@polyn/immutable

Define object schema's for validation, and construction of immutable objects

Usage no npm install needed!

<script type="module">
  import polynImmutable from 'https://cdn.skypack.dev/@polyn/immutable';
</script>

README

@polyn/immutable

tests Coverage Status

Define object schemas for validation, and construction of immutable objects. @polyn/immutable uses native JavaScript features (namely, Object.freeze) to make your objects immutable. It uses @polyn/blueprint to validate the schemas that you define, and also supports custom validators (i.e. if you prefer JSON Schemas).

Unlike Object.freeze, @polyn/immutable acts on your objects recursively: nested objects are frozen, as well as the values inside arrays.

@polyn/immutable also mitigates parameter/property pollution attacks at the model level: instances of immutable include only properties that exist on the schema, and only if they meet the validation you expressed.

Usage

Node

$ npm install --save @polyn/immutable

'use strict'

const { immutable } = require('@polyn/immutable')
const Product = immutable('Product', {
  id: 'string',
  title: 'string',
  description: 'string',
  price: 'decimal:2',
  type: /^book|magazine|card$/,
  metadata: {
    keywords: 'string[]',
    isbn: 'string?'
  }
})

const product = new Product({
  id: '5623c1263b952eb796d79e03',
  title: 'Swamplandia',
  description: 'From the celebrated...',
  price: 9.99,
  type: 'book',
  metadata: {
    keywords: ['swamp'],
    isbn: '0-307-26399-1'
  }
})

console.log(product)
// prints: ValidatedImmutable { id: '5623c1263b952eb796d79e03', ... }

try {
  const product2 = new Product({
    id: '5623c1263b952eb796d79e03',
    title: null, // this is required!
    description: 'From the celebrated...',
    price: 9.99,
    type: 'book',
    metadata: {
      keywords: ['swamp'],
      isbn: '0-307-26399-1'
    }
  })
} catch (e) {
  console.log(e.message)
  // will print Invalid Product: Product.title {string} is invalid
}

This library is bound to a specific version of @polyn/blueprint. If you install blueprint separately, and choose a different version, node will have two versions of blueprint in scope at runtime. All registrations will be performed on the version you install explicitly, which will make them unavailable to @polyn/immutable. Your best option is to just not do this.

Just install @polyn/immutable. @polyn/blueprint comes with it implicitly.

If there is a reason you want a different version of @polyn/blueprint than the one packaged with @polyn/immutable, checkout Using a Different Version of Blueprint in the cookbook.

Browser

$ npm install --save @polyn/immutable

<script src="./node_modules/@polyn/blueprint/dist/blueprint.min.js" />
<script src="./node_modules/@polyn/immutable/dist/immutable.min.js" />
<script>
  ((immutable) => {
    'use strict';

    const Product = immutable('Product', {
      id: 'string',
      title: 'string',
      description: 'string',
      price: 'decimal:2',
      type: /^book|magazine|card$/,
      metadata: {
        keywords: 'string[]',
        isbn: 'string?'
      }
    })

    const product = new Product({
      id: '5623c1263b952eb796d79e03',
      title: 'Swamplandia',
      description: 'From the celebrated...',
      price: 9.99,
      type: 'book',
      metadata: {
        keywords: ['swamp'],
        isbn: '0-307-26399-1'
      }
    })

    console.log(product)
    // prints: ValidatedImmutable { id: '5623c1263b952eb796d79e03', ... }
  })(window.polyn.immutable.immutable)
</script>

Updating Models With patch

The objects that immutable creates include patch on their prototype (it's not enumerable; doesn't show up in Object.keys). This function accepts an object, and will produce a new instance, favoring the values in that object over the existing values (like PATCH in REST).

const { immutable } = require('@polyn/immutable')
const { gt } = require('@polyn/blueprint')

const Person = immutable('Person', {
  firstName: 'string',
  lastName: 'string',
  age: gt(0)
})

const person = new Person({
  firstName: 'John',
  lastName: 'Doe',
  age: 21
})

console.log(person)
// prints { firstName: 'John', lastName: 'Doe', age: 21 }

const modified = person.patch({ age: 22 })

console.log(person)
// prints { firstName: 'John', lastName: 'Doe', age: 21 }
console.log(modified)
// prints { firstName: 'John', lastName: 'Doe', age: 22 }

Updating Models With toObject

Sometimes it's easier to make a copy, modify it, and then create a new instance ourselves, than it is to patch an existing instance. Here's how:

const { immutable } = require('@polyn/immutable')
const { gt } = require('@polyn/blueprint')

const Person = immutable('Person', {
  firstName: 'string',
  lastName: 'string',
  age: gt(0)
})

const person = new Person({
  firstName: 'John',
  lastName: 'Doe',
  age: 21
})

console.log(person)
// prints { firstName: 'John', lastName: 'Doe', age: 21 }

const mutable = person.toObject()
mutable.age = 22
const modified = new Person(mutable)

console.log(person)
// prints { firstName: 'John', lastName: 'Doe', age: 21 }
console.log(modified)
// prints { firstName: 'John', lastName: 'Doe', age: 22 }

To learn more about blueprint schemas and validation, checkout the @polyn/blueprint README

The Importance of Strict Mode

This library uses Object.freeze to make objects immutable, so it shares the behaviors of Object.freeze. If you 'use strict', and attempt to set a property on an immutable object, the operation will throw an error (usually a TypeError). If you don't 'use strict', and attempt to set a property on an immutable object, the operation will silently do nothing.

Nothing can be added to or removed from the properties set of a frozen object. Any attempt to do so will fail, either silently or by throwing a TypeError exception (most commonly, but not exclusively, when in strict mode).

-- MDN: Object.freeze()

Scope (It's not Managed)

While strict mode will help us identify cases where we attempt to mutate an object, JavaScript scope provides some avenues to possible confusion, and limitations we chose for this library.

For instance, what happens when we mutate a value that was originally defined outside of the scope of an immutable? What about mutable scope that is managed inside of a function that is attached to an immutable (i.e. getters and setters)? The expected behavior is illustrated below:

'use strict'

// define an immutable
const MakeNumber = immutable('MakeNumber', {
  makeOne: 'function',
  gettersAndSetters: {
    immutableTwo: 'number',
    get: 'function',
    set: 'function'
  },
  one: 'number'
})

// define variables that meet the immutable's schema
let makeOne = () => 1
let makeTwo = () => {
  let nonDeterministicNumber = 2

  return {
    immutableTwo: nonDeterministicNumber,
    get: () => nonDeterministicNumber,
    set: (num) => {
      nonDeterministicNumber = num
    }
  }
}
let one = 1

// create an instance of our immutable, using the variables
// we defined above
const makeNumber = new MakeNumber({
  makeOne,
  gettersAndSetters: makeTwo(),
  one
})

// each of `makeNumber.makeOne()`, and `makeNumber.one` returns 1
expect(makeNumber.makeOne()).to.equal(1)
expect(makeNumber.one).to.equal(1)

// each of `gettersAndSetters.immutableTwo`, and
// `gettersAndSetters.get()` returns 2
expect(makeNumber.gettersAndSetters.immutableTwo).to.equal(2)
expect(makeNumber.gettersAndSetters.get()).to.equal(2)

// mutate the original properties
makeOne = () => 2
one = 2
// note that this doesn't throw
// immutable doesn't mutate the values you pass to it

// now, each of `makeOne()`, and `one` in this
// immediate scope returns 2
expect(makeOne()).to.equal(2)
expect(one).to.equal(2)

// however, each of the immutable `makeNumber.makeOne()`,
// and `makeNumber.one` still returns 1
expect(makeNumber.makeOne()).to.equal(1)
expect(makeNumber.one).to.equal(1)

// execute a function that mutates the inner scope of gettersAndSetters
makeNumber.gettersAndSetters.set(3)

// The immutable `makeNumber.gettersAndSetters.immutableTwo`
// still returns 2 :yay:
expect(makeNumber.gettersAndSetters.immutableTwo).to.equal(2)

// but `makeNumber.gettersAndSetters.get()` now returns 3
// because it's a factory that exposes non-deterministic inner scope
expect(makeNumber.gettersAndSetters.get()).to.equal(3)

We have the option of allowing existing scope to be managed by the developer, or to fully clone all functions and manage them in this library. The latter is likely to produce more astonishment than the prior, so this library assumes you will manage your scopes. If you don't want inner scope to change, make it immutable, or simply don't change it (if you're using this library, there's a good chance you believe getters and setters are an anti-pattern anyway).

Using JSON Schema or Other Validators

Also see the Using JSON Schema with AJV example

You can register your own validator by creating a new instance of Immutable. The syntax for defining your own validator is:

/**
 * Creates a validator that will compare an input object to a schema
 * @param {string} name - the name of the model being validated
 * @param {object} schema - the schema this model should match
 */
function Validator (name, schema) {
  return {
    validate: (input) => {
      // throw if invalid
      // optionally: return { value: { /*...*/ } }
    }
  }
}

If your validate function returns an object with a value property, the value will be used instead of the input to construct the immutable instance. This allows you to set defaults, and intercept values

TypeScript Support

This library exports types. A brief example from strictly-typed-input is shown here. If you'd like to see more, the examples above, as well as others, are implemented in TypeScript in ts-examples.

import { gt, optional } from '@polyn/blueprint'
import { immutable, IValidatedImmutable } from '@polyn/immutable';
import { v4 as uuid } from 'uuid';

interface IPerson extends IValidatedImmutable<IPerson> {
  readonly id: string;
  readonly firstName: string;
  readonly lastName: string;
  readonly age: number;
}

// notice the `id` is optional
interface IPersonInput {
  readonly id?: string;
  readonly firstName: string;
  readonly lastName: string;
  readonly age: number;
}

const Person = immutable<IPerson, IPersonInput>('Person', {
  id: optional(UUID_REGEX).withDefault(uuid),
  firstName: 'string',
  lastName: 'string',
  age: gt(0)
});

const person: IPerson = new Person({
  firstName: 'John',
  lastName: 'Doe',
  age: 21
});

expect(UUID_REGEX.test(person.id)).to.equal(true);
expect(person.firstName).to.equal('John');
expect(person.lastName).to.equal('Doe');
expect(person.age).to.equal(21);
expect(() => new Person({
  firstName: '',
  lastName: '',
  age: -1
})).to.throw(Error, 'Invalid Person: expected `firstName` {string} to not be an empty string, expected `lastName` {string} to not be an empty string, expected `age` to be greater than 0');

Cookbook

Using JSON Schema with AJV

In the following example, we'll use ajv to validate JSON Schemas.

const { Immutable } = require('@polyn/immutable')

/**
 * Creates a validator that uses ajv to validate data against
 * the given JSON Schema
 * @param {string} name - the name of the model being validated
 * @param {object} schema - the JSON Schema this model should match
 */
function AjvValidator (name, schema) {
  const makeErrorText = (errors) => {
    return errors && errors
      .map((error) => `${name}${error.dataPath} ${error.message}`)
      .join(', ')
  }
  return {
    validate: (input) => {
      // allErrors: don't exit on first error
      const ajv = new Ajv({ allErrors: true })
      const isValid = ajv.validate(schema, input)

      if (!isValid) {
        throw new Error(`Invalid ${name}: ${makeErrorText(ajv.errors)}`)
      }
    }
  }
}

const { immutable } = new PolynImmutable({ Validator: AjvValidator })

const Person = immutable('Person', {
  $id: 'https://example.com/person.schema.json',
  $schema: 'http://json-schema.org/schema#',
  title: 'Person',
  type: 'object',
  properties: {
    firstName: {
      type: 'string',
      description: 'The person\'s first name.'
    },
    lastName: {
      type: 'string',
      description: 'The person\'s last name.'
    },
    age: {
      description: 'Age in years which must be equal to or greater than zero.',
      type: 'integer',
      minimum: 0
    }
  }
})

const person = new Person({
  firstName: 'John',
  lastName: 'Doe',
  age: 21
})

console.log(person)
// prints: ValidatedImmutable { firstName: 'John', lastName: 'Doe', age: 21 }

try {
  const person2 = new Person({
    firstName: 1,
    lastName: 2,
    age: -1
  })
} catch (e) {
  console.log(e.message)
  // prints ( `\` line breaks for readability):
  // Invalid Person: \
  // Person.firstName should be string, \
  // Person.lastName should be string, \
  // Person.age should be >= 0
}

NOTE this implementation isn't the fastest approach to using ajv, but it avoids collisions on the ajv.errors singleton

Using a Different Version of Blueprint

This isn't necessary if you just install @polyn/immutable, and avoid conflicting versions of @polyn/blueprint. If you have a reason for doing this, you can create your own Validator, and a new instance of PolynImmutable.

const { blueprint, gt, is } = require('@polyn/blueprint')
const { PolynImmutable } = require('@polyn/immutable')

const isBlueprint = (input) => {
  return is.object(input) &&
    is.string(input.name) &&
    is.function(input.validate) &&
    is.object(input.schema)
}

function Validator (name, schema) {
  let bp

  if (isBlueprint(name)) {
    // a blueprint was passed as the first argument
    bp = name
  } else {
    bp = blueprint(name, schema)
  }

  if (bp.err) {
    throw bp.err
  }

  return {
    validate: (input) => {
      const validationResult = bp.validate(input)

      if (validationResult.err) {
        throw validationResult.err
      }

      return validationResult
    }
  }
}

const { immutable } = new PolynImmutable({ Validator })

const Person = immutable('Person', {
  firstName: 'string',
  lastName: 'string',
  age: gt(0)
})

const person = new Person({
  firstName: 'John',
  lastName: 'Doe',
  age: 21
})

Schema Inheritance

const { blueprint, registerBlueprint } = require('@polyn/blueprint')
const { immutable } = require('@polyn/immutable')

const productBp = blueprint('Product', {
  id: 'string',
  title: 'string',
  description: 'string',
  price: 'decimal:2',
  type: /^book|magazine|card$/,
  metadata: {
    keywords: 'string[]'
  }
})

registerBlueprint('Author', {
  firstName: 'string',
  lastName: 'string'
})

const Product = immutable(productBp)
const Book = immutable('Book', {
  ...productBp.schema,
  ...{
    metadata: {
      ...productBp.schema.metadata,
      ...{
        isbn: 'string',
        authors: 'Author[]'
      }
    }
  }
})

const product = new Product({
  id: '5623c1263b952eb796d79e02',
  title: 'Happy Birthday',
  description: 'A birthday card',
  price: 9.99,
  type: 'card',
  metadata: {
    keywords: ['bday']
  }
})

const book = new Book({
  id: '5623c1263b952eb796d79e03',
  title: 'Swamplandia',
  description: 'From the celebrated...',
  price: 9.99,
  type: 'book',
  metadata: {
    keywords: ['swamp'],
    isbn: '0-307-26399-1',
    authors: [{
      firstName: 'Karen',
      lastName: 'Russell'
    }]
  }
})

console.log(product)
console.log(book)

Deep Equals With Functions

Sometimes, your schema might have a type that returns a function:

const Product = immutable('Product', {
  type: /^book$/i,
  comparableType: ({ input }) => {
    return {
      value: () => input.toLowerCase()
    }
  }
})

@polyn/immutable could put this function on Products prototype so the are comparable with deep equals, but this can lead to astonishing outputs because the scope of the function would be different than the other properties. Instead, there are two ways to make instances of ValidatedImmutable comparable when they have functions on them.

Functions on the Prototype

If all of the functions on the object are safe to be put on the prototype, you can set the functionsOnPrototype option to do that when you defined the schema:

const { expect } = 'chai'
const { immutable } = require('@polyn/immutable')

const Product = immutable('Product', {
  type: /^book$/i,
  comparableType: ({ input }) => {
    return {
      value: () => input.toLowerCase()
    }
  }
}, { functionsOnPrototype: true })

const p1 = new Product({ type: 'book' })
const p2 = new Product({ type: 'book' })

expect(p1).to.deep.equal(p2)

Remember - the prototype is shared across ALL INSTANCES of Product in this example. If your functions act on non-deterministic values they may produce side-effects that effect all instances.

Cast to Object Without Functions

If you don't want functions to be on the prototype, you can instead pass the removeFunctions option to toObjects in order to compare:

const { expect } = 'chai'
const { immutable } = require('@polyn/immutable')

const Product = immutable('Product', {
  type: /^book$/i,
  comparableType: ({ input }) => {
    return {
      value: () => input.toLowerCase()
    }
  }
})

const p1 = new Product({ type: 'book' })
const p2 = new Product({ type: 'book' })

expect(p1.toObject({ removeFunctions: true }))
  .to.deep.equal(p2.toObject({ removeFunctions: true }))

Why @polyn/immutable

This won't be a case for immutability - that's been made many times, elsewhere. Why use this library for immutability instead of another one?

I generally want objects to meet two criteria:

  • to be immutable
  • to be valid

This library deals with both of those problems in a way that's easier to grok than other languages or libraries that I've used. Unlike strictly typed languages, with JavaScript we can define the type, and the parameters beyond type that make an object, or property valid in a single location, and this library makes that trivial.