Value Objects - DDD w/ TypeScript
We cover this topic in The Software Essentialist online course. Check it out if you liked this post.
Also from the Domain-Driven Design with TypeScript article series.
In Domain-Driven Design, Value Objects are one of two primitive concepts that help us to create rich and encapsulated domain models.
Those concepts are Entities and Value Objects.
Value Objects are best understood by understanding how it's different from an Entity. Their main difference is in how we determine identity between two Value Objects and how we determine identity between two Entities.
Entity Identity
We use an Entity to model a domain concept when we care about the model's identity and being able to distinguish that identity from other instances of the model.
The way that we determine that identity helps us determine whether it's an Entity or a Value Object.
A common example is modeling a user.
In this example, we'd say that a User is an Entity because the way that we determine the difference between two different instances of a User is through it's Unique Identifier.
The Unique Identifier we use here is either a randomly-generated UUID or an auto-incremented SQL id that becomes a Primary Key that we can use for lookup from some persistence technology.
Value Objects
With Value Objects, we establish identity through the structural equality of two instances.
Structural Equality
Structural equality means that two objects have the same content. This is different from referential equality / identity which means that the two objects are the same.
To identify two Value Objects from each other, we look at the actual contents of the objects and compare based on that.
For example, there might be a Name property on the User Entity.
How can we tell if two Names are the same?
It's pretty much like comparing two strings, right?
"Nick Cave" === "Nick Cave" // true
"Kim Gordon" === "Nick Cave" // falseThis is easy.
Our User entity could look like this:
interface UserProps {
name: string
}
class User extends Entity<UserProps> {
get name (): string {
return this.props.name;
}
constructor (props: UserProps) {
super(props);
}
}This is OK, but it could be better. Lemme ask a question:
What if we wanted to limit the length of a user's name. Let's say that it can be no longer than 100 characters, and it must be at least 2 characters.
A naive approach would be to write some validation logic before we create an instance of this User, maybe in a service.
class CreateUserService {
public static createUser (name: string) : User{
if (name === undefined || name === null || name.length <= 2 || name.length > 100) {
throw new Error('User must be greater than 2 chars and less than 100.')
} else {
return new User(name)
}
}
}This isn't ideal. What if we wanted to handle Editing a user's name?
class EditUserService {
public static editUserName (user: User, name: string) : void {
if (name === undefined || name === null || name.length <= 2 || name.length > 100) {
throw new Error('User must be greater than 2 chars and less than 100.')
} else {
user.name = name;
// save
}
}
}- This isn't really the right place to be doing this.
- We've just repeated the same validation logic.
This is actually how a lot of projects start to spin out of scope. We end up putting too much domain logic and validation into the services, and the models themselves don't accurately encapsulate the domain logic.
We call this an Anemic Domain Model.
We introduce value object classes to strictly represent a type and encapsulate the validation rules of that type.
Value Objects
We had this before, a basic class for our User entity wiith a string-ly typed name property.
interface UserProps {
name: string
}
class User extends Entity<UserProps> {
get name (): string {
return this.props.name;
}
constructor (props: UserProps) {
super(props);
}
}If we were to create a class for the name property, we could co-locate all of the validation logic for a name in that class itself.
The upper bound (max length), the lower bound (min length), in addition to any algorithm that we wanted to implement in order to strip out whitespace, remove bad characters, etc- it could all go in here.
Using a static factory method and a private constructor, we can ensure that the preconditions that must be satisfied in order to create a valid name.
interface NameProps {
value: string
}
class Name extends ValueObject<NameProps> {
get value (): string {
return this.props.value;
}
// Can't use the `new` keyword from outside the scope of the class.
private constuctor (props: NameProps) {
super(props);
}
public static create (name: string) : Name {
if (name === undefined || name === null || name.length <= 2 || name.length > 100) {
throw new Error('User must be greater than 2 chars and less than 100.')
} else {
return new Name({ value: name })
}
}
}Then, in the User class, we'll update the name attribute in UserProps to be of type Name instead of string.
interface UserProps {
name: Name;
}
class User extends Entity<UserProps> {
get name (): Name {
return this.props.name;
}
private constructor (props: UserProps) {
super(props);
this.name = props.name;
}
public static create (props: IUser) {
if (props.name === null || props.name === undefined) {
throw new Error('Must provide a name for the user');
} else {
return new User(props);
}
}
}We apply the static factory method here as well.
Value Object class
Here's an example of a Value Object class.
import { shallowEqual } from "shallow-equal-object";
interface ValueObjectProps {
[index: string]: any;
}
/**
* @desc ValueObjects are objects that we determine their
* equality through their structrual property.
*/
export abstract class ValueObject<T extends ValueObjectProps> {
public readonly props: T;
constructor (props: T) {
this.props = Object.freeze(props);
}
public equals (vo?: ValueObject<T>) : boolean {
if (vo === null || vo === undefined) {
return false;
}
if (vo.props === undefined) {
return false;
}
return shallowEqual(this.props, vo.props)
}
}Check out the equals method. Notice that we use shallowEquals in order to determine equality. This is one way to accomplish structural equality.
When it makes sense, subclasses of this Value Object base class can also be extended to include convenience methods like greaterThan(vo?: ValueObject<T>) or lessThan(vo?: ValueObject<T>). It wouldn't make sense in this example, but it might if we were talking about something like LoggingLevels or BusinessRatings.
In future articles, we'll talk about:
- entity design
- better error handling technique for object creation
- moving anemic code out of services and into domain models
- writing DTOs to create data contracts
This is part of the Domain-Driven Design with TypeScript series. If this was useful to you, let me know in the comments & subscribe to the newsletter to get notified when new articles come out. Cheers!
More in this series so far..
An Introduction to Domain-Driven Design - DDD w/ TypeScript
We cover this topic in The Software Essentialist online course. Check it out if you liked this post.
Discussion
Liked this? Sing it loud and proud 👨🎤.
Stay in touch!
Enjoying so far? Join 15000+ Software Essentialists getting my posts delivered straight to your inbox each week. I won't spam ya. 🖖
View more in Domain-Driven Design
You may also enjoy...
A few more related articles
Want to be notified when new content comes out?
Join 15000+ other Software Essentialists learning how to master The Essentials of software design and architecture.
• Open sourced on
• Deployed on
19 Comments
Commenting has been disabled for now. To ask questions and discuss this post, join the community.
Yo your articles are actually the bomb, helping me out so much with my research project right now!
Woo! Happy to hear that, man :)
Really nice article. I have a question about the instantiation of the Name class, would you just inject it into the User class? Ex: User.create(Name.create(name)) or is there another DDD pattern to delegate its instantiation? Thanks
Thanks Matteo!
Good question. For the most part, yes! But you want to make sure that your Value Objects are valid. That can be remedied by using the Result<T> class + a private constructor + a factory method so that you never pass in an invalid Value Object to an Entity.
With respect to delegation, to reconstruct Domain entities and transform them between layers, I use the Data Mapper pattern. I use it to:
That's a good separation of concerns, specifically delegating a construct to be responsible for converting domain objects to the appropriate representation.
You'll often find that you'll need to create Domain Objects from HTTP controller requests as well in order to pass input values to application layer Use Cases.
For a Use Case like Create User (UserEmail, Name, Password), try the `Result<T>.combine([])` method to check your objects are correct before passing them into your Use Case.
Thanks for great explanation. What about case when we need to determine equality of passed and persisted passwords? Should we encapsulate hashing inside Password VO and check equality between recounsiled from persistence and passed passwords?
Absolutely. Check out an example here: https://github.com/stemmlerjs/ddd-forum/blob/master/src/modules/users/domain/userPassword.ts
And when you're reconstituting the `UserPassword` from persistence and turning it into a domain object again, you can signal to the value object that it's hashed.
```typescript
const userPasswordOrError = UserPassword.create({ value: raw.user_password, hashed: true });
```
Another example here: https://github.com/stemmlerjs/ddd-forum/blob/master/src/modules/users/mappers/userMap.ts
Great stuff in here! I'm going through all your content to get up to speed on DDD and implement parts of it into my project. I'm currently using TypeORM which allows me to define entities like this:
```ts
@Entity()
export class User {
@PrimaryGeneratedColumn()
id!: number;
@Column("text")
@IsName()
name!: string;
}
```
In the above example, you can define a validator function called `IsName` directly on the entity. Would you recommend still extracting this validation into its own class?
Hey Rahul!
Ah, that's pretty cool :)
Well, if it was possible using TypeORM, I'd recommend you create a new type/class called `Name` and locate all the validation logic for `Name` there instead using a factory method, because:
I hope that was helpful!
Great article!
A few questions:
-Why do you store the value in a separate "props" object? Why not store the value directly on the value object class?
-Do you recommend using VOs for most values in entities, or is this more for just values that need validation?
-Are there any significant performance impacts of wrapping values in value objects in your experience? It seems like it could cause higher memory usage if there are a lot of objects (say a large list of value objects).
Thanks, Nathan! Good questions, btw.
I use a "props" object as a form of encapsulation. Placing the values directly on the object is equivalent to defining both a getter and a setter for each value. We don't to do that because if we did, anyone would be able to change the internal value of our value objects, circumventing any validation logic we've defined in the static factory method. Instead, let value objects be immutable, and prevent them from being changed using the equals (=) syntax.
Two reasons to use VO's in my opinion.
1) For validation and encapsulating that logic, as you've identified.
2) Encoding business requirements. Read the section "Wrapping primitives with domain specific types to express a common language and prevent misuse" from Make Illegal States Unrepresentable.
I haven't had performance issues returning lots of value objects (I typically implement some form of pagination), but I have had performance issues returning Aggregates as read models.
When implementing DDD, some of us choose to use a single model for reads and writes, while some of us choose to implement CQRS and delegate a representation of a particular model for reads (value object) and one for writes (aggregate). Typically, the aggregate involves a LOT more information in order to protect model invariants, so if we had to reconstitute several of these for reads (which are more frequent), it's quite possible to get hit by performance constraints. Worth looking into.
Thanks for your good articles. Not only the contents are good, but also English sentences are easy to understand and help for me (I'm not good at English).
Hi,
Thanks for a very nice article. However I think there is a typo in constructor in Name class (name.ts).
Hey Khalil!
Thank you so much for this amazing content.
Few questions:
Looking forward for your answer and for more brilliant articles!
Very clear and concise article! Great job!
And I like your ValueObject base class. By the way: I always have to develop more or less the same base class and collections (ValueObjectMap, ValueObjectList, etc...) for value objects in my projects.
Have you open sourced a lib for this?
If not, do you mind if I create one and quote your article?
Thanks for your articles they are a great Help !
I have a question:
Is it possible (in typescript) to create dynamic Value Object (it's probably not call that ^^)
interface UserProps { firstname: LimitedText(2, 100); lastname: LimitedText(2, 50); }Where the only difference between the 2 value object would be the min and max of character.
Or do I have to do this
interface UserProps { firstname: Firstname; lastname: Lastname; }which is not very DRY beaucause FirstName & LastName are basically the same excepts for their min/max.
I would love to hear your thougths on that
Sooo good, thank you very much for this...
Thanks for this incredible article!
Greate article
Khalil this reads very well; great work here and thanks for sharing the knowledge.
I might be wrong, but I guess there is a copy-paste error in the domain/user.ts constructor.
Is this piece of code legit?
Btw, really nice serie of article, thanks.
Hey dude, your posts are amazing, im a newbie in oop and i appreciate your Amazon work so thanks for your time. Cheers
Thank you so much. I've implemented it in this way
export class Entity<Properties> { protected properties: Partial<Properties> = {} constructor(properties: Properties) { this.assignProperties<Properties>(properties) } protected assignProperties<T>(fields: T) { Object.entries(fields).forEach(([key, value]) => { ;(this.properties as any)[key] = value }) } }export class Student extends Entity<StudentProperties> { constructor(fields: StudentProperties) { // pass fields to the parent Entity class to load the properties globally :) super(fields) // now I have all my validations on a single place (UwU) I'm so happy with this fields.schoolEmail && Student.checkEmailFormat(fields.schoolEmail) }Thank you Khalif for this article. It's really helpful.
In my project middleware (infrastructure layer), I use an npm validation package (Joi) to validate all inputs before they get to the application/business layer. However, my business layer deals only with "primitive" types. I am realizing that it's wasn't a great choice as I am heading to duplicate some validations that I defined in joi package schemas.
Is Joi package a false good solution?
Is there a clean way to use both a validation package (joi) and value object/entity validations?
Thank you again!
Is very helpul your article. Tks for share your knoledge.