The PHP parser currently recognizes RFC/PMID/ISBN links even if they are buried in unrelated text, for example:
fooRFC 1234bar
We should add \b restrictions to the regexp to ensure that magic links stand apart from other text.
See also bug 28950, which asks for the whitespace restrictions in magic links to be loosened somewhat.
Version: 1.24rc
Severity: normal
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Magic link fixes. | mediawiki/services/parsoid | master | +384 -320 |
Source code file name (and path) welcome in case a contributor would like to give fixing the regexes a shot. Marking as easy.
Unfortunately parser changes are never quite so "easy" -- even though the source code change is small, before deployment we need to grep through all our existing wikis to be sure that no one is using RFC links that will be broken. Perhaps some language wiki uses prefixes for possessives and quite likes the current behavior. We won't know until we look.
But sure, patch welcome! Just don't be surprised if it's not immediately committed. (Oh, and be sure to include parser tests with your patch.)
Source code file name is includes/parser/Parser.php -- here's the code in question:
Hm. The bot seems to be asleep. Here's a patch for this: https://gerrit.wikimedia.org/r/133650
Change 178711 had a related patch set uploaded (by Cscott):
WIP: Magic link fixes.