Repository Pattern

Overview

This skill helps implement the repository pattern used in this workout tracker application. The pattern provides a clean abstraction over Dexie (IndexedDB) with type-safe interfaces, consistent error handling, and standardized CRUD operations.

Architecture Overview

Layered Approach:

code

Interfaces → Implementations → Provider → Public API
    ↓              ↓              ↓           ↓
db/interfaces  db/impl/dexie  db/provider  db/index

Flow:

•Define repository interface in db/interfaces.ts
•Implement with Dexie in db/implementations/dexie/[entity].ts
•Register in factory provider (db/implementations/dexie/index.ts)
•Export public getter (db/index.ts)
•Use in features via getEntityRepository()

Core Workflow

Follow these 6 steps when creating a new repository:

Step 1: Define Interface

Location: src/db/interfaces.ts

Define the repository interface with standard CRUD methods:

typescript

export type EntityRepository = {
  getAll(): Promise<ReadonlyArray<DbEntity>>
  getById(id: string): Promise<DbEntity | undefined>
  create(entity: Omit<DbEntity, 'id' | 'createdAt'>): Promise<DbEntity>
  update(id: string, updates: Partial<Omit<DbEntity, 'id' | 'createdAt'>>): Promise<void>
  delete(id: string): Promise<void>
  // ... entity-specific queries
}

Add to RepositoryProvider:

typescript

export type RepositoryProvider = {
  // ... existing
  entity: EntityRepository
}

Guidelines:

•Use ReadonlyArray<T> and Readonly<T> for arguments
•getById returns undefined (not throw) when not found
•create returns the created entity with generated ID
•update and delete return void
•Use Omit<> to exclude auto-generated fields (id, createdAt)

Step 2: Add Schema Type

Location: src/db/schema.ts

Define database type with Db prefix:

typescript

/**
 * Entity stored in database.
 * Uses null instead of undefined for explicit "no value" semantics.
 */
export type DbEntity = {
  id: string
  name: string
  value: string | null        // Use null, not undefined
  createdAt: number
  updatedAt: number | null    // null until first update
}

Key Conventions:

•Always use Db prefix for database types
•Use null for "no value" (not undefined)
•Store user input numbers as string (e.g., kg: string, reps: string)
•Use discriminated unions with kind property for variants
•Include type guards if needed: export function isDbEntity(x: unknown): x is DbEntity

Step 3: Update Database Class

Location: src/db/implementations/dexie/database.ts

Add table and indexes:

typescript

export class WorkoutTrackerDb extends Dexie {
  // ... existing tables
  entities!: Table<DbEntity, string>

  constructor() {
    super('WorkoutTracker')

    // Increment version number
    this.version(3).stores({
      // ... existing tables
      entities: 'id, name, createdAt',  // Index: primary + frequently queried fields
    })
  }
}

Indexing Guidelines:

•Always index primary key (automatic)
•Index fields used in where(), orderBy(), equals()
•Index foreign keys for joins
•Compound indexes for junction tables: '[field1+field2], field1, field2'

Step 4: Implement Repository

Location: src/db/implementations/dexie/[entity].ts

Create factory function returning repository implementation:

typescript

import type { EntityRepository } from '@/db/interfaces'
import type { DbEntity } from '@/db/schema'
import { createDatabaseError, tryCatch } from '@/lib/tryCatch'
import type { WorkoutTrackerDb } from './database'
import { generateId } from './database'

/**
 * Dexie implementation of EntityRepository.
 */
export function createDexieEntityRepository(db: WorkoutTrackerDb): EntityRepository {
  return {
    async getAll(): Promise<ReadonlyArray<DbEntity>> {
      const [error, entities] = await tryCatch(
        db.entities.orderBy('createdAt').reverse().toArray(),
      )
      if (error) {
        throw createDatabaseError('LOAD_FAILED', 'retrieve entities', error)
      }
      return entities
    },

    async getById(id: string): Promise<DbEntity | undefined> {
      const [error, entity] = await tryCatch(db.entities.get(id))
      if (error) {
        throw createDatabaseError('LOAD_FAILED', `retrieve entity with id ${id}`, error)
      }
      return entity
    },

    async create(
      entity: Omit<DbEntity, 'id' | 'createdAt'>,
    ): Promise<DbEntity> {
      const newEntity: DbEntity = {
        ...entity,
        id: generateId(),
        createdAt: Date.now(),
      }

      const [error] = await tryCatch(db.entities.add(newEntity))
      if (error) {
        throw createDatabaseError('SAVE_FAILED', 'create entity', error)
      }

      return newEntity
    },

    async update(
      id: string,
      updates: Partial<Omit<DbEntity, 'id' | 'createdAt'>>,
    ): Promise<void> {
      const [error, updatedCount] = await tryCatch(
        db.entities.update(id, {
          ...updates,
          updatedAt: Date.now(),  // Auto-inject timestamp
        }),
      )

      if (error) {
        throw createDatabaseError('SAVE_FAILED', `update entity with id ${id}`, error)
      }

      if (updatedCount === 0) {
        throw createDatabaseError('NOT_FOUND', `entity with id ${id} not found`)
      }
    },

    async delete(id: string): Promise<void> {
      const [error] = await tryCatch(db.entities.delete(id))
      if (error) {
        throw createDatabaseError('SAVE_FAILED', `delete entity with id ${id}`, error)
      }
      // Soft delete: no NOT_FOUND check
    },
  }
}

Key Patterns:

•Use tryCatch() wrapper for all operations (preferred pattern)
•Two-phase error checking: operation failure + not found
•Auto-inject timestamps: createdAt, updatedAt
•Use generateId() for new IDs
•Soft delete: no error if entity doesn't exist

Step 5: Register in Factory Provider

Location: src/db/implementations/dexie/index.ts

Import and add to provider:

typescript

import { createDexieEntityRepository } from './entity'

export function createDexieRepositoryProvider(): RepositoryProvider {
  return {
    activeWorkout: createDexieActiveWorkoutRepository(db),
    workouts: createDexieWorkoutsRepository(db),
    // ... existing repositories
    entity: createDexieEntityRepository(db),  // ADD THIS
  }
}

Step 6: Export Public Getter

Location: src/db/index.ts

Add getter function:

typescript

export function getEntityRepository(): EntityRepository {
  return getRepositoryProvider().entity
}

Usage in Features

typescript

import { getEntityRepository } from '@/db'
import type { DbEntity } from '@/db/schema'

export function useEntities() {
  const entities = ref<ReadonlyArray<DbEntity>>([])
  const entityRepo = getEntityRepository()

  async function loadEntities() {
    entities.value = await entityRepo.getAll()
  }

  async function createEntity(name: string, value: string | null) {
    const newEntity = await entityRepo.create({ name, value, updatedAt: null })
    entities.value = [...entities.value, newEntity]
  }

  async function updateEntity(id: string, updates: Partial<DbEntity>) {
    await entityRepo.update(id, updates)
    await loadEntities()
  }

  async function deleteEntity(id: string) {
    await entityRepo.delete(id)
    entities.value = entities.value.filter(e => e.id !== id)
  }

  onMounted(() => loadEntities())

  return {
    entities: readonly(entities),
    createEntity,
    updateEntity,
    deleteEntity,
  }
}

Key Principles

1. Error Handling

Preferred: tryCatch wrapper

typescript

const [error, result] = await tryCatch(operation)
if (error) {
  throw createDatabaseError('ERROR_CODE', 'description', error)
}

Error codes:

•SAVE_FAILED - Create, update, delete operations
•LOAD_FAILED - Read operations
•NOT_FOUND - Entity doesn't exist

2. Timestamp Management

Auto-inject timestamps in repository methods:

•createdAt: Date.now() in create()
•updatedAt: Date.now() in update()
•lastUsedAt: Date.now() when accessing entity

3. ID Generation

Always use generateId() from database.ts:

typescript

import { generateId } from './database'

const newEntity = {
  ...entity,
  id: generateId(),  // crypto.randomUUID()
}

4. Soft Delete

Delete operations don't throw if entity doesn't exist:

typescript

async delete(id: string): Promise<void> {
  await tryCatch(db.entities.delete(id))
  // No NOT_FOUND check - silent success
}

5. Type Safety

•Use Readonly<T> and ReadonlyArray<T> for function parameters
•Use Omit<> to exclude auto-generated fields
•Use discriminated unions with exhaustive checking
•Define type guards for runtime type checking

File Reference

Critical files when creating repository:

•src/db/interfaces.ts - Interface definition + RepositoryProvider
•src/db/schema.ts - Db-prefixed type definitions
•src/db/implementations/dexie/database.ts - Table + indexes
•src/db/implementations/dexie/[entity].ts - Implementation
•src/db/implementations/dexie/index.ts - Factory registration
•src/db/index.ts - Public getter export

Utility imports:

•@/lib/tryCatch - Error handling utilities
•@/db/implementations/dexie/database - generateId()

Detailed References

For complete examples and advanced patterns, see:

•
references/examples.md - Complete end-to-end examples:
- •Example 1: Simple CRUD repository (Notes)
- •Example 2: Complex transformations (Tags with many-to-many)
- •Example 3: Extending Settings with function overloads
•
references/patterns.md - Detailed pattern catalog:
- •Error handling patterns (direct throw vs tryCatch)
- •CRUD patterns (getAll, create, update, delete, timestamps)
- •Type transformation patterns (helper utilities, deep cloning)
- •Advanced patterns (function overloads, singleton, transactions, bulk ops)
- •Schema design patterns (discriminated unions, indexing, embedded vs referenced)

Common Scenarios

Scenario 1: Simple CRUD Repository

Need basic storage for an entity? See examples.md → Example 1 (Notes).

Quick checklist:

•Define interface with getAll/getById/create/update/delete
•Add DbEntity type with Db prefix
•Add table with indexes
•Implement using tryCatch pattern
•Register and export

Scenario 2: Complex Relationships

Need many-to-many relationships or complex queries? See examples.md → Example 2 (Tags).

Pattern: Junction table + transaction handling + usage tracking.

Scenario 3: Extending Settings

Adding new setting? See examples.md → Example 3.

Pattern: Add discriminated union member + function overload + default value.

Scenario 4: Conversions Between Types

Need to convert between templates and workouts? See patterns.md → Type Transformation Patterns.

Pattern: Helper utilities with exhaustive switch statements.

Scenario 5: Bulk Operations

Import/export or batch delete? See patterns.md → Advanced Patterns → Bulk Operations.

Pattern: Transactions + Promise.all() + bulkAdd().

Testing Support

Mock repositories for unit tests:

typescript

import { createMockRepositories } from '@/__tests__/helpers/mockRepositories'

const mockRepos = createMockRepositories()
mockRepos.entity.getAll.mockResolvedValue([...])

Integration tests with fake-indexeddb are automatically set up via test helpers.

Migration Strategy

When updating schema version:

•Increment version number in database.ts
•Add new table/indexes in .stores({})
•Dexie handles migrations automatically
•For data migrations, use .upgrade() callback

typescript

this.version(3)
  .stores({
    entities: 'id, name, createdAt',
  })
  .upgrade(tx => {
    // Optional data migration logic
    return tx.table('entities').toCollection().modify(entity => {
      entity.newField = 'default'
    })
  })

Project-Specific Repositories

`Db*` Types vs Domain Types

Aspect	Database (`Db*`)	Domain
File	`src/db/schema.ts`	`src/types/`
Prefix	`DbWorkout`, `DbSet`	`Workout`, `Set`
No value	`null`	`undefined`
Optimized for	Storage	App logic

Available Repositories

SettingsRepository - Key-value store with defaults:

const repo = getSettingsRepository()
await repo.get('theme')           // 'light' | 'dark' | 'system'
await repo.get('defaultRestTimer') // number
await repo.set({ key: 'theme', value: 'dark' })
await repo.getAll()               // All settings merged with defaults
await repo.reset('theme')

CustomExercisesRepository - Exercise CRUD:

const repo = getCustomExercisesRepository()
await repo.getAll()
await repo.getById(id)
await repo.add({ id: generateId(), name: 'Squat', ... })
await repo.update(id, { name: 'Back Squat' })
await repo.delete(id)

WorkoutsRepository - Completed workouts:

const repo = getWorkoutsRepository()
await repo.getAll()
await repo.getById(id)
await repo.create(convertWorkoutToDb(workout))
await repo.delete(id)

ActiveWorkoutRepository - Singleton active workout:

const repo = getActiveWorkoutRepository()
await repo.load()
await repo.save(dbActiveWorkout)
await repo.delete()
await repo.exists()

BenchmarksRepository - Benchmark workouts:

const repo = getBenchmarksRepository()
await repo.getAll()
await repo.getById(id)
await repo.create({ id: generateId(), name: 'Fran', ... })
await repo.update(id, { name: 'Fran (Scaled)' })
await repo.delete(id)

TemplatesRepository - Workout templates:

const repo = getTemplatesRepository()
await repo.getAll()
await repo.getById(id)
await repo.create(template)
await repo.update(id, changes)
await repo.delete(id)

Using Converters

Always convert when crossing domain/database boundary:

import { convertWorkoutToDb, convertDbToWorkout } from '@/db/converters'

// Domain → Database
const dbWorkout = convertWorkoutToDb(workout)
await getWorkoutsRepository().create(dbWorkout)

// Database → Domain
const dbWorkout = await getWorkoutsRepository().getById(id)
const workout = convertDbToWorkout(dbWorkout)

Partial Updates with buildPartialUpdate

Dexie's update() overwrites all keys in the object. Use buildPartialUpdate to only modify provided fields:

import { buildPartialUpdate } from '@/db/partialUpdate'

const NULLABLE_FIELDS = ['equipment', 'muscle', 'image']

// Only includes keys present in updates
// Converts undefined → null for nullable fields
const dbUpdates = buildPartialUpdate(updates, NULLABLE_FIELDS)
await repo.update(id, dbUpdates)

Why: Without filtering, { name: 'Squat', equipment: undefined } would set equipment to null even if you only meant to update the name.

Project-Specific Gotchas

1. Use `null` in Database, `undefined` in Domain

IndexedDB doesn't support undefined:

// Database types
type DbExercise = {
  equipment: Equipment | null  // Use null
}

// Domain types
type Exercise = {
  equipment?: Equipment  // Use undefined
}

2. Always Reset Database in Tests

import { resetDatabase } from '@/__tests__/setup'

beforeEach(async () => {
  await resetDatabase()
})

3. Convert Types at Boundaries

// BAD - Type mismatch
await getWorkoutsRepository().create(workout)

// GOOD - Convert first
const dbWorkout = convertWorkoutToDb(workout)
await getWorkoutsRepository().create(dbWorkout)

Repository Pattern

Overview

Architecture Overview

Core Workflow

Step 1: Define Interface

Step 2: Add Schema Type

Step 3: Update Database Class

Step 4: Implement Repository

Step 5: Register in Factory Provider

Step 6: Export Public Getter

Usage in Features

Key Principles

1. Error Handling

2. Timestamp Management

3. ID Generation

4. Soft Delete

5. Type Safety

File Reference

Detailed References

Common Scenarios

Scenario 1: Simple CRUD Repository

Scenario 2: Complex Relationships

Scenario 3: Extending Settings

Scenario 4: Conversions Between Types

Scenario 5: Bulk Operations

Testing Support

Migration Strategy

Project-Specific Repositories

Db* Types vs Domain Types

Available Repositories

Using Converters

Partial Updates with buildPartialUpdate

Project-Specific Gotchas

1. Use null in Database, undefined in Domain

2. Always Reset Database in Tests

3. Convert Types at Boundaries

`Db*` Types vs Domain Types

1. Use `null` in Database, `undefined` in Domain