PyThaiNLP 5.3 Change Log

[wannaphong]

[WIP]

This release modernizes the codebase, delivering a 62x reduction in peak import memory (thanks to @what-in-the-nim), better support for offline and read-only environments, and achieving typed package status through improved type annotations.

This will be the last minor update in 5.x series. We expected the next version to be 6.0, a major update that may introduce breaking changes.

Minimum required Python version is now 3.9. The library ensures compatibility across Python 3.9–3.14.

Many updates were AI-assisted; see Pull Requests for specific prompts and implementation details.

• Add Python 3.14 support (update nlpO3 to 1.4.0)
• Fix royin romanization Fix royin romanization for Thai consonant clusters and syllable boundaries #1172
• Fix final consonant classification in check_marttra() Fix check_marttra() Thai final consonant classification #1173
• Lazy load dictionaries to reduce memory usage Reduce peak memory on import by 62x #1186
• Fix Kho Khon alphabet issue in tltk transliteration Fix KeyError when transliterating text containing obsolete Thai letter ฅ #1187
• Migrate configurations to pyproject.toml Consolidate configuration into pyproject.toml #1188 Integrate tox.ini into pyproject.toml #1226 Remove obsolete configuration files (.pep8speaks.yml and .pyup.yml) #1239
• Update type hints; Use Python 3.9 features Modernize to Python 3.9: type hints with future annotations, exception handling, and code quality #1189 Fix type hint mismatches with None defaults throughout codebase #1190 & many more PRs.
• Remove duplicated entries in volubilis dictionary volubilis_words_th.txt: Remove trailing whitespaces + remove duplicate entries #1200
• Remove star imports Remove star imports and enable F403/F405 Ruff checks #1207
• Remove requests dependency Replace requests with urllib.request to reduce core dependencies #1211
• Make package zip-safe Make package zip-safe by migrating to importlib.resources #1212
• Ensure thread-safety for tokenizers Ensure thread-safety for word_tokenize() wrapper functions #1213
• Better support for offline and read-only environment Add PYTHAINLP_OFFLINE support, is_offline_mode(), and PYTHAINLP_DATA env var to corpus path handling #1309 Lazy create db.json; Introduce PYTHAINLP_READ_ONLY with scoped write protection #1317
• Improved documentation; code cleanup; more tests

[wannaphong]

@bact I thinking we should releasing the next PyThaiNLP version as new major (6.0). It will allow us can break API-compatible without concerns.

[bact]

I don't think we have break anything so far. But that's a good point.

What about releasing what we have now as 5.3 + announce that the next version will be 6.0 that can break things?

Maybe we can remove some legacy features in 6.0. For example those that need very old or unmaintained dependencies.

[wannaphong]

I don't think we have break anything so far. But that's a good point.

What about releasing what we have now as 5.3 + announce that the next version will be 6.0 that can break things?

Maybe we can remove some legacy features in 6.0. For example those that need very old or unmaintained dependencies.

Agree

[bact]

Ok. Let's release 5.3.
Maybe this week?

[bact]

@wannaphong I don't have anything to add to this 5.3. If you're ok, maybe we can release it today or tomorrow and announce during FOSSASIA.

[wannaphong]

@wannaphong I don't have anything to add to this 5.3. If you're ok, maybe we can release it today or tomorrow and announce during FOSSASIA.

Do it 💯 Many features are slow than I expect from my PhD student work. I think version 5.3 will also be a celebration of PyThaiNLP's 10th anniversary!

[wannaphong]

I am sad. I don't have time for FOSSASIA this year. :(

[bact]

Ok. Will take the notes here and in the CHANGELOG.md for the release notes.

For 6.0, maybe aim it around June, make it 3-month cycle.

One of primary targets is to reduce size of codebase by removing outdated, legacy, unsupported, dependency-difficult features.

[bact]

5.3.0 is here

• https://github.com/PyThaiNLP/pythainlp/releases/tag/v5.3.0
• https://pypi.org/project/pythainlp/5.3.0/

https://pythainlp.org/docs/ is to be regenerated for 5.3 API document.

Thanks again to @what-in-the-nim