Changelog¶

All notable changes to this project are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]¶

Changed¶

• Expanded API documentation and docstrings for ConsoleRenderer, processor helpers, testing helpers, the stdlib bridge, and Core.
• Internal cleanup: removed unused internals (Core._print_error, _frames._format_exception, _utils.eval_dict) and consolidated duplicated profile builders and name-splitting logic. No public API change.
• Renamed the internal Core.min_level_no attribute to Core.min_level.

0.9.0 - 2026-10-02¶

Added¶

• plainlog.testing — a framework-agnostic capture_logs() context manager and PlainlogCapture object for asserting on logs in pytest or unittest.
• Logger.flush() — block until queued records have been processed.

Changed¶

• Core.configure() no longer closes the processors it replaces — reconfiguration is an attach/detach operation.
• Logger.configure() was replaced by configure_log(), which can apply a named profile, explicit processors, or both. Profiles now return a processor list, or None to leave the pipeline unchanged.
• configure_log(..., close_before_configure=True) closes the previously registered processors before installing the new pipeline.
• Core.close() now releases the resources of every registered processor but keeps the worker thread running. The previous behaviour (release and stop) lives in the new Core.shutdown(), which is also what runs at interpreter exit.

0.8.0 - 2026-09-27¶

Changed¶

• Record schema flattened: the extra key is gone. Keyword arguments passed to logger.info(...), bind() and contextualize() now become top-level fields on the record instead of being nested under record["extra"].
• SimpleFormatter and JsonFormatter write the rendered string to record["formatted_message"] instead of overwriting record["message"].
• JsonFormatter serializes a fixed set of keys plus additional_keys; arbitrary user fields are no longer included automatically.
• SubProcessor runs its nested processors on a shallow copy of the record (previously deepcopy).
• Logger.error(), critical() and exception() request caller info (function, line, ...) by default.
• Documentation and benchmark examples now create a child logger with logger.new() and log through it rather than logging to the root logger. The benchmark also logs through a bound name so CPython's imported-name call-site overhead does not skew the comparison.
• Logger.new() auto-detected names keep the class/nesting prefix from co_qualname on Python 3.11+ (e.g. module.Class.method), falling back to module.method on 3.10; <locals> frame markers are stripped.

Removed¶

• eval_extra, eval_lambda_extra and remove_extra_items processors — fields are plain top-level record keys now.
• The extra parameter of Logger(...) and the verbose parameter of Logger(...), Logger.new() and Logger.configure(); pass keyword arguments directly and use caller_info=True for caller info.

0.7.0 - 2026-09-20¶

Added¶

• develop_no_color profile — like develop but without ANSI colors.

Changed¶

• Logger.new() inherits the parent logger's verbose setting when verbose is not given, instead of always defaulting to False.
• SimpleFormatter now renders a full ISO timestamp and appends extra instead of dropping it.
• The default profile now uses SimpleFormatter (writing to stdout).
• develop honors a stream kwarg; default, file and fingerscrossed_file accept a format kwarg.
• Profiles now set verbose explicitly, so applying a profile no longer inherits a previous verbose setting. Logger.configure(verbose=None) and Logger.new(verbose=None) keep inheriting, as before.

Removed¶

• DefaultFormatter — use SimpleFormatter instead.
• Profiles simple, fast and console_no_color (console_no_color is replaced by develop_no_color).

Fixed¶

• std_handler_default and std_handler_develop now forward their kwargs (e.g. stream, format) to the wrapped profile.
• Documentation corrections: profile count, changelog comparison links, and the missing callable logger(...) form plus manual context() / reset_context() usage.
• Corrected Logger docstrings: removed the non-existent core attribute and refreshed the configure(), new() and __call__() descriptions.

0.6.0 - 2026-09-18¶

Added¶

• redact_fields and redact_by_pattern processors — mask sensitive extra values by exact key name or substring pattern. contributed by @kashyapm94

Changed¶

• License changed from Apache-2.0 OR MIT to BSD 3-Clause.

0.5.0 - 2026-09-13¶

Changed¶

• Everything is a processor. configure(handler=...) is replaced by configure(processors=[...]). The Core runs an ordered processor list. The usual pipeline is filter, format, handle.
• Formatters are processors. SimpleFormatter, DefaultFormatter, JsonFormatter and ConsoleRenderer now set record["message"] and return the record instead of returning a string.
• Merged formatters.py and handlers.py into plainlog.processors. The plainlog.formatters and plainlog.handlers modules are gone.
• Handler renames. StreamHandler → Stream, FileHandler → FileWriter, AsyncHandler → AsyncBridge, FingersCrossedHandler → FingersCrossed.
• JsonHandler, ConsoleHandler and DefaultHandler removed. Build the equivalent pipeline from a formatter plus Stream.
• configure(print_errors=...) removed. Processor errors are now emitted by the new print_processor_error processor instead of a Core flag.
• Tests consolidated into tests/test_processors.py.
• Documentation reorganized; docs/handlers.md is now docs/processors.md.

Added¶

• SubProcessor — runs a nested processor pipeline on a copy of the record.
• print_processor_error — prints a record's processor error to stderr.
• ProcessorProtocol, ProcessorCloseProtocol and UniversalProcessorProtocol in plainlog._base.
• Fork support: the Core worker is restarted in forked child processes.
• verbose=True to Logger, adding caller info (function, line, ...) to the record.

Removed¶

• BaseHandler, ProcessingHandler and the preprocessor concept.
• preformat_message.

Fixed¶

• Thread-safety of concurrent logging and reconfiguration.
• Async handler (AsyncBridge) reliability.

0.4.0 - 2026-08-28¶

Changed¶

• Log level is now a plain int. The Level NamedTuple was removed. record["level"] is an int (e.g. 10, 20, 30), and the human-readable name is stored alongside it as record["level_name"] (a str).
• Record schema simplified. The separate context and kwargs keys are now merged into the record's extra dict. Caller/context variables and per-call keyword arguments are all available under record["extra"].
• Exception handling. exc_info is replaced by exception, a RecordException (pickle-safe) stored under record["exception"].

Added¶

• level_name (the human-readable level name) is written directly onto every record alongside level.
• Performance improvements on the logging hot path: time() is used instead of datetime for the record timestamp, and benchmarks against stdlib logging were improved.

Removed¶

• Core.level() method. Level validation is handled internally by _validate_level (which now uses logging._checkLevel).

Fixed¶

• Bug with caller-level resolution in DevelopHandler.
• Documentation references and examples updated to the new record schema.

0.3.0¶

Changed¶

• Renamed internal entry points; the main interface now lives on logger (logger.debug, logger.log, logger.configure, ...). The Core is no longer part of the public surface.

Added¶

• Initial documentation site and runnable doc examples (pytest-examples).
• More tests and benchmark coverage.