The valid chars in titles are configurable in MediaWiki via the $wgLegalTitleChars global. We should expose this through the API (general section I presume), and use it in our tokenizer to recognize the relevant chars.
Version: unspecified
Severity: normal
See https://gerrit.wikimedia.org/r/60852 which fixed our link parser to recognize the default $wgLegalTitleChars. The PHP parser uses this regexp to match links at Parser.php around line 1780:
$tc = Title::legalChars() . '#%';
$e1 = "/^([{$tc}]+)(?:\\|(.+?))?]](.*)\$/sD";
Where Title::legalChars() just returns $wgLegalTitleChars.
Change 173051 had a related patch set uploaded by Arlolra:
(Bug 47651) Expose legalTitleChars through the API
Change 173051 had a related patch set uploaded by Arlolra:
Expose legaltitlechars through the API
mediawiki-title makes use of general.legaltitlechars
https://github.com/wikimedia/parsoid/commit/9cdbd2bc7387f4155b671495b6a78ea5fe0a6672