Bug: %SEARCH{}% does not handle correctly topic names containing non-alphanum chars

If the topic name in %SEARCH{}% command contains non-alphanum chars (for exemple "-", or "ã", ...), topic(s) are not returned by SEARCH.

CannotExcludeL10NTopicFromSearch exposes the same kind of problem, about the excludetopic parameter and "non english words".

Test case

while
%SEARCH{ "blabla" topic="Foo001*" nosearch="on" format="found: $topic"}%
returns "found: Foo001-001".

(in my case, "-" are authorised in wiki pages because they represent part numbers, but the problem is the same with "non english" chars : éèàã ... as mentionned in CannotExcludeL10NTopicFromSearch).

Environment

-- NicolasRaibaut - 17 Jan 2006

Impact and Available Solutions

Follow up

Fix record

The problem comes from _makeTopicPattern subroutine in lib/TWiki/Search.pm.

This subroutine, used to reformat searched topic names and excluded topic names, assumes topic names are strictly made of alphanumeric chars [A-Z][a-z][0-9] : s/[^\*\_$TWiki::regex{mixedAlphaNum}]//go;

All other chars are stripped.

-- NicolasRaibaut - 17 Jan 2006

(partial) workaround : add in the s/[^\*\_$TWiki::regex{mixedAlphaNum}]//go; command "authorised chars"... (but this does not solve problems exposed in CyrillicWikiWordError and GermanUmlauteBreakWikiWords).

Ideally, _makeTopicPattern should :

• suppress only forbidden chars specified in $securityFilter in TWiki.cfg,
• and/or manage the character set autorised in WikiWords (in the same way as lib/TWiki/Render.pm, after applying patchs proposed by CyrillicWikiWordError and GermanUmlauteBreakWikiWords, but this should be a little bit more complicated...)

-- NicolasRaibaut - 19 Jan 2006

Discussion