Technical Community Newsletter/2024/October
 |
The Wikimedia Technical Community Newsletter shares highlights and news from the past quarter and information about events coming up in the next quarter.
What's new
[edit]PHPUnit parallel testing
[edit]Since the middle of June we've been slowly rolling out support for and improvements in parallel PHPUnit tests, per the announcement on the tech mailing list.
- The PHPUnit "extensions" suite now runs in parallel for CI jobs of all components except for Core, reducing the runtime for some jobs by almost 50%.
- Mediawiki core now has new Composer tasks including
phpunit:prepare-parallel:extensions,phpunit:parallel:databaseandphpunit:parallel:databaselesswhich are used by CI per the above, and can also be used by developers locally to run tests in parallel.
Running the tests in parallel means reordering the tests, which can cause hard-to-debug CI run failures where test state is not sufficiently isolated. We have written a short guide to debugging failed parallel test runs to help developers.
Article Country Model exceeds initial performance goals
[edit]Led by Isaac Johnson as part of the Encyclopedic Content objective, the WMF Research team has developed the Article Country Model, a powerful tool for classifying Wikipedia articles by country. This machine learning model, which achieves 97% precision and 77% recall, addresses a crucial need in organizing and accessing knowledge and knowledge gaps.
The Article Country Model combines three key data sources (Wikidata properties, Wikipedia categories, and wikilinks) to capture the various ways articles relate to countries, from direct connections like geographical locations to more nuanced cultural associations. By expanding their data sources and refining the approach, the team improved the model's ability to correctly identify article topics, especially in languages like English where we saw a 25% increase in accuracy.
Moving forward, the WMF Language and Product Localization team plans to integrate the Article Country Model into topic suggestion workflows for content translation.
Coming soon
[edit]CodexPHP
[edit]The Design System Team is working in partnership with community developers to provide a standalone library for generating Codex-styled HTML markup in PHP (phab:T373708). We are aiming to publish an initial v0.1.0 release to Composer in the coming weeks.
We hope this will simplify the process for WMF and community developers to use Codex components in their projects, regardless of whether or not they are presenting rich JS applications to their users. A special thank you to DoÄu for initiating this effort and contributing the bulk of the code.
Community metrics: Phabricator, Gerrit, and GitLab
[edit]- Number of tasks created in Q3/2024: 7223
- Number of tasks closed in Q3/2024: 6848
- Number of different people who created tasks in Q3/2024: 1000
- Number of different people who closed tasks in Q3/2024: 480
362 people wrote patches (143 of them being volunteers) in Q3/2024.
144 people (23 of them being volunteers) requested 1642 merges in 165 repositories in Q3/2024
WM:TechBlog Posts
[edit]- Wikidata and Artificial Intelligence: Simplified Access to Open Data for Open-Source Projects
- An Introduction to Wikipedia: A free online course for beginners
- Wikidata at the Open-Source Conference
- Language and Internationalization/Newsletters/4
- Creating racehorse parent-child relationship trees with Wikidata Graph Builder
- Empowering Africaâs Wikimedia Tech Talent: Wiki Mentor Africa â First Cameroon Experience
Upcoming conferences and workshops
[edit]- MediaWiki Users and Developers Conference Fall 2024
- Conference on Climate Justice, Indigenous Voices and Wikimedia Platforms
Get involved!
[edit]Are you interested in learning more or volunteering to improve Wikimedia Technology? Get involved!