Data Modeling

Overview

We use MongoDB for persistence with @nestjs/mongoose. The combination gives us schema flexibility alongside type-safe DTOs. Our tutorial resource (Task) demonstrates the recommended patterns below, but apply the same principles to production collections.

Schemas

Use @Schema() + @Prop() decorators to define models.
Enable timestamps to track created/updated dates ({ timestamps: true }).
Prefer explicit field types (e.g., @Prop({ type: String, required: true, trim: true })).
Add defaults for booleans and enums to keep writes predictable.

@Schema({ timestamps: true })
export class Task {
  @Prop({ type: String, required: true, trim: true })
  title!: string;

  @Prop({ type: String, required: false, trim: true })
  description?: string;

  @Prop({ type: Boolean, default: false, index: true })
  done!: boolean;
}

export type TaskDocument = Task & Document;
export const TaskSchema = SchemaFactory.createForClass(Task);

Indexing strategy

Add indexes for frequent query filters (done, createdAt ranges).
Use compound indexes for combinations (ownerId + status).
For text search, use MongoDB Atlas search or define text indexes with caution; they impact write performance.
Periodically run db.tasks.getIndexes() in non-prod to audit index drift.

TaskSchema.index({ done: 1, updatedAt: -1 });
TaskSchema.index({ title: 1 }, { unique: false, collation: { locale: "en", strength: 2 } });

DTOs and validation

Map DTOs with class-validator + class-transformer to enforce input shape.
Keep DTOs separate from persistence models (e.g., CreateTaskDto, UpdateTaskDto).
Use @IsString(), @IsOptional(), @IsBoolean() etc. to guarantee data quality.
Apply ValidationPipe globally with { whitelist: true, transform: true } in main.ts.

export class CreateTaskDto {
  @IsString()
  @IsNotEmpty()
  title!: string;

  @IsOptional()
  @IsString()
  description?: string;
}

export class UpdateTaskDto {
  @IsOptional()
  @IsString()
  title?: string;

  @IsOptional()
  @IsBoolean()
  done?: boolean;
}

Relationships & references

Prefer storing foreign keys as ObjectIds (Types.ObjectId).
Populate references lazily to avoid large payloads; use .populate() only where needed.
For denormalized views, create dedicated read models or use MongoDB aggregation pipelines.

Environment awareness

Development uses local Docker or Atlas (see Local Development).
Production uses managed clusters; keep connection strings in secrets (see Configuration & Environment).
Always include read/write user roles scoped per environment.

Migrations

MongoDB migrations can be handled via scripts or tools like migrate-mongo.
Document schema changes in PR descriptions and update relevant tutorial content when field requirements change.