schema Validation
Of course it can, and the CommonJS module loader of Node.js even makes it easy to use require('file.json') so even loading JSON is easy!
With ES6, JavaScript introduced classes. And TypeScript uses classes way more as a first class citizen. Well in JavaScript it's simply to adopt a class prototype, and in TypeScript interfaces give JSON types.
It seems like there's isn't too much utility in json-decoder? These existing mechanisms are just fine if they work for you, but there are limitations of JSON:
- You get everything, not just what you want (all those extra properties)
- JSON properties might not be named appropriately or desired for your class needs
- JSON properties are limited in types, no support for
Map,Set,URL,Datesor custom types - JSON property values might need coercing, where there is a
String, maybe aNumber, or an object needs to be a custom object - You may need a single property value pulled up from a sub-object property, or a single element of an array.
- Custom translation functions called to covert values
- And more...
To install the module use either yarn or npm
$ yarn add https://github.com/happycodelucky/json-decoder-node.git
$ npm install -S https://github.com/happycodelucky/json-decoder-node.git
Reflect metadata additions are not yet formally part of EcmaScript so are not available in JavaScript or Node.js. reflect-metadata is a package that must only be installed once per project. As such, it's is included here as a peer dependency.
To install in any project, use:
$ yarn add reflect-metadata
To install in any model, use:
$ yarn add --peer reflect-metadata
Simplest of all declarations is to declare a class as JSON decodable by using the @jsonDecoable class decorator. This is required to declare decoding support.
Prototype properties can be decorated with @jsonProperty, inheriting the prototype property name as the name of the JSON property to map the value to.
import { jsonDecodable, jsonProperty } from './json-decoder' // Decodable Account class @jsonDecodable() class Account { // Mapped 'id' property @jsonProperty public readonly id: number // Mapped 'name' property @jsonProperty public readonly name: string }
To decode an Account class, use JsonDecoder.decode with a JSON object (or JSON string) and the decodable class to decode, in this case Account
import { JsonDecoder } from 'json-decoder' import { Account } from './account' const json = { id: 1001, name: 'John Smith', } // Decode the above JSON for 'Account' class const account = JsonDecoder.decode<Account>(json, Account)
The class you are authoring and the JSON used as a source might not be a one to one match. You class may want to use different property names or even structure the representation different to the source JSON.
This is where "aliases" come in and the jsonPropertyAlias comes into play.
The following uses the decorator describing the JSON property accountId mapped to the class prototype property id.
import { jsonDecodable, jsonPropertyAlias } from './json-decoder' @jsonDecodable() class Account { // Alias 'id' to 'accountId' property @jsonPropertyAlias('account-id') public readonly id: number // ... }
Note the JSON object below is using account-id instead of id
import { JsonDecoder } from 'json-decoder' import { Account } from './account' const json = { 'account-id': 1001, name: 'John Smith', } const account = JsonDecoder.decode<Account>(json, Account)
The decoder will automatically map the id value based on the alias given in the @jsonPropertyAlias.
Simple property renaming does not help with restructuring the original JSON. The decoder can help here too with Key Paths alias. Key paths are dot/@-notion paths to JSON values.
If the desire was to flatten a JSON object's properties this can be achieved by using a path similar to that used in TypeScript/JavaScript itself
{ id: 1001, name: { title: 'Dr', fullName: 'John Smith', suffix: 'III', } }
The of interest name properties can be flatten
import { jsonDecodable, jsonPropertyAlias } from './json-decoder' @jsonDecodable() class Account { @jsonPropertyAlias('name.title') public readonly title: string @jsonPropertyAlias('name.fullName') public readonly fullName: string // ... }
The @-notation can be used to address indexes of arrays if needed.
{ id: 1001, name: { title: 'Dr', fullName: 'John Smith', suffix: 'III', }, profiles: [2001, 2002, 2451], }
The decode could map index 0 of profiles by using profiles@0. At the same time map all the profiles from the JSON.
import { jsonDecodable, jsonPropertyAlias, jsonProperty } from './json-decoder' @jsonDecodable() class Account { // ... @jsonPropertyAlias('profiles@0') public readonly primaryProfile: number @jsonProperty public readonly profiles: Array<number> // ... }
Key paths can be a long as the JSON supports. [email protected] will attempt to look up YAP, the end value. If at any point any sub-property returns undefined then undefined will be returned. If null is returned for any but the last property then undefined will be returned, else null will be returned.
@ and . are synonymous with one another (subject to change) but provide a more readable structure of intent. That is '.' sub-property, '@' index. a.b.c is the same as a@b@c.
As a convenience, you may use @index as an alias, which will attempt to map a source JSON object that is an array, at index to the class prototype property.
Using an example of a message protocol from a network request, the message is composed of a [header, body] structure.
[ { message: 'ACCOUNT' }, { name: 'John Smith' }, ]
Decoding the above message use @jsonPropertyAlias('@1.name'), where @1 represents the body of the message.
import { jsonDecodable, jsonPropertyAlias } from './json-decoder' @jsonDecodable() class AccountMessage { // ... @jsonPropertyAlias('@1.name') public readonly name: string // ... }
Type marshalling is a built-in feature of the decoder. Type marshalling allows a property of one type to be automatically converted into another. This allows more flexibility and implicit type conversion to happen automatically.
The @jsonType decorator allows the declaration of the class to marshal the property value to. Natively supported types are:
- Array
- Boolean
- Date (as timestamp or string)
- Map
- Number
- Set
- URL
If class passed to @jsonType() is also a @jsonDecodable class, then json-decoder will attempt to deserialize the object associated with the current @jsonProperty using the specified class.
import { jsonDecodable, jsonType } from './json-decoder' @jsonDecodable() class Account { // Use 'Number' marshalling @jsonProperty @jsonType(Number) public readonly id: number }
The above will attempt to convert a JSON property value of id into a Number. The marshalling attempts to perform a smart conversion based on the property in the JSON. For Numbers this could be Integers or Floats, and supports Base10, Base16 (Hex) and Base2 (Binary) number strings similar to JavaScript notation.
import { JsonDecoder } from 'json-decoder' import { Account } from './account' const json = { id: '0x3E9', name: 'John Smith', } const account = JsonDecoder.decode<Account>(json, Account)
jsonType can also be used to marshal iterable collections such as Array, Map and Set. To specify a collection with a specific type of element, @jsonType({ collection: Array | Map | Set, element: Type }) (an array of one type element).
{
id: 1001,
profiles: ['2001', '2002', '2451'],
}To marshal profiles into an Array of Number
import { jsonDecodable, jsonType } from './json-decoder' @jsonDecodable() class Account { // ... @jsonType({ collection: Array, element: Number }) public readonly profiles: Array<number> // ... }
element types can also be nested collections. For example:
import { jsonDecodable, jsonType } from './json-decoder' @jsonDecodable() class Account { // ... @jsonType({ collection: Map, element: { collection: Map, element: String }}) public readonly nestedMap: Map<string, Map<string, string>> // ... }
Array marshalling can also convert a single property value into an array. If the JSON object only had a single scalar property value instead, it will creating an Array of one element.
All element marshaling still apply here too
{
id: 1001,
profiles: '2001',
}With marshaling arrays, a JSON property with an array can be marshalled into a single property value. In this case only the first element of the array is taken.
{ id: 1001, profiles: ['2001', '2002', '2451'], }
// Marshal `profiles` into a single value @jsonType(Number) public readonly primaryProfile: number // ... @jsonType({ collection: Array, element: Number }) public readonly profiles: Array<number>
When decoded primaryProfile will be set to 2001 as a Number
Sometimes when marshaling a given attribute, one may wish to apply custom business logic to calculate the value that is assigned to the @jsonProperty. To accomplish this, a function may be optionally passed as a second parameter to @jsonType with a signature of:
(value: Type): OtherType
For example, if a source JSON document contains an unbounded numeric value but once deserialized, it should be limited to a maximum decimal value:
{ id: 1001, size: 9999, }
const MAX_VALUE = 1000.01 @jsonType(Number) public readonly id: number @jsonPropertyAlias('size') @jsonType(Number, (data: number): number => { return Math.min(MAX_VALUE, data) }) public readonly decimalSize: Number
Sometimes when instantiating a @jsonDecodable class, it becomes necessary to execute some custom initialization code in the class constructor. However, json-decoder will not use the constructor unless directed to do so. To cause the class constructor to get invoked when decode() is called, pass { useConstructor: true } to @jsonDecodable.
import { jsonDecodable, jsonType } from './json-decoder' @jsonDecodable({ useConstructor: true }) class Account { constructor() { super() // do custom initialization } @jsonProperty @jsonType(Number) public readonly id: number }
Similarly, if one is using inheritance to describe @jsonDecodable classes, it may be desirable to call decode on a base class (which may be abstract), and dynamically instantiate a derived class based on the contents of the JSON.
@jsonDecodable({ useConstructor: true }) abstract class BaseClass { @jsonProperty name: string @jsonProperty id: number @jsonDecoderFactory static decoderClassFactory(json: JsonObject): DerivedClass_1 | DerivedClass_2 { return json['some-attribute'] === 'some value' ? new DerivedClass_1() : new DerivedClass_2() } abstract toJSON(): JsonObject }
json-decoder supports validation of JSON documents using JSON Schema. To enable schema validation, add the @jsonSchema() decorator to your class definition.
import { jsonDecodable, jsonType } from './json-decoder' interface MyData { foo: number bar?: string } const schema: JSONSchemaType<MyData> = { $id: "http://schemas.example.com/my-data", $schema: "http://json-schema.org/draft-07/schema#", description: "This is my data.", type: "object", properties: { foo: {type: "integer"}, bar: {type: "string", nullable: true} }, required: ["foo"], additionalProperties: false } @jsonDecodable({ useConstructor: true }) @jsonSchema(schema) class Account { // Use 'Number' marshalling @jsonProperty public readonly foo: number @jsonProperty bar?: string }
When decode() is called, the schema validation will be executed automatically. If validation fails for any reason an Error will be thrown containing each of the validation errors encountered.