Header and link with spaces:
== Nbsp test == [[#Nbsp test]]
id="Nbsp_test" href="#Nbsp_test"
Header and link with non-breaking spaces:
== Nbsp test == [[#Nbsp test]]
id="Nbsp.C2.A0test" href="#Nbsp_test"
We have non-breaking space → space replacement for a long time since T17248. We need to make exactly the same replacement for headers.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | MaxSem | T75092 Anchors to section names for non-ASCII letters are encoded in the URL | |||
| Resolved | matmarex | T172493 mw.util.escapeId should log deprecation warning | |||
| Resolved | MaxSem | T152540 Migrate to HTML5 section ids | |||
| Resolved | MaxSem | T175725 Deploy HTML5 sections to WMF production | |||
| Resolved | MaxSem | T90902 Non-breaking space in header ID breaks anchor |
This problem is regularly occurring on Russian Wikipedia due to its active use of non-breaking space before dashes and other entities, and many topics on the technical forum cover it, so it's better be fixed.
This change would have to be made in both Parsoid and PHP core. We can leave the legacy IDs alone; we should probably only normalize the HTML5 ids. With $wgFragmentMode=['html5','legacy'];:
$ (echo '==Foo bar==' ; echo '[[#Foo bar]] [[#Ж]]' ) | pance/parse.php <div class="mw-parser-output"><h2><span id="Foo.C2.A0bar"></span><span class="mw-headline" id="Foo bar">Foo bar</span><span class="mw-editsection"><span class="mw-editsection-bracket">[</span><a href="/~cananian/mediawiki/index.php?title=CLIParser&veaction=edit&section=1" class="mw-editsection-visualeditor" title="Edit section: Foo bar">edit</a><span class="mw-editsection-divider"> | </span><a href="/~cananian/mediawiki/index.php?title=CLIParser&action=edit&section=1" title="Edit section: Foo bar">edit source</a><span class="mw-editsection-bracket">]</span></span></h2> <p><a href="#Foo_bar">#Foo bar</a> <a href="#Ж">#Ж</a>
Note that there's an underscore substitution happening in the [[#...]] processing, but it's not happening in the generated HTML5 id. (The thing that looks like a space in the HTML5 id is actually the non-breaking space  .)
Added @MaxSem, as the above inconsistency in links and headings will make html5 ids break.
Change 388365 had a related patch set uploaded (by MaxSem; owner: MaxSem):
[mediawiki/core@master] Remove nbsp and similar characters from section IDs
Change 388365 merged by jenkins-bot:
[mediawiki/core@master] Remove nbsp and similar characters from section IDs
Change 416890 had a related patch set uploaded (by C. Scott Ananian; owner: C. Scott Ananian):
[mediawiki/services/parsoid@master] Remove and similar characters from section IDs
Change 416890 merged by jenkins-bot:
[mediawiki/services/parsoid@master] Remove and similar characters from section IDs
Mentioned in SAL (#wikimedia-operations) [2018-03-12T20:44:51Z] <arlolra> Updated Parsoid to 16ced34 (T188670, T90902)
On fr.wp, « X » is automatically replaced with « X ».
Legacy IDs seem to have been turned from .C2.AB.C2.A0X.C2.A0.C2.BB into .C2.AB_X_.C2.BB. Can you confirm that? Is that reversible?
@Pols12 that seems like possibly an interaction with french space armoring (T197902).
I can confirm that in legacy mode the .C2.A0 are being converted to underscore:
$ (echo '« X »' ; echo; echo '[[#« X »]]' ; echo; echo '==« X »==') | php maintenance/parse.php parse.php: warning: reading wikitext from STDIN. Press CTRL+D to parse. <p>« X » </p><p><a href="#.C2.AB_X_.C2.BB">#« X »</a> </p> <h2><span class="mw-headline" id=".C2.AB_X_.C2.BB">« X »</span><span class="mw-editsection"><span class="mw-editsection-bracket">[</span><a href="/~cananian/mediawiki/index.php?title=CLIParser&veaction=edit&section=1" class="mw-editsection-visualeditor" title="Edit section: « X »">edit</a><span class="mw-editsection-divider"> | </span><a href="/~cananian/mediawiki/index.php?title=CLIParser&action=edit&section=1" title="Edit section: « X »">edit source</a><span class="mw-editsection-bracket">]</span></span></h2>
But if I am understanding correctly, prior to this patch we would generate id=".C2.AB.C2.A0..." but href=".C2.AB_.... That is, the tag ids didn't match the href attributes and all the links were broken. This patch fixed the situation, by changing the ids to match the hrefs we'd been generating all along.
OK, thanks!
The link I’ve seen was created from a gadget which read directly the source code to get the ID, so it let generating correct URLs. But the gadget is currently used by only 2025 (including 242 active) users, so I think there is not a lot of issues of this kind.