Before we deploy numeric sorting on English Wikipedia, we should test it on a real wiki with actual categories and defaultsort keys. Johan has offered to start a proposal on Swedish Wikipedia to turn on numeric sorting so that it can be more thoroughly tested.
The actual change will be switching svwiki's collation from uca-sv to uca-sv-u-kn.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | kaldari | T128483 Fix category headers for pages that begin with numbers | |||
| Resolved | kaldari | T8948 Natural number sorting in category listings | |||
| Resolved | kaldari | T142113 Test numeric sorting on Swedish Wikipedia |
Will the number of bot-created articles be a problem? They're categorized as well, after all.
Question that came up: uca-sv-u-kn takes Swedish digit grouping and decimal comma into account, right? Have we specifically tested it?
(Swedish digit grouping differs from English. 100.000,00 would usually be written 100 000,00 in Swedish, occasionally as 100,000.00, as opposed to 100.000,00 in English.)
Do we have a nifty solution for articles that have solved this problem using DEFAULTSORT already, e.g {{STANDARDSORTERING:0011 Freunde}}? Maybe running a script to find articles where defaultsort starts with 0/00/000 but otherwise matches the article name? Or will it even affect things at all? The numerical value is still the same.
@Johan: Any time a number is separated with a non-numeric character, UCA collation will treat the separator as a string and the separated pieces as separate numbers. This is true in English, Swedish, and all UCA collations. Separators are considered ambiguous characters and thus not treated as parts of numbers. This can be worked around (in the case of large integers) by removing the separators in DEFAULTSORT keys. So for example, if you had an article entitled "9,999 bottles of beer", you could add a DEFAULTSORT key of "9999 bottles of beer".
Change 304262 had a related patch set uploaded (by Kaldari):
Updating $tailoringFirstLetters for Swedish
Change 304262 abandoned by Kaldari:
Updating $tailoringFirstLetters for Swedish Per https://ssl.icu-project.org/trac/browser/icu/trunk/source/data/coll/sv.txt
Reason:
Already in the list :P
Posted in the discussion that I'm reading consensus as "the change is fine as long as it doesn't sabotage anything that's working right now". Giving folks a chance to protest, but I don't see any reason anyone would if they haven't so far.
Change 306216 had a related patch set uploaded (by Kaldari):
Switching Swedish Wikipedia to uca-sv-u-kn collation
Change 306216 merged by jenkins-bot:
Switching Swedish Wikipedia to uca-sv-u-kn collation
Mentioned in SAL [2016-08-23T18:12:13Z] <thcipriani@tin> Synchronized wmf-config/InitialiseSettings.php: SWAT: [[gerrit:306216|Switching Swedish Wikipedia to uca-sv-u-kn collation (T142113)]] (duration: 00m 58s)
Deployed to Swedish Wikipedia:
https://sv.wikipedia.org/wiki/Kategori:Musikgrupper_med_syskon
Seems to be working well so far. updateCollation.php script is still running. Will probably take a few hours to finish.
No problems reported so far, except for that it doesn't work well with separators. But we knew that.