🔙 Legacy Module Integration
This document outlines how existing legacy code is integrated into the new modular architecture of the application and the strategy for its migration. The core principle is to isolate legacy code and provide a controlled way for newer modules to interact with legacy functionality without becoming directly dependent on it.
note
This document should be read in conjunction with Module Structure and Module Organization to get a complete understanding of the modular architecture.
Overview
The Thunderbird for Android project is transitioning from a monolithic architecture to a modular one. During this
transition, we need to maintain compatibility with existing legacy code while gradually migrating to the new
architecture. The legacy:*, mail:*, and backend:* modules contain functionality that is still essential for the
project but does not yet adhere to the new modular architecture. These modules are integrated into the new architecture
through the :app-common module, which acts as a bridge or adapter to provide access to legacy functionality without
directly depending on it.
The key components in this integration strategy are:
- Legacy Modules:
legacy:*,mail:*, andbackend:*modules containing existing functionality - Interfaces: Well-defined interfaces in
feature:*:apiandcore:*modules - App Common Bridge: The
:app-commonmodule that implements these interfaces and delegates to legacy code - Dependency Injection: Configuration that provides the appropriate implementations to modules
Integration Approach “The App Common Bridge”
Newer application modules (such as features or core components) depend on well-defined Interfaces
(e.g., those found in feature:*:api modules). Typically, a feature will provide its own modern Implementation
(e.g., :feature:mail:impl) for its API.
However, to manage dependencies on code still within legacy:*, mail:*, and backend:* modules and prevent it
from spreading, we use app-common as bridge or adapter to provide an alternative implementation for these. In
this role, app-common is responsible for:
- Implementing interfaces:
app-commonprovides concrete implementations for interfaces that newer modules define. - Delegating to legacy code: Internally, these
app-commonimplementations will delegate calls, adapt data, and manage interactions with the underlyinglegacy:*,mail:*, andbackend:*modules. - Containing glue code: All logic required to connect the modern interfaces with the legacy systems is encapsulated within
app-common.
This approach ensures that:
- Newer modules are decoupled from legacy implementations: They only interact with the defined interfaces, regardless of whether the implementation is the modern feature
implor theapp-commonbridge. - Legacy code is isolated.
- A clear path for refactoring is maintained: Initially, the application might be configured to use the
app-commonbridge. As new, native implementations in feature modules (e.g.,:feature:mail:impl) mature, the dependency injection can be switched to use them, often without changes to the modules consuming the interface.
Bridge Pattern Flow
The typical flow is:
- Interfaces: Interfaces are defined, usually within the
apimodule of a feature (e.g.,:feature:mail:api) or a core module. These interfaces represent the contract for a piece of functionality. - New Module Dependency: Newer modules (e.g.,
:feature:somefeature:implor other parts of:app-common) depend on these defined interfaces, to avoid dependency on concrete legacy classes. - Implementation: The
:app-commonmodule provides concrete implementations for these interfaces. - Delegation to Legacy: Internally, these implementations within
:app-commondelegate the actual work to the code residing in the legacy modules (e.g.,legacy:*,mail:*,backend:*). - Dependency Injection: The application’s dependency injection framework is configured to provide instances of these
:app-commonbridge implementations when a newer module requests an implementation of the interface.
This pattern ensures that newer modules remain decoupled from the specifics of legacy code.
The following diagram illustrates this pattern, showing how both a feature’s own implementation and app-common can relate to the interfaces, with app-common specifically bridging to legacy systems:
graph TB
subgraph FEATURE[Feature Modules]
direction TB
INTERFACES["`**Interfaces**<br> (e.g., :feature:mail:api)`"]
IMPLEMENTATIONS["`**Implementations**<br> (e.g., :feature:mail:impl)`"]
OTHER_MODULES["`**Other Modules**<br>(depend on Interfaces)`"]
end
subgraph COMMON[App Common Module]
direction TB
COMMON_APP["`**:app-common**<br>Integration Code`"]
end
subgraph LEGACY[Legacy Modules]
direction TB
LEGACY_K9["`**:legacy**`"]
LEGACY_MAIL["`**:mail**`"]
LEGACY_BACKEND["`**:backend**`"]
end
OTHER_MODULES --> |uses| INTERFACES
IMPLEMENTATIONS --> |depends on| INTERFACES
COMMON_APP --> |implements| INTERFACES
COMMON_APP --> |delegates to / wraps| LEGACY_K9
COMMON_APP --> |delegates to / wraps| LEGACY_MAIL
COMMON_APP --> |delegates to / wraps| LEGACY_BACKEND
classDef common fill:#e6e6e6,stroke:#000000,color:#000000
classDef common_module fill:#999999,stroke:#000000,color:#000000
classDef feature fill:#d9ffd9,stroke:#000000,color:#000000
classDef feature_module fill:#33cc33,stroke:#000000,color:#000000
classDef legacy fill:#ffe6e6,stroke:#000000,color:#000000
classDef legacy_module fill:#ff9999,stroke:#000000,color:#000000
linkStyle default stroke:#999,stroke-width:2px
class COMMON common
class COMMON_APP common_module
class FEATURE feature
class INTERFACES,IMPLEMENTATIONS,OTHER_MODULES feature_module
class LEGACY legacy
class LEGACY_MAIL,LEGACY_BACKEND,LEGACY_K9 legacy_module
Implementation Techniques
Several techniques are used to implement the bridge pattern effectively:
-
Wrapper Classes: Creating immutable data classes that wrap legacy data structures, implementing interfaces from the new architecture. These wrappers should not contain conversion methods but should delegate this responsibility to specific mapper classes.
-
Adapter Implementations: Classes in
:app-commonthat implement interfaces from the new architecture but delegate to legacy code internally. -
Data Conversion: Dedicated mapper classes that handle mapping between legacy and new data structures, ensuring clean separation of concerns.
Example: Account Profile Bridge
A concrete example of this pattern is the account profile bridge, which demonstrates a complete implementation of the bridge pattern across multiple layers:
- Modern Interfaces:
AccountProfileRepositoryinfeature:account:apidefines the high-level contract for account profile managementAccountProfileLocalDataSourceinfeature:account:coredefines the data access contract
- Modern Data Structure:
AccountProfileinfeature:account:apiis a clean, immutable data class that represents account profile information in the new architecture. - Repository Implementation:
DefaultAccountProfileRepositoryinfeature:account:coreimplements theAccountProfileRepositoryinterface and depends onAccountProfileLocalDataSource. - Bridge Implementation:
DefaultAccountProfileLocalDataSourceinapp-commonimplements theAccountProfileLocalDataSourceinterface and serves as the bridge to legacy code. - Legacy Access: The bridge uses
DefaultLegacyAccountWrapperManagerto access legacy account data:LegacyAccountWrapperManagerincore:android:accountdefines the contract for legacy account accessLegacyAccountWrapperincore:android:accountis an immutable wrapper around the legacyLegacyAccountclass
- Data Conversion: The bridge uses a dedicated mapper class to convert between modern
AccountProfileobjects and legacy account data. - Dependency Injection: The
appCommonAccountModuleinapp-commonregistersDefaultAccountProfileLocalDataSourceas implementations of the respective interface.
This multi-layered approach allows newer modules to interact with legacy account functionality through clean, modern interfaces without directly depending on legacy code. It also demonstrates how bridges can be composed, with higher-level bridges (AccountProfile) building on lower-level bridges (LegacyAccountWrapper).
Testing Considerations
Testing bridge implementations requires special attention to ensure both the bridge itself and its integration with legacy code work correctly:
- Unit Testing Bridge Classes:
- Test the bridge implementation in isolation by faking/stubbing the legacy dependencies
- Verify that the bridge correctly translates between the new interfaces and legacy code
- Focus on testing the conversion logic and error handling
- Integration Testing:
- Test the bridge with actual legacy code to ensure proper integration
- Verify that the bridge correctly handles all edge cases from legacy code
- Test Doubles:
- Create fake implementations of bridge classes for testing other components
- Example:
FakeLegacyAccountWrapperManagercan be used to test components that depend onLegacyAccountWrapperManager
- Migration Testing:
- When migrating from a legacy bridge to a new implementation, test both implementations with the same test suite
- Ensure behavior consistency during the transition
Testing Examples
Below are examples of tests for legacy module integration, demonstrating different testing approaches and best practices.
Example 1: Unit Testing a Bridge Implementation
This example shows how to test a bridge implementation (DefaultAccountProfileLocalDataSource) in isolation by using a fake implementation of the legacy dependency (FakeLegacyAccountWrapperManager):
class DefaultAccountProfileLocalDataSourceTest {
@Test
fun `getById should return account profile`() = runTest {
// arrange
val accountId = AccountIdFactory.create()
val legacyAccount = createLegacyAccount(accountId)
val accountProfile = createAccountProfile(accountId)
val testSubject = createTestSubject(legacyAccount)
// act & assert
testSubject.getById(accountId).test {
assertThat(awaitItem()).isEqualTo(accountProfile)
}
}
@Test
fun `getById should return null when account is not found`() = runTest {
// arrange
val accountId = AccountIdFactory.create()
val testSubject = createTestSubject(null)
// act & assert
testSubject.getById(accountId).test {
assertThat(awaitItem()).isEqualTo(null)
}
}
@Test
fun `update should save account profile`() = runTest {
// arrange
val accountId = AccountIdFactory.create()
val legacyAccount = createLegacyAccount(accountId)
val accountProfile = createAccountProfile(accountId)
val updatedName = "updatedName"
val updatedAccountProfile = accountProfile.copy(name = updatedName)
val testSubject = createTestSubject(legacyAccount)
// act & assert
testSubject.getById(accountId).test {
assertThat(awaitItem()).isEqualTo(accountProfile)
testSubject.update(updatedAccountProfile)
assertThat(awaitItem()).isEqualTo(updatedAccountProfile)
}
}
private fun createTestSubject(
legacyAccount: LegacyAccountWrapper?,
): DefaultAccountProfileLocalDataSource {
return DefaultAccountProfileLocalDataSource(
accountManager = FakeLegacyAccountWrapperManager(
initialAccounts = if (legacyAccount != null) {
listOf(legacyAccount)
} else {
emptyList()
},
),
dataMapper = DefaultAccountProfileDataMapper(
avatarMapper = DefaultAccountAvatarDataMapper(),
),
)
}
}
Key points:
- The test creates a controlled test environment using a fake implementation of the legacy dependency
- It tests both success cases and error handling (account not found)
- It verifies that the bridge correctly translates between legacy data structures and domain models
- The test is structured with clear arrange, act, and assert sections
Example 2: Creating Test Doubles for Legacy Dependencies
This example shows how to create a fake implementation of a legacy dependency (FakeLegacyAccountWrapperManager) for testing:
internal class FakeLegacyAccountWrapperManager(
initialAccounts: List<LegacyAccountWrapper> = emptyList(),
) : LegacyAccountWrapperManager {
private val accountsState = MutableStateFlow(
initialAccounts,
)
private val accounts: StateFlow<List<LegacyAccountWrapper>> = accountsState
override fun getAll(): Flow<List<LegacyAccountWrapper>> = accounts
override fun getById(id: AccountId): Flow<LegacyAccountWrapper?> = accounts
.map { list ->
list.find { it.id == id }
}
override suspend fun update(account: LegacyAccountWrapper) {
accountsState.update { currentList ->
currentList.toMutableList().apply {
removeIf { it.uuid == account.uuid }
add(account)
}
}
}
}
Key points:
- The fake implementation implements the same interface as the real implementation
- It provides a simple in-memory implementation for testing
- It uses Kotlin Flows to simulate the reactive behavior of the real implementation
- It allows for easy setup of test data through the constructor parameter
Example 3: Testing Data Conversion Logic
This example shows how to test data conversion logic in bridge implementations:
class DefaultAccountProfileDataMapperTest {
@Test
fun `toDomain should convert ProfileDto to AccountProfile`() {
// Arrange
val dto = createProfileDto()
val expected = createAccountProfile()
val testSubject = DefaultAccountProfileDataMapper(
avatarMapper = FakeAccountAvatarDataMapper(
dto = dto.avatar,
domain = expected.avatar,
),
)
// Act
val result = testSubject.toDomain(dto)
// Assert
assertThat(result.id).isEqualTo(expected.id)
assertThat(result.name).isEqualTo(expected.name)
assertThat(result.color).isEqualTo(expected.color)
assertThat(result.avatar).isEqualTo(expected.avatar)
}
@Test
fun `toDto should convert AccountProfile to ProfileDto`() {
// Arrange
val domain = createAccountProfile()
val expected = createProfileDto()
val testSubject = DefaultAccountProfileDataMapper(
avatarMapper = FakeAccountAvatarDataMapper(
dto = expected.avatar,
domain = domain.avatar,
),
)
// Act
val result = testSubject.toDto(domain)
// Assert
assertThat(result.id).isEqualTo(expected.id)
assertThat(result.name).isEqualTo(expected.name)
assertThat(result.color).isEqualTo(expected.color)
assertThat(result.avatar).isEqualTo(expected.avatar)
}
}
Key points:
- The test verifies that the mapper correctly converts between legacy data structures (DTOs) and domain models
- It tests both directions of the conversion (toDomain and toDto)
- It uses a fake implementation of a dependency (FakeAccountAvatarDataMapper) to isolate the test
- It verifies that all properties are correctly mapped
Best Practices for Testing Legacy Module Integration
- Isolate the Bridge: Test the bridge implementation in isolation by using fake or mock implementations of legacy dependencies.
- Test Both Directions: For data conversion, test both directions (legacy to domain and domain to legacy).
- Cover Edge Cases: Test edge cases such as null values, empty collections, and error conditions.
- Use Clear Test Structure: Structure tests with clear arrange, act, and assert sections.
- Create Reusable Test Fixtures: Create helper methods for creating test data to make tests more readable and maintainable.
- Test Reactive Behavior: For reactive code (using Flows, LiveData, etc.), use appropriate testing utilities (e.g., Turbine for Flow testing).
- Verify Integration: In addition to unit tests, create integration tests that verify the bridge works correctly with actual legacy code.
- Test Migration Path: When migrating from a legacy bridge to a new implementation, test both implementations with the same test suite to ensure behavior consistency.
Migration Strategy
The long-term strategy involves gradually migrating functionality out of the legacy modules:
- Identify Functionality: Pinpoint specific functionalities within legacy modules that need to be modernized.
- Define Interfaces: Ensure clear interfaces are defined (typically in feature
apimodules) for this functionality. - Entity Modeling: Create proper domain entity models that represent the business objects as immutable data classes.
- Implement in New Modules: Re-implement the functionality within new, dedicated feature
implmodules or core modules. - Update Bridge (Optional): If
:app-commonwas bridging to this specific legacy code, its bridge implementation can be updated or removed. - Switch DI Configuration: Update the dependency injection to provide the new modern implementation instead of the legacy bridge.
- Retire Legacy Code: Once no longer referenced, the corresponding legacy code can be safely removed.
Migration Example
Using the account profile example, the migration process would look like:
- Identify: Account profile functionality in legacy modules needs modernization.
- Define Interfaces:
AccountProfileRepositoryinterface is defined infeature:account:apiAccountProfileLocalDataSourceinterface is defined infeature:account:core
- Entity Modeling: Create
AccountProfileas an immutable data class infeature:account:api. - Implement: Create a new implementation of
AccountProfileLocalDataSourcein a modern module, e.g.,feature:account:impl. - Update Bridge: Update or remove
DefaultAccountProfileLocalDataSourceinapp-common. - Switch DI: Update
appCommonAccountModuleto provide the new implementation instead ofDefaultAccountProfileLocalDataSource. - Retire: Once all references to legacy account code are removed, the legacy code and lower-level bridges (
LegacyAccountWrapperManager,DefaultLegacyAccountWrapperManager) can be safely deleted.
This approach ensures a smooth transition with minimal disruption to the application’s functionality.
Dependency Direction
A strict dependency rule is enforced: New modules (features, core) must not directly depend on legacy modules.
The dependency flow is always from newer modules to interfaces, with :app-common providing the implementation.
If :app-common bridges to legacy code, that is an internal detail of :app-common.
The legacy module integration diagram below explains how legacy code is integrated into the new modular architecture:
graph TB
subgraph APP[App Modules]
direction TB
APP_TB["`**:app-thunderbird**<br>Thunderbird for Android`"]
APP_K9["`**:app-k9mail**<br>K-9 Mail`"]
end
subgraph COMMON[App Common Module]
direction TB
COMMON_APP["`**:app-common**<br>Integration Code`"]
end
subgraph FEATURE[Feature Modules]
direction TB
FEATURE1[Feature 1]
FEATURE2[Feature 2]
FEATURE3[Feature from Legacy]
end
subgraph CORE[Core Modules]
direction TB
CORE1[Core 1]
CORE2[Core 2]
CORE3[Core from Legacy]
end
subgraph LIBRARY[Library Modules]
direction TB
LIB1[Library 1]
LIB2[Library 2]
end
subgraph LEGACY[Legacy Modules]
direction TB
LEGACY_CODE[Legacy Code]
end
APP_K9 --> |depends on| COMMON_APP
APP_TB --> |depends on| COMMON_APP
COMMON_APP --> |integrates| FEATURE1
COMMON_APP --> |integrates| FEATURE2
COMMON_APP --> |integrates| FEATURE3
FEATURE1 --> |uses| CORE1
FEATURE1 --> |uses| LIB2
FEATURE2 --> |uses| CORE2
FEATURE2 --> |uses| CORE3
COMMON_APP --> |integrates| LEGACY_CODE
LEGACY_CODE -.-> |migrate to| FEATURE3
LEGACY_CODE -.-> |migrate to| CORE3
classDef app fill:#d9e9ff,stroke:#000000,color:#000000
classDef app_module fill:#4d94ff,stroke:#000000,color:#000000
classDef common fill:#e6e6e6,stroke:#000000,color:#000000
classDef common_module fill:#999999,stroke:#000000,color:#000000
classDef feature fill:#d9ffd9,stroke:#000000,color:#000000
classDef feature_module fill:#33cc33,stroke:#000000,color:#000000
classDef core fill:#e6cce6,stroke:#000000,color:#000000
classDef core_module fill:#cc99cc,stroke:#000000,color:#000000
classDef library fill:#fff0d0,stroke:#000000,color:#000000
classDef library_module fill:#ffaa33,stroke:#000000,color:#000000
classDef legacy fill:#ffe6e6,stroke:#000000,color:#000000
classDef legacy_module fill:#ff9999,stroke:#000000,color:#000000
linkStyle default stroke:#999,stroke-width:2px
class APP app
class APP_K9,APP_TB app_module
class COMMON common
class COMMON_APP common_module
class FEATURE feature
class FEATURE1,FEATURE2,FEATURE3 feature_module
class CORE core
class CORE1,CORE2,CORE3 core_module
class LIBRARY library
class LIB1,LIB2 library_module
class LEGACY legacy
class LEGACY_CODE legacy_module