The 'math' table will no longer be created on a default install unless you've explicitly enabled the Math plugin at install time; the usual update.php procedure will add it in.
Postgres, Oracle, MSSQL, and DB2 variants are included -- broken out from the core files -- but have not been tested.
I know there has been some code duplication in parser test infrastructure but could only find one instance of the parser test temporary table setup to remove the 'math' table from (the extension adds it back via the hook). If the phpunit-based runner breaks, please track it down and fix it there too.
* Remove cl_type from paging in categorylinks - it's not
really needed there. Although cl_type is in WHERE but not
in ORDER BY clause the worst thing that can happen
is to have a filesort going again through <500 entries
selected by index. Or will FORCE INDEX work anyway?
* Revert schema change, as we don't need cl_type there
anyway (or even if we wanted to compare, it should
work as expected by using INT values against ENUM).
This commit does not include a patch for SQLite, because ALTER TABLE MODIFY is apparently not supported by SQLite as far as I could tell by Googling. Leaving resolution of this issue for SQLite to the SQLite experts; maybe SQLite's enum implementation is saner than MySQL and it doesn't even need this schema change, I don't know.
* Added a maintenance script which generates a list of first letters. Unified Han are omitted for performance, and because they shouldn't be used as headings anyway. A future collation specific to Chinese would provide the KangXi radicals as "first letters".
* Provided a precomputed list of first letters. Used Unicode 6.0.0 data and ICU 4.2.
* Moved collation functionality from Language to a Collation class hierarchy with factory function. Removed the recently-added methods from Language and updated all callers.
* Changed Title::getCategorySortkey() to separate its parts with a line break instead of a null character. All collations supported by the intl extension ignore the null character, i.e. "ab" == "a\0b". It would have required a lot of hacking to make it work.
* Fixed the uppercase collation to handle non-ASCII characters, redundantly with r80436. I don't think it's necessary to change the collation name as was done there, so I reverted that in the course of my conflict merge. A --force option to updateCollation.php might be nice though.
Add patch to kill bad original ar_page_revid index (did exist as 2 definitions under the same name). Add patch to add index ar_revid per FIXME, though, hopefully only a temporary index
Updated tables.sql with new index
Bug 25503, reported by Dmitriy. When switching it to binary, I figured
that since it would always be one byte per character now, there would be
room to make it the full 255 bytes. In fact, if cl_to is utf8, the
(cl_to, cl_type, cl_sortkey, cl_from) index will be 255*3 + 1 + 255 + 4
= 1025 bytes, longer than the max of 1000. Shortening cl_sortkey to 230
bytes fixes this (fix tested by Dmitriy).
I didn't add an updater because if you're already running the current
schema without problems, you don't need this change.
Per review by Tim, I made two changes:
1) Fix cl_sortkey to be varbinary(255).
2) Expand cl_collation to varbinary(32), and change $wgCollationVersion
to $wgCategoryCollation, to account for the variety of collations we
might have. tinyint is too small. I could have gone with int, but
that's annoyingly inscrutable in practice, as we all know from namespace
fields.
To make the upgrade easier for non-trunk users, I updated the old patch
file to incorporate the new changes, using the updatelog table so that
people upgrading from 1.16 won't have to do two alters on categorylinks.
I didn't test the upgrade-from-1.16 code path yet, so if anyone tests
that and it seems not to break, commenting to that effect would be
appreciated.
Also removed wfDeprecated() from archive(). Do *not* add this to
functions that are still actively used in core. If you think this
function is so terrible that it really mustn't be used, remove callers
yourself, don't pester every single developer with messages in the hope
that someone else will do it for you.
* When a revision cannot be found in the revision table, look for it in the archive table.
* If both deleted and undeleted revisions are covered by the same log entry, show two links (see bug 23663)
Like langlinks, this stores the interwiki prefix (as iwl_prefix) and full page title (as iwl_title), attached to the page doing the liking (as iwl_from -> page_id).
Unlike langlinks, there can be multiple entries stored per interwiki prefix.
Updater to add the table confirmed on MySQL, untested on SQLite but should work.
Someone may still need to add and test a PostgreSQL updater.
Refactored makeWhereFrom2d() out of LinkBatch to Database so it could be re-used for the similar mapping for the interwiki links, which need a string prefix rather than an int namespace key.
Also cleaned it up internally to reuse existing code for building where clauses from arrays. (Tim & Domas -- if the previous more verbose code was there to reduce function call and array processing overhead on very large link lists, feel free to unroll it again if the difference is measurable. Just swap the var names around from the old LinkBatch code and escape the base key value if it's not an integer, it'll be functionally equivalent.)
This should not affect any existing behavior. (Except that it reorders
some error conditions in attemptAutoCreate(), but probably no one cares
about that.) It adds a new database table, but it will be unused unless
you enable external authentication.
An outline of the rationale for this system, and the design planning, is
at <http://www.mediawiki.org/wiki/ExternalAuth>. Essentially,
AuthPlugin puts too much of a burden on plugin authors, requiring them
to write a lot of policy logic instead of just handling the actual
interface to the external user database. This system uses a standard
framework to decide policy questions, and auth plugins only need to
provide some low-level, clearly-specified data.
There are lots of features still missing, marked in the code, but basic
functionality is present. The commit includes initial support for one
type of external authentication, the forum software vBulletin (which I
happen to know well, and want to integrate with my MediaWiki).
I'm encouraging the inclusion of ExternalAuth plugins in core because in
this framework, the amount of code required to add an additional backend
is quite small -- well under 100 lines in this case. I'd hope to see a
lot more of these, and it seems unreasonable to make an armada of tiny
extensions instead of letting them live happily in their own directory
out of everyone's way.
* The serialized message cache, which would have been redundant, has been removed. Similar performance characteristics can be achieved with $wgLocalisationCacheConf['manualRecache'] = true;
* Added a maintenance script rebuildLocalisationCache.php for offline rebuilding of the localisation cache.
* Extension i18n files can now contain any of the variables which can be set in Messages*.php. It is possible, and recommended, to use this feature instead of the hooks for special page aliases and magic words.
* $wgExtensionAliasesFiles, LanguageGetMagic and LanguageGetSpecialPageAliases are retained for backwards compatibility. $wgMessageCache->addMessages() and related functions have been removed. wfLoadExtensionMessages() is a no-op and can continue to be called for b/c.
* Introduced $wgCacheDirectory as a default location for the various local caches that have accumulated. Suggested $IP/cache as a good place for it in the default LocalSettings.php and created this directory with a deny-all .htaccess.
* Patched Exception.php to avoid using the message cache when an exception is thrown from within LocalisationCache, since this tends to fail horribly.
* Removed Language::getLocalisationArray(), Language::loadLocalisation(), Language::load()
* Fixed FileDependency::__sleep()
* In Cdb.php, fixed newlines in debug messages
In MessageCache::get():
* Replaced calls to $wgContLang capitalisation functions with plain PHP functions, reducing the typical case from 99us to 93us. Message cache keys are already documented as being restricted to ASCII.
* Implemented a more efficient way to filter out bogus language codes, reducing the "foo/en" case from 430us to 101us
* Optimised wfRunHooks() in the typical do-nothing case, from ~30us to ~3us. This reduced MessageCache::get() typical case time from 93us to 38us.
* Removed hook MessageNotInMwNs to save an extra 3us per cache hit. Reimplemented the only user (LocalisationUpdate) using the new hook LocalisationCacheRecache.
* Include DB table options on log_search table creation
* Adjust tabs->spaces in table def... sounds silly, but the tabs sometimes confuse things when cut-and-pasting into mysql command line. :)
