주요 콘텐츠로 건너뛰기

Excessive Entities

The entities layer in Feature-Sliced Design is one of the lower layers that's primarily for business logic. That makes it widely accessible — all layers except for shared can access it. However, its global nature means that changes to entities can have a widespread impact, requiring careful design to avoid costly refactors.

Excessive entities can lead to ambiguity (what code belongs to this layer), coupling, and constant import dilemmas (code scattered across sibling entities).

How to keep entities layer clean

0. Consider having no entities layer

You might think that your application won't be Feature-Sliced if you don't include this layer, but it is completely fine for the application to have no entities layer. It doesn't break FSD in any way, on the contrary, it simplifies the architecture and keeps the entities layer available for future scaling. For example, if your application acts as a thin client, most likely it doesn't need entities layer.

What are thick and thin clients?

Thick vs. thin client distinction refers to how the application processes data:

  • Thin clients rely on the backend for most data processing. Client-side business logic is minimal and involves only exchanging data with the backend.
  • Thick clients handle significant client-side business logic, making them suitable candidates for the entities layer.

Keep in mind that this classification is not strictly binary, and different parts of the same application may act as a "thick" or a "thin" client.

1. Avoid preemptive slicing

In contrast to previous versions, FSD 2.1 encourages deferred decomposition of slices instead of preemptive, and this approach also extends to entities layer. At first, you can place all your code in the model segment of your page (widget, feature), and then consider refactoring it later, when business requirements are stable. Remember: it is much easier and safer to move something into entities later, than refactor code inside entities that can affect any upper level slice functionality.

2. Avoid Unnecessary Entities

Do not create an entity for every piece of business logic. Instead, leverage types from shared/api and place logic in the model segment of a current slice. For reusable business logic, use the model segment within an entity slice while keeping data definitions in shared/api:

📂 entities
📂 order
📄 index.ts
📂 model
📄 apply-discount.ts // Business logic using OrderDto from shared/api
📂 shared
📂 api
📄 index.ts
📂 endpoints
📄 order.ts

3. Exclude CRUD Operations from Entities

CRUD operations, while essential, often involve boilerplate code without significant business logic. Including them in the entities layer can clutter it and obscure meaningful code. Instead, place CRUD operations in shared/api:

📂 shared
📂 api
📄 client.ts
📄 index.ts
📂 endpoints
📄 order.ts // Contains all order-related CRUD operations
📄 products.ts
📄 cart.ts

For complex CRUD operations (e.g., atomic updates, rollbacks, or transactions), evaluate whether the entities layer is appropriate, but use it with caution.

4. Store Authentication Data in shared

Prefer shared layer to creating a user entity for authentication data, such as tokens or user DTOs returned from the backend. These are context-specific and unlikely to be reused outside authentication scope:

  • Authentication responses (e.g., tokens or DTOs) often lack fields needed for broader reuse or vary by context (e.g., private vs. public user profiles).
  • Using entities for auth data can lead to cross-layer imports (e.g., entities into shared) or usage of @x notation, complicating the architecture.

Instead, store authentication-related data in shared/auth or shared/api:

📂 shared
📂 auth
📄 use-auth.ts // authenticated user info or token
📄 index.ts
📂 api
📄 client.ts
📄 index.ts
📂 endpoints
📄 order.ts

5. Minimize Cross-Imports

FSD permits cross-imports via @x notation, but they can introduce technical issues like circular dependencies. To avoid this, design entities within isolated business contexts to eliminate the need for cross-imports:

Non-Isolated Business Context (Avoid):

📂 entities
📂 order
📂 @x
📂 model
📂 order-item
📂 @x
📂 model
📂 order-customer-info
📂 @x
📂 model

Isolated Business Context (Preferred):

📂 entities
📂 order-info
📄 index.ts
📂 model
📄 order-info.ts

An isolated context encapsulates all related logic (e.g., order items and customer info) within a single module, reducing complexity and preventing external modifications to tightly coupled logic.