Difference between revisions of "Issue:Magic word in titlematch is not escaped"
(→Reply: line 2300) |
|||
Line 21: | Line 21: | ||
== Reply == | == Reply == | ||
Macros like PAGENAME can be used anywhere within a DPL query - not ony in 'titlematch'. The problem is that the expansion happens outside DPL and DPL does not know that there once has been a call of such a macro when it parses its parameter content. Are you talking about the place where "linksto", "linksfrom", "titlematch" etc. conditions are being converted to SQL statements (~ line 1900, ~2300)? A more general approach would be to run the sanitizer before all parameters are being analysed, i.e. at the top of the parameter parsing loop ( ~ line 360 ..). But I am not sure if this would create unwanted side effects. --[[User:Gero|Gero]] 07:49, 16 February 2010 (UTC) | Macros like PAGENAME can be used anywhere within a DPL query - not ony in 'titlematch'. The problem is that the expansion happens outside DPL and DPL does not know that there once has been a call of such a macro when it parses its parameter content. Are you talking about the place where "linksto", "linksfrom", "titlematch" etc. conditions are being converted to SQL statements (~ line 1900, ~2300)? A more general approach would be to run the sanitizer before all parameters are being analysed, i.e. at the top of the parameter parsing loop ( ~ line 360 ..). But I am not sure if this would create unwanted side effects. --[[User:Gero|Gero]] 07:49, 16 February 2010 (UTC) | ||
+ | :I think the sanitizer should be run for anything that can be a page title, to ensure that magic words like PAGENAME work. For some arguments (formatting?) you want to keep html codes (although you could still encode them with & amp;) | ||
+ | :I inserted the sanitizer call here to temporarily fix this problem, starting at line 2357: | ||
+ | <pre> | ||
+ | // TitleMatch ... | ||
+ | |||
+ | if ( count($aTitleMatch)>0 ) { | ||
+ | |||
+ | $sSqlWhere .= ' AND ('; | ||
+ | |||
+ | $n=0; | ||
+ | |||
+ | foreach ($aTitleMatch as $link) { | ||
+ | $link = Sanitizer::decodeCharReferences($link); | ||
+ | |||
+ | if ($n>0) $sSqlWhere .= ' OR '; | ||
+ | |||
+ | if ($acceptOpenReferences) { | ||
+ | |||
+ | if ($bIgnoreCase) $sSqlWhere .= 'LOWER(pl_title)' . $sTitleMatchMode . strtolower($dbr->addQuotes($link)) ; | ||
+ | |||
+ | else $sSqlWhere .= 'pl_title' . $sTitleMatchMode . $dbr->addQuotes($link) ; | ||
+ | |||
+ | } else { | ||
+ | |||
+ | if ($bIgnoreCase) $sSqlWhere .= 'LOWER(' . $sPageTable.'.page_title)' . $sTitleMatchMode . strtolower($dbr->addQuotes($link)) ; | ||
+ | |||
+ | else $sSqlWhere .= $sPageTable.'.page_title' . $sTitleMatchMode . $dbr->addQuotes($link) ; | ||
+ | |||
+ | } | ||
+ | |||
+ | $n++; | ||
+ | |||
+ | } | ||
+ | |||
+ | $sSqlWhere .= ')'; | ||
+ | |||
+ | } | ||
+ | </pre> | ||
+ | |||
+ | :And the same for notitlematch. -- [[User:Nx|<span style="color:teal">'''''Nx'''''</span>]] / [[User talk:Nx|''talk'']] 21:20, 16 February 2010 (UTC) |
Latest revision as of 22:20, 16 February 2010
Description: | |
Extension / Version: | DPL / 1.9.0 |
Type / Status: | Bug / open |
Problem
Observe: User:Nx/sand'box
The first query is correctly escaped, the second one, which uses {{PAGENAME}} is not. -- Nx / talk 12:14, 14 February 2010 (UTC)
- Ok, upon further investigation, this isn't really a dpl bug. {{PAGENAME}} et al. return an escaped title: Nx/sand'box -- Nx / talk 13:18, 14 February 2010 (UTC)
And the solution is to run Sanitizer:decodeCharReferences (this is what the Title:newFromText does too), e.g. in the foreach cycle:
$link = Sanitizer::decodeCharReferences($link);
-- Nx / talk 13:37, 14 February 2010 (UTC)
Reply
Macros like PAGENAME can be used anywhere within a DPL query - not ony in 'titlematch'. The problem is that the expansion happens outside DPL and DPL does not know that there once has been a call of such a macro when it parses its parameter content. Are you talking about the place where "linksto", "linksfrom", "titlematch" etc. conditions are being converted to SQL statements (~ line 1900, ~2300)? A more general approach would be to run the sanitizer before all parameters are being analysed, i.e. at the top of the parameter parsing loop ( ~ line 360 ..). But I am not sure if this would create unwanted side effects. --Gero 07:49, 16 February 2010 (UTC)
- I think the sanitizer should be run for anything that can be a page title, to ensure that magic words like PAGENAME work. For some arguments (formatting?) you want to keep html codes (although you could still encode them with & amp;)
- I inserted the sanitizer call here to temporarily fix this problem, starting at line 2357:
// TitleMatch ... if ( count($aTitleMatch)>0 ) { $sSqlWhere .= ' AND ('; $n=0; foreach ($aTitleMatch as $link) { $link = Sanitizer::decodeCharReferences($link); if ($n>0) $sSqlWhere .= ' OR '; if ($acceptOpenReferences) { if ($bIgnoreCase) $sSqlWhere .= 'LOWER(pl_title)' . $sTitleMatchMode . strtolower($dbr->addQuotes($link)) ; else $sSqlWhere .= 'pl_title' . $sTitleMatchMode . $dbr->addQuotes($link) ; } else { if ($bIgnoreCase) $sSqlWhere .= 'LOWER(' . $sPageTable.'.page_title)' . $sTitleMatchMode . strtolower($dbr->addQuotes($link)) ; else $sSqlWhere .= $sPageTable.'.page_title' . $sTitleMatchMode . $dbr->addQuotes($link) ; } $n++; } $sSqlWhere .= ')'; }