Use getTime().utc in both message constructors, document the timezone change, and cover concurrent construction and structured formatting. Restore ordinary logging calls in the parallel writer fixture.
Validation: 32 unit tests and 29 regression tests pass. Timestamp sanitizer checks pass under UTC and non-UTC timezones. The full suite retains the known allocator reports in factory-state and allocator.
Document and test explicit refresh boundaries for isEnabled guards, and
only read the shared root threshold when no namespace threshold applies.
Include the portable Nimble configuration loader.
Build replacement appenders and thresholds before installing them and
advancing the local version. Release partial replacements and old caches
outside the configuration lock. In order to make the `ensureFreshness`
version check as cheap as posible and only incur overhead when we
actually need to update the local cache, isolate the refresh logic in a
separate scope to prevent initialization and cleanup of temporary
variables when the version has not changed
Add regression coverage for repeated clone failures, cache preservation,
cleanup, and retry without another configuration change. Update
documentation to reflect the remaining freshness issue.
Serialize writer submission and shutdown, enqueue explicit stop messages after accepted records, and finish the final file batch before joining. Keep drained process-wide channel storage until exit to avoid the Nim 2.2.10 Channel.close mutex teardown defect.
Cover empty records, multiple destinations, joined producers, and concurrent shutdown. Include writer lifecycle sanitizer tests in nimble test and update usage documentation.
Validation: all 32 unit tests, eight writer exit scenarios, and writer lifecycle ThreadSanitizer tests pass. The full suite retains the known configuration refresh, timestamp, and default allocator failures.
Add SharedLogState for synchronized value data and factories for
independently owned appender state and captured formatter closures.
Reject unsafe configuration state, update the test collector to value
snapshots, and document the breaking APIs in version 3.0.0.
Expand ORC ownership and sanitizer coverage and reject stale test
executables.
Primarily intended for use in libraries or other contexts where you want
to fall back to std/logging if the application is not using or hasn't
configured namespaced_logging.
By default the StdLoggingAppender only logs when no namespaced_logging
appenders are configured, but it can also be configured to always
forward log messages, even when namespaced_logging is configured.
Autoconfiguration implementation really needs access to internal fields
and data structures to work properly.
Additionally introduces the concept of GlobalLogService takeover
internally, which allows existing LogService instances to become aware
of a new GlobalLogService. This is needed for
`useForAutoconfiguredLogging` to work as one would naturally expect,
where Loggers that may have already been created (explicitly or
implicitly) by library or third-party code are kept up to date when the
application explicitly configures logging.
Previously, log message parsing and construction were performed before
we checked if we were actually going to log anything (based on our
thresholds). This change moves the logic for both the LogMessage
construction, as well as any logic in the calling context (string
concatenation etc.) after the threshold check.
This is to enable use-cases like this:
logger.debug("Something interesting happened\p context: " &
expensiveCallToConstructFullContext())
and not have to pay the performance penalty of the string concatenation
in production settings when debug logging is turned off.
This commit represents a complete architectural overhaul of the namespaced
logging library, transitioning from a shared-memory approach to a robust
thread-safe design with proper synchronization primitives.
- **NEW**: internal `GlobalLogService` with atomic configuration versioning
- **NEW**: Thread-local configuration caching with freshness checking
- **NEW**: Separate `LogService` (copyable) and `ThreadLocalLogService` (ref)
- **REMOVED**: Manual thread state reloading in favor of automatic freshness
- All shared state protected by locks and atomics
- Configuration changes use atomic version numbers for efficient sync
- Proper cleanup with `=destroy` implementing graceful shutdown
- Thread-safe appender cloning via `clone()` method pattern
- **NEW**: Dedicated writer threads for console and file output
- **NEW**: Channel-based message passing to writer threads
- **NEW**: Batched file I/O with optimized write patterns
- **NEW**: Graceful thread shutdown on service destruction
- **NEW**: Configurable error handling with `ErrorHandlerFunc`
- **NEW**: `defaultErrorHandlerFunc` with stderr fallback
- **NEW**: Thread-safe error reporting with separate lock
- **NEW**: Robust error recovery in all I/O operations
- **NEW**: `autoconfigured` module for zero-config usage
- **NEW**: `formatSimpleTextLog` as default formatter
- **NEW**: Optional logger support for ergonomic usage
- **NEW**: Generic `CustomLogAppender[T]` with state parameter
- **NEW**: `FileLogAppender` with proper multithreaded file I/O
- **BREAKING**: Logger `name` field renamed to `scope`
- **BREAKING**: Configuration methods renamed (e.g., `setRootLevel` → `setRootThreshold`)
- **NEW**: Comprehensive test suite with 20+ test cases
- **NEW**: `testutil` module with thread-safe test infrastructure
- **NEW**: Cross-thread synchronization testing
- **NEW**: File I/O testing with temporary files
- **REMOVED**: Old test suite replaced with more comprehensive version
- Atomic version checking prevents unnecessary config copies
- Writer threads use efficient polling with 100ms idle sleep
- File writer batches messages and optimizes file operations
- Thread-local caching reduces lock contention
1. **API Changes**:
- `LogService` returned by `iniLogService` is fundamentally different
- `threadLocalRef()` required for thread-local operations
- `reloadThreadState()` removed (automatic freshness)
- Logger field `name` → `scope`
2. **Configuration**:
- `setRootLevel()` → `setRootThreshold()`
- `setThreshold()` API simplified
- `clearAppenders()` removed
3. **Appenders**:
- `initThreadCopy()` → `clone()`
- `appendLogMessage()` signature changed
- Custom appenders now generic with state parameter
```nim
let ls = initLogService()
ls.addAppender(initConsoleLogAppender())
reloadThreadState(ls)
let logger = ls.getLogger("app")
logger.info("Hello world")
```
```nim
let ls = initLogService()
let tlls = threadLocalRef(ls)
tlls.addAppender(initConsoleLogAppender())
let logger = tlls.getLogger("app")
logger.info("Hello world")
```
```nim
import namespaced_logging/autoconfigured
addLogAppender(initConsoleLogAppender())
info("Hello world")
```
- Hide initLoggingNamespace. getLoggerForNamespace should be used
instead.
- setLevelForNamespace now calls getLoggerForNamespace to ensure the
namespace exists when a caller tries to set the logging level.
Previously it was possible to call setLevelForNamespace before the
namespace was initialized, meaning that the setting had no effect.
This is especially problematic for cases where a library is using
namespaced logging with the filter set to a high level by default.
The code using that library may want to enable debug logs, and is
likely to call setLevelForNamespace in initialization code that runs
prior to the library code which instantiates the namespace.
jdb