v6.0.0

Breaking Changes

• Automatic .env loading has been removed. The library no longer reads .env files or depends on dotenv, even optionally. Configuration now comes only from existing process.env values or LoggerAdaptToConsole({ envOptions }).
• logUsingWinston has been renamed to logUsingConsoleJson. If you were importing or referencing the old helper directly, update to the new name.

New Features

• Built-in structured redaction. LoggerAdaptToConsole() now accepts a Pino-style redact option:
  • shorthand array form: redact: ['password', 'headers.authorization', 'items[*].token']
  • object form with custom censor: redact: { paths: [...], censor: 'MASKED' }
• Redaction runs on the final output object. It applies after transformOutput, before serialization/write, and onLog receives the redacted result.
• Minification-resistant caller detection. @filename now prefers V8 callsites, skips internal logger frames by function identity where possible, and falls back to stack parsing only when needed.

Improvements

• Jest replaced Mocha. The test runner is now Jest via ts-jest, with a new jest.config.cjs and TypeScript configured for Jest globals.
• Safer browser/bundled filename behavior. Documentation now explicitly treats @filename as best-effort when a consumer bundles the library and their app into a single browser file.
• No more manual .env scanning logic. The legacy Env.loadDotEnv() implementation is now a no-op compatibility shim instead of touching the filesystem.
• Logging pipeline cleanup. Internal naming and structure were updated to match the current architecture:
  • logUsingConsoleJson replaces stale Winston naming
  • final output shaping, transform hooks, redaction, and serialization are now separated more clearly
• Test suite cleanup. Removed confusing no-op await console.*(...) usage, converted the helper stack-format test to assertions instead of console output, and added targeted coverage for:
  • reserved message handling through the real console.log(...) path
  • fallback-pattern skipping in browser-style stack parsing
  • callsite-based filename detection
  • structured redaction behavior
• Stronger configuration coverage. Added an audit test to ensure every documented CONSOLE_LOG_* setting in README.md is referenced by the test suite.
• More deterministic permutation coverage. Added a table-driven test that exercises permutations of string messages, errors, one or more context objects, and explicit { level: ... } overrides and asserts a normalized final payload.

Follow-up Behavior Changes

• Reserved message handling is now consistent.
  • console.log('dude', { message: 'hi there' }) now produces message: "dude - hi there"
  • object-valued message is preserved under @messageObject instead of being flattened into top-level keys
  • when only message: { ... } is provided, the canonical message field remains present with the usual placeholder text

Bug Fixes

• @packageName now works in Node ESM startup paths. The logger no longer relies solely on require(...) to discover package.json, so LoggerAdaptToConsole() in "type": "module" projects no longer emits the misleading "<not-yet-set> Please await..." placeholder.
• Early logs are buffered while async package-name lookup completes. Startup logs emitted immediately after LoggerAdaptToConsole() are now replayed once package-name resolution finishes, instead of racing initialization.
• Redaction no longer mutates caller-owned input objects. Sensitive fields are still redacted in the final emitted log object, but the original application objects passed to console.log(...) remain unchanged.

Documentation

• ARCHITECTURE.md was rewritten and updated to reflect the real implementation instead of the older Winston-based design.
• README.md now documents:
  • Jest-based contributor/test setup
  • no automatic .env loading
  • best-effort @filename semantics in bundled browsers
  • the new redact API with examples and guidance