Wikimedia Language engineering/Reports/2017/January

December 2016

Wikimedia Language engineering monthly report for January 2017

February 2017

Monthly updates from the Wikimedia Language team focused on technical development of the software they maintain. Names of people assumed to contribute as volunteers are in bold text.

Content Translation

Content Translation allows people to translate Wikipedia articles. Also check out out the newsletter.

About 20 patches were merged this month.

Niklas Laxström fixed a bug related to long and non-ASCII headers. T154279
Fomafix maintained the code by replacing a one type of deprecated jQuery construct.
Fomafix updated a few API queries to use newer format.
Santhosh Thottingal fixed some scrolling and position related regressions.
Santhosh Thottingal made the code to pass the target title to parsoid, which is needed for template translation. T154297
Santhosh Thottingal fixed publishing failure related to templates.
Kartik Mistry fixed many errors reported by Stylelint. T152079
Amir Aharoni made the user interface to always call template parameters as fields. 151865
Fomafix updated various URLs in the source code to use HTTPS.
Fomafix updated the code to pass arrays to API module without converting them to strings.
Santhosh Thottingal and Kartik Mistry replaced jshint, jscs, jsbeautify with eslint in cxserver.
Santhosh Thottingal changed the h1,h2.. ids to fixed length sha256 based ids.

Usage data

Details	Numbers
Articles created (during the month)	12040
Articles created till end of month	167040
Number of new translators (during the month)	1032
Highest number of articles created by one user (cumulative number till end of the month)	3990
Highest number of translators for a Wikipedia (cumulative number till end of the month)	4194 (es)
Total number of translators with only 1 published article (cumulative number till end of the month)	16297
Total number of translators with only 2 published articles (cumulative number till end of the month)	3332
Total number of translators with only 3 published articles (cumulative number till end of the month)	1587
Total number of translators with only 4 published articles (cumulative number till end of the month)	877
Total number of translators with >= 5 and < 20 published articles (cumulative number till end of the month)	2592
Total number of translators with >= 20 and < 100 published articles (cumulative number till end of the month)	919
Total number of translators with >= 100 published articles (cumulative number till end of the month)	250
Articles deleted (during the month)	647 / 1194 (see comment)

There's a spike in deletions because one-time spike of bad machine transations in Malagasy Wikipedia (mg), after which 547 pages were deleted. Without it, the number is 647.

Comparisons

Articles published.
Deletion ratio.
Distribution of users per number of translations they published.

Translate

Translate extension allows translation of software interface, manuals, announcements, banners etc. on Wikimedia special projects, translatewiki.net and other sites. Translate is part of MLEB.

About 10 patches were merged in this month.

Phantom42 added support for some HTML tags to appear as insertables during translation. T106632
Huji Lee made translation memory match percentages localised. T153514
Fomafix maintained the code by replacing a one type of deprecated jQuery construct.
Justin Du maintained the code by accessing the UtfNormal library in non-deprecated way. T153994
Max Semenik maintained the code by correcting some class annotations.
Niklas Laxström rewrote the Android file format support code to use shared code for plural handling.
Niklas Laxström fixed the handling of HTML entities in the Android file format. T148549
Niklas Laxström fixed a regression that prevented translation admins from performing message renames properly. T148892
Aaron Schulz strengthened page translation code against replication lag.

Usage data

Translatable pages: Meta 3322 (+42) ; MediaWiki.org 3287 (+34) ; Commons 585 (+84) [1][2][3]
Translate extension is installed on 215 wikis (+2) Wikiapiary

Universal Language Selector

Universal Language Selector provides language choice, input methods and web fonts. It is part of MLEB.

Fomafix maintained the code by replacing a one type of deprecated jQuery construct.
Fomafix improved the code to avoid flash during page rendering. First patch of two was merged in January.
Fomafix maintained the code by removing jquery.i18n which is now bundled with MediaWiki core. T153635
Fomafix maintained the code by using mediawiki.storage module from MediaWiki core. T153635
Ed Sanders updated build configuration.
Amir Aharoni and Kartik Mistry updated the bundled upstream libraries jquery.uls and jquery.ime.

Other extensions

Babel (part of MLEB):

This, that and the other fixed issues related to category names containing language codes with variants. T146778
This, that and the other made edit summaries of automated edits to use the content language as opposed to interface language. T46623

cldr (Common Locale Data Repository) (part of MLEB):

Ed Sanders updated build configuration.

CleanChanges (part of MLEB):

Rudloff fixed code that was causing PHP notices after a MediaWiki core change. T151825

LocalisationUpdate (part of MLEB):

No changes.

TranslationNotifications:

No changes.

MLEB

MLEB 2017.01 was released this month. It contains a fix for reported incompatibility in the Babel extension, among other changes.

MediaWiki core internationalization

This is a new, experimental section in this report. Coverage might not be comprehensive.

Federico Leva added many new fallbacks to existing languages. gerrit
Leszek Manicki removed unused code in LanguageConverters.
This, that and the other made invalid UTF-8 sequences to be replaced with U+FFFD in edit summaries. T95353
Justin Du added a WebAPI to set the language of a wiki page. T74958
David Sn added a {{PAGELANGUAGE}} magic word to get the language of a wiki page. T59603
Amir Aharoni fixed a mistake in grammar transformations for he.
Jack Phoenix enhanced the MessagesPreLoad hook to pass a language code.
Aaron Schulz maintained MessageCache
Bartosz Dziewoński fixed display of block durations given in relative format. T55907

Project Milkshake

Project Milkshake consists of four re-usable jquery internationalization libraries.

jquery.i18n

No changes

jquery.uls

The Discoverer updated the language name of gom-Deva and gom-Latn. T137222

jquery.webfonts

No changes.

jquery.ime

The Discoverer updated the name of input method for gom. T137222

translatewiki.net

Stats

About 85 (+5) new users
About 375 (+30) translators made about 38.000 (-8.000) edits
About 95 (+20) translators reviewed about 9.500 (=) translations

Sources: [4] [5] [6] [7] [8]

TwnMainPage

No changes

Deployments and other site related updates

Marko Obrovac upgraded cxserver to use Node v6.9.1.
Kartik Mistry added more information in MT error logs.

Cross team work/requirements

Started discussion with Analytics team to improve metrics data collection for Language team projects

Team status

Development was slow during this month due to travels and holidays.