- Add detailed documentation for `cleanup()` and `cleanupFailedStart()` methods
- Clarify the purpose, behavior, and differences between the methods
- Document parameters, return values, and error handling
- Add @remarks sections for important implementation details
- Remove waitForReady option from RhizomeNode.start() and always wait for readiness
- Update test files to use simplified start() method
- Clean up debug logging in lossless view
- Rename docker orchestrator test file to remove version suffix
- Removed unused dockerOptions from DockerOrchestrator and all manager classes
- Updated container-manager to use NanoCpus for CPU limits instead of CpuQuota/CpuPeriod
- Fixed resource limit enforcement in container configuration
- Improved test reliability and error handling
- Cleaned up test files and removed unused code
This commit includes a major refactoring of the Docker orchestrator implementation
along with improvements to the testing infrastructure:
- Refactored Docker orchestrator to handle dynamic port assignment
- Added comprehensive test utilities in docker-test-utils.ts
- Improved error handling and resource cleanup in test environments
- Enhanced NodeStatus interface with containerId and networkId
- Added support for different storage types in NodeConfig
- Fixed request port handling in TestOrchestrator
- Added proper cleanup method to NodeOrchestrator interface
The changes ensure more reliable container management and better test isolation
while maintaining backward compatibility with existing implementations.
BREAKING CHANGE: The NodeOrchestrator interface now requires a cleanup() method.
- Add DockerOrchestrator implementation with container management
- Refactor InMemoryOrchestrator to use RhizomeNode directly
- Fix orchestrated test files for single and two-node scenarios
- Update package.json with dockerode dependency
- Clean up old test files and imports
- Refactored isDeltaNegated to properly handle complex negation chains
- Added recursive negation detection to account for negations of negations
- Replaced console.log with debug statements in tests
- Improved test coverage and documentation
- Fixed TypeScript type issues
- Update ResolverPlugin interface to allow resolve() to return undefined
- Modify CustomResolver to skip properties with undefined values
- Add test case for resolvers that return undefined
- Update existing tests to handle cases where no properties are resolved
This change makes the behavior more explicit when resolvers don't return a value,
which is particularly important for numeric aggregations where 0 might be a valid result.
- Moved the resolve method implementation from individual resolvers to the base Lossy class
- Updated initializer methods to accept a LosslessViewOne parameter
- Removed redundant resolve methods from LastWriteWins, TimestampResolver, CustomResolver, and AggregationResolver
- Ensured consistent behavior across all resolver implementations
- All tests passing with the refactored code
- Moved CommonSchemas from src/schema/schema.ts to src/test-utils/schemas.ts
- Updated all test files to import CommonSchemas from the new location
- Fixed the Document schema to match test expectations by making the author field required
- Added additional fields to the Document schema to match the original implementation
- Ensured all tests pass with the new implementation
Addresses PR feedback that CommonSchemas is only used in tests and should be moved to test files.
Addresses PR feedback about the outdated comment in concurrent-writes.ts. The comment now accurately reflects that the resolution uses the LastWriteWins resolver's tie-breaking algorithm rather than delta processing order.
