1.7.2

2023-12-03 release

• Reformatted code using black with default parameters
• Automatically refactor code using Sourcery

Translator

Fixes

• Persistent RabbitMQ messages and queues: messages are no longer lost upon RabbitMQ restart (happens once in 300 days, but still happens)
• Tweaks on single-word definitions: fetch English-language entries' definition if a single-word definition is given for an entry.
• Use caching for NLLB

Wikistats generator

• Implement linear regression to determine future milestones

1.7.1

Minor tweaks on translation module

1.7.0: use HAProxy

Feature

• Integrate HAProxy with supervisor

Fixes

• check definitions for foreign-language untranslated word and use word-to-word translations
• word-to-word translations are enabled if the word is in a whitelist
• if the language section was not detected, skip the section altogether

1.6.0: Integrate NLLB and more

New features

• Integration of NLLB as a translation method
• Integration of OpenMT as a translation method
• Add parser for fr.wiktionary
• Add parser for zh.wiktionary
• Add RabbitMQ producer/consumer to push pages asynchronously
• Add tests for Python 3.9 up to 3.11

Improvements/Fixes

• Improve parser for en.wiktionary
• Create config files for a multi-process PostgREST

Minor API change, refactoring

Summary

• Refactoring: number of in-code serialisers has been reduced
• Support read operations on additional data outside PostgREST

What's Changed

• Update to 1.3.0 by @radomd92 in #26

Full Changelog: 1.2.31...1.3.0

Fixes / more pacific service manager / database lookup

• Service manager no longer kills already-spawned process.
• wiktionary_irc kills its predecessor by itself, like a grown-up, thus simplifying its launcher structure.
• Faster database lookup API through memory caching: use that API on read-heavy operations if you have a lot of spare RAM to use. This is practical for big Wiktionary dump processing as table lookups were painfully slow. It takes time to load at the beginning, but once loaded a laptop could breeze through 10,000s of pages in a blink of an eye.
• ORM-related Word objects can be serialised into Entry ones
• Database updates
• Webui: language view (very slow, but still workable)
• time_this decorator is now much more precise (down to the millionth of second)
• More categories to the category generator

Code refactoring, fixes

• Service Manager now controls the ports on which associated processed listen on.
• Added estonian parser
• Following the release of a new Python 3.6, Travis file has been edited to make it a little bit more future-proof
• Language and word databases have been updated

Updates, fixes

• Database updates
• Added hungarian and icelandic parsers
• Added subcategory crawlers to increase in efficiency on low page count non-lemmata word categories.
• Fixed postprocessors and lemma retrieval

Fixes

• Raise custom exceptions on parsers and on Output class
• Use Entry object in entry_translator and in translation API instead of tuples
• Chnage pip3.6 to a Python module call

Update script

Automatically updates botjagwar periodically.