This commit significantly increases the test coverage across the application and fixes several underlying bugs that were discovered while writing the new tests.
The key changes include:
- **New Tests:** Added extensive integration and unit tests for GraphQL resolvers, application services, and data repositories, substantially increasing the test coverage for packages like `graphql`, `user`, `translation`, and `analytics`.
- **Authorization Bug Fixes:**
- Fixed a critical bug where a user creating a `Work` was not correctly associated as its author, causing subsequent permission failures.
- Corrected the authorization logic in `authz.Service` to properly check for entity ownership by non-admin users.
- **Test Refactoring:**
- Refactored numerous test suites to use `testify/mock` instead of manual mocks, improving test clarity and maintainability.
- Isolated integration tests by creating a fresh admin user and token for each test run, eliminating test pollution.
- Centralized domain errors into `internal/domain/errors.go` and updated repositories to use them, making error handling more consistent.
- **Code Quality Improvements:**
- Replaced manual mock implementations with `testify/mock` for better consistency.
- Cleaned up redundant and outdated test files.
These changes stabilize the test suite, improve the overall quality of the codebase, and move the project closer to the goal of 80% test coverage.
This commit introduces a significant refactoring to improve the application's quality, test coverage, and production readiness, focusing on core localization and business logic features.
Key changes include:
- Consolidated the `CreateTranslation` and `UpdateTranslation` commands into a single, more robust `CreateOrUpdateTranslation` command. This uses a database-level `Upsert` for atomicity.
- Centralized authorization for translatable entities into a new `CanEditEntity` check within the application service layer.
- Fixed a critical bug in the `MergeWork` command that caused a UNIQUE constraint violation when merging works with conflicting translations. The logic now intelligently handles language conflicts.
- Implemented decrementing for "like" counts in the analytics service when a like is deleted, ensuring accurate statistics.
- Stabilized the test suite by switching to a file-based database for integration tests, fixing test data isolation issues, and adding a unique index to the `Translation` model to enforce data integrity.
- Refactored manual mocks to use the `testify/mock` library for better consistency and maintainability.
This commit addresses all the high-priority tasks outlined in the TASKS.md file, significantly improving the application's observability, completing key features, and refactoring critical parts of the codebase.
### Observability
- **Centralized Logging:** Implemented a new structured, context-aware logging system using `zerolog`. A new logging middleware injects request-specific information (request ID, user ID, trace ID) into the logger, and all application logging has been refactored to use this new system.
- **Prometheus Metrics:** Added Prometheus metrics for database query performance by creating a GORM plugin that automatically records query latency and totals.
- **OpenTelemetry Tracing:** Fully instrumented all application services in `internal/app` and data repositories in `internal/data/sql` with OpenTelemetry tracing, providing deep visibility into application performance.
### Features
- **Analytics:** Implemented like, comment, and bookmark counting. The respective command handlers now call the analytics service to increment counters when these actions are performed.
- **Enrichment Tool:** Built a new, extensible `enrich` command-line tool to fetch data from external sources. The initial implementation enriches author data using the Open Library API.
### Refactoring & Fixes
- **Decoupled Testing:** Refactored the testing utilities in `internal/testutil` to be database-agnostic, promoting the use of mock-based unit tests and improving test speed and reliability.
- **Build Fixes:** Resolved numerous build errors, including a critical import cycle between the logging, observability, and authentication packages.
- **Search Service:** Fixed the search service integration by implementing the `GetWorkContent` method in the localization service, allowing the search indexer to correctly fetch and index work content.
The integration tests for admin-only mutations were failing due to an authorization issue. The root cause was that the JWT token used in the tests did not reflect the user's admin role, which was being set directly in the database.
This commit fixes the issue by:
1. Updating the `CreateAuthenticatedUser` test helper to generate a new JWT token after a user's role is changed. This ensures the token contains the correct, up-to-date role.
2. Removing all uses of `auth.ContextWithAdminUser` from the integration tests, making the JWT token the single source of truth for authorization.
This change also removes unused imports and variables that were causing build failures after the refactoring. All integration tests now pass.
This commit addresses a broken build state caused by a mid-stream architectural refactoring. The changes align the existing code with the new Domain-Driven Design (DDD-lite) structure outlined in `refactor.md`.
Key changes include:
- Defined missing domain interfaces for `Auth`, `Localization`, and `Search`.
- Refactored application services to use a `Commands` and `Queries` pattern.
- Updated GraphQL resolvers to call application services instead of accessing repositories directly.
- Fixed dependency injection in `cmd/api/main.go` by removing the non-existent `ApplicationBuilder` and manually instantiating services.
- Corrected numerous test files (`integration`, `unit`, and `repository` tests) to reflect the new architecture, including fixing mock objects and test suite setups.
- Added missing database migrations for test schemas to resolve "no such table" errors.
This effort successfully gets the application to a compilable state and passes a significant portion of the test suite, laying the groundwork for further development and fixing the remaining test failures.
This change introduces a service layer to encapsulate the business logic
for each domain aggregate. This will make the code more modular,
testable, and easier to maintain.
The following services have been created:
- author
- bookmark
- category
- collection
- comment
- like
- tag
- translation
- user
The main Application struct has been updated to use these new services.
The integration test suite has also been updated to use the new
Application struct and services.
This is a work in progress. The next step is to fix the compilation
errors and then refactor the resolvers to use the new services.
This commit introduces a new application layer to the codebase, which decouples the GraphQL resolvers from the data layer. The resolvers now call application services, which in turn call the repositories. This change improves the separation of concerns and makes the code more testable and maintainable.
Additionally, this commit introduces dataloaders to solve the N+1 problem in the GraphQL resolvers. The dataloaders are used to batch and cache database queries, which significantly improves the performance of the API.
The following changes were made:
- Created application services for most of the domains.
- Refactored the GraphQL resolvers to use the new application services.
- Implemented dataloaders for the `Author` aggregate.
- Updated the `app.Application` struct to hold the application services instead of the repositories.
- Fixed a large number of compilation errors in the test files that arose from these changes.
There are still some compilation errors in the `internal/adapters/graphql/integration_test.go` file. These errors are due to the test files still trying to access the repositories directly from the `app.Application` struct. The remaining work is to update these tests to use the new application services.
