v1.1.0

This release brings a major infrastructure upgrade with the integration of long-standing contributions from @cquest, improved caching strategy, and modernized packaging.

✨ What's New

🎯 Integration of @cquest's Work (#9)

After several years, we've finally merged the valuable contributions from @cquest's cquest-11rc1 branch (originally from 2016-2018):

Enhanced Phonemicization Rules:

• Better handling of complex vowel combinations (ae, ei, oeu)
• Improved nasal sound processing (m → n before labial/dental consonants)
• Special cases like "oeufs" → "eu"
• Enhanced support for "y" at word beginning
• More accurate silent consonant handling (including "gn": seigneur → senieur)
• Better duplicate letter removal
• Compiled regex patterns for better performance

Extended Synonym Coverage (#6):

• Added ~48 new synonym mappings (175 → 223 entries)
• Additional address abbreviations and variations
• Includes mappings like "clef/clefs → cle/cles", "gir → giratoire"
• Fixed duplicate entries

🔧 LRU Cache Implementation (#7)

Replaced the simple dictionary cache with a memory-efficient LRU (Least Recently Used) cache:

• Configurable cache size via PHONEMICIZE_CACHE_SIZE setting
• Default: 500,000 entries (~86 MB), suitable for most French address datasets
• Recommendations:
  • 500K entries (~86 MB): Default, suitable for most datasets
  • 1M entries (~172 MB): For larger datasets with more unique words
  • 250K entries (~43 MB): For memory-constrained environments
• Fixed race condition in cache initialization
• Prevents unbounded memory growth while maintaining performance

🔧 Infrastructure & Development

• Modernized Packaging (#5)

  • Migrated from setup.py to pyproject.toml (PEP 517/518 compliant)
  • Modern build system with setuptools>=65.0
  • Support for Python 3.9 through 3.14

• CI/CD Setup (#8)

  • Added GitHub Actions workflow for automated testing
  • Redis integration in CI pipeline
  • Automated pytest and coverage reporting

• Development Environment

  • Added dev dependencies: pytest, pytest-cov, build, twine

📝 Documentation

• Added cache configuration guidance in README
• Memory usage recommendations for different cache sizes
• Improved wording and examples

🔄 Upgrading from 1.0.1

This release is fully backward compatible:

pip install --upgrade addok-fr

Optional: Configure cache size in your Addok config if needed:

PHONEMICIZE_CACHE_SIZE = 500_000  # Adjust based on your needs

No other configuration changes required.

---

Full Changelog: 1.0.1...v1.1.0