* added paramters to cli interface
--page={pagename} only process that page
--log : show some output about the process
--debug : show debug output in cli
* changed pagecount on SpecialPage to use COUNT instead of selecting all rows
Use the unicode character property \pL together with look-behind and
look-ahead assertions rather than '\b' escape sequences (which do not work
with non-ASCII characters).
To be able to use the InternalParseBeforeLinks hook (to be implemented), a
different signature was needed. This entailed subsequent changes to calling
functions.
Prevously, the magic words were removed by hooking onto ParserBeforeTidy.
Since the Parser class provides its own mechanism for removing magic words
that are surrounded by double underscores, we now use this (undocumented)
hook.
The anonymous callback functions that were handed over to
preg_replace_callback called private static methods of the LinkTitles class,
which worked fine with PHP 5.5.9, but not with PHP 5.3.3.
The functions were not turned (back) into static methods, and the callback
is given in the array form, as described in the PHP documentation.
WikiPage::doEditContent was extremely slow, because it updated link tables
etc. For batch processing, we now use WikiPage::doQuickEditContent, which is
considerably faster.
The checks for whether a page is a potential link target or not (depending
on the absence of the __NOAUTOLINKTARGET__ magic word and if it is not a
redirect to the current page) have now been moved into the callback
functions, so that they are only performed if a page really is a candidate
for linking (i.e, its title occurs on the currently edited page).
The change resulted in a ~10-fold increase in speed.
Added two new configuration variables, $wgLinkTitlesCheckRedirect and
$wgLinkTitlesEnableNoTargetMagicWord, that can be used to increase
performance. If both are overridden to be false, the target page content
will not be fetched from the database to check for redirect or the
occurrence of a magic word, which saves a lot of time.
Added code to check if a target page redirects to the current page, and
prevents linking to that target page. Note that this does not catch
situations where a page title exists twice, once with first capital letter
and once with first lowercase letter (should happen on test systems only).
The preg_replace_callback callback functions were removed as static
functions and declared inline; the callback function for $wgCapitalLinks ==
false was simplified to speed up the process.
