Page MenuHomePhabricator

Replace outdated translations with source text on translation pages
Closed, ResolvedPublic

Description

Wikitext for Wikidata:Glossary/it with fuzzied sections

When a section of a translated page is fuzzied, it is enclosed in spans with the id mw-translate-fuzzy. Currently these span tags have a line break after the <span> and before the </span> in wikitext. This breaks line-sensitive wikimarkup, such as lists. I have attached a txt with the source of a page with such marking. Copypaste that into your favorite wiki and preview, and you will see what I mean. The solution is to remove the linebreaks fuzzying creates.


Version: unspecified
Severity: normal
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=50103
https://bugzilla.wikimedia.org/show_bug.cgi?id=58429

Attached:

Details

Reference
bz44328

Event Timeline

bzimport raised the priority of this task from to High.
bzimport set Reference to bz44328.
jhsoby created this task.Jan 24 2013, 11:39 PM

I don't really have a solution for this. A span shouldn't necessarily add this line break. Is it a CSS thing, or why are these line breaks introduced?

I think that the solution is to no longer add fuzzy translations in the translated page. I wrote this "feature request" a long time ago, and it's on the twn "Issues and features" page[1] as: "Fuzzied page translation message should be replaced by the source language part. We never realised this before, but it was seen in UserBase that page links can break on retags, etc. Page consistency must be preserved, and that can only happen if outdated parts are not in the aggregate."

Any thoughts?

If the unit changes enough to break page consistency, the unit marker should have been removed before remarking the page.

(In reply to comment #3)

If the unit changes enough to break page consistency, the unit marker should
have been removed before remarking the page.

In a way I agree, but what I often see are changes like this: initially, mark up was added to units, then a more knowledgeable translation admin comes up, improves marking of the page, and a lot of translations need to be updated; the units basically need stripping of markup. That of course breaks the pages, but throwing away the translations doesn't make sense, and those translations will not be suggested by TM. because of that, I think it's best to not include fuzzy translations in translated pages.

gerrit change 52215

milupo wrote:

Is this bug the reason that you can't see on the translated page any more which messages have changed in Wikidata? At present I can see in the version history only what FuzzyBot has changed. On the translated pages the translated text is replaced by English text but the changed text is not marked and even the translation percentage is not changed. Also in the translation tool is not shown which messages have been changed, I have to open every message to find out the fuzzied messages because neither the word FUZZY is shown nor the background color of fuzzied messages differs from that of up-to-date messages.

Please mark changed messages on translated pages and update translation percentage, as before.
Please mark changed message in Translation tool with either with FUZZY or/and a separate background color

At present a translator has a small chance only to see which messages have been changed.

Outdated messages can be found in the list of untranslated messages. In the new UI there will be separate tab for outdated messages and clear label too.

I can not confirm the issue with percentages are there are no steps to reproduce.

Outdated messages will not be marked in translation pages, because they are replaced with the source text.

milupo wrote:

I didn't know that there will be a new UI for the translation tool. Well, you are right, I can display the untranslated resp. outdated messages only. But I see in the version history only which pages were changed. And, I must have those pages in my watching list. Otherwise I don't see that fuzzybot changed those pages. For me as translator it is a regression that outdated messages are not marked on translated pages any more. I don't understand why a well-tried feature is removed.

milupo wrote:

Also the message diff in the Translation tool doesn't work correctly. Instead of the previous text corresponding to the current message the text of the whole page is shown and I don't have any comparison possibility more.

(In reply to comment #8)
Translation pages are still updated and the percentage indicates that 100% is not translated. I don't see any regression here.

(In reply to comment #9)
This is a transitional issue because we made internal changes how the changes are tracked. It will go away with time.

milupo wrote:

(To comment 8) No, that's not true, the percentage on the translation page indicates how much is translated. Please consider that I speak about Wikidata pages! And although FuzzyBot now changes the translated directly replacing translated text by English text, it didn't change the percentage. I saw it on Wikidata:Introduction. FuzzyBot changed the outdated messages replacing translated text by English text but the percentage remained at 100%. Now I updated translation of Wikidata:introduction, and now 100% is the correct percentage again.

BTW: There is a small cache issue: When I have translated resp. updated the last untranslated resp. outdated messsage, its translation is not taken over to the translation page. I must do a small edit in the translation tool, then it works.

(To comment 9): Thanks for your information.

Changing bug summary to reflect what was done, per commit message.

FYI this change is causing Translate quite a bit of bad press, i.e. https://www.wikidata.org/w/index.php?title=Wikidata:Project_chat&oldid=17460543#Fuzzybot (Meta users have been less whiny but not happier, in private).

I suggest that it's *very* prominently advertised in the MLEB release notes, otherwise much more of bad feelings will be seen on other wikis.

From the perspective of the reader of a translated page, it almost always much more preferable to see the outdated translation (clearly marked as such, of course), than displaying the untranslated text without any indication that an outdated translation is available. At least in the use cases I'm familiar with.

On Meta, this change made FuzzyBot remove large swathes of carefully translated text from the readers' view. For example, here because of a missing wikilink in one case and a typo in one number in the other: https://meta.wikimedia.org/w/index.php?title=Special:Contributions/FuzzyBot&offset=20130430101100&limit=16&target=FuzzyBot . And I'm not optimistic that these translations will be restored soon in a manual cleanup - it appears that Nemo's repeated call on Translators-l for volunteers to participate in this work has not met with enough resonance.

I am not familiar with the HTML/CSS formatting problem on Wikidata which appears to form the rationale for https://gerrit.wikimedia.org/r/#/c/52215/ , but was its solution really worth all those detrimental effects on readers and the extra work required from translation volunteers? This is a deeply unsatisfactory situation.

(In reply to comment #14)

On Meta, this change made FuzzyBot remove large swathes of carefully
translated text from the readers' view.

Stats: http://lists.wikimedia.org/pipermail/translators-l/2013-May/002277.html
Most of the disruption was caused by bug 47178, made worse by bug 46562. However, Tilman, given that this bug completely changed the meaning of invalidation, you may want to file an enhancement request to change the default status of the checkbox to "Do not invalidate".