This has been done by replacing simple associative arrays with an AbuseFilterVariableHolder, which recognises helper classes called AFComputedVariables.
Computation may occur during the abuse filter analysis, or later when testing and reviewing filters.
* Deprecate parseTokens in favour of a parse-as-you-go approach, faster and uses less memory.
* Display variables in lower_case so they aren't SHOUTING_AT_PEOPLE.
* Tell people if they try to use variables that don't exist, rather than silently returning NULL.
* Write array literals with one item per line. This makes diffs which add or remove items far easier to interpret, and makes merging such changes feasible. And it looks nicer too.
* Use line breaks to show the logical structure of your code. This enhances readability. Bring similar elements in a list into alignment, in order to reveal the differences between those elements at a glance.
* Removed a fun game of spot-the-difference in AbuseFilterHistoryPager::getQueryInfo(). If I want fun games I'll play UFO:AI.
* Moved some oddly placed assignments (in expressions) to their own statements: such assignments reduce readbility.
* Add searching and filtering functionality to the existing 'test' interface.
* Add an 'examine' interface designed for testing filters against a previous change, selectable through the search interface in either the 'test' or the 'examine' view.
* Minor enabling change in ChangesList core, to allow subclassing.
-- Introduce tagging of individual changes (revisions, logs, and on recentchanges). The tags are customisable, and currently settable by the Abuse Filter and by the TorBlock extension. The tags can be styled on the various pages on which they appear.
-- Introduces a schema change, three new tables (valid_tag, change_tag, and tag_summary).
* Remove one dupe call of addExtensionStyle from SpecialAbuseFilter::execute
* Bump $wgAbuseFilterStyleVersion per CSS changes in r46325
* Add a space between drop down and submit button
* Consolidate details on the last edit.
* Allow specification of a warning message to display instead of the standard warning (was already implemented in code, but there was no UI for it.
* Reimplement tools, starting with a spiffy AJAX expression tester. Didn't bother implementing as non-AJAX, as it's just an extra.
* Change exchange protocol between native parser and PHP app. Resulted in 2-3-fold performance boost.
* Fix minor bug from last night.
* Introduce a syntax check on-save.
* Introduce filter history (requires DB schema change).
* Deprecate now non-working /tools interface (will come back).
* Add a bunch of profiling, so we can figure out where performance bottlenecks are.
* Cache the in-object token cache in memcached, bringing per-condition run times from 4ms to a cooler 296us
* Improve normalisation: use AntiSpoof.
* Allow callers to prevent the use of a condition counter. It's kinda annoying in batch processes.
* Improved caching of tokenisation - reduces average time to process a condition from 4ms right down to 200us
* Some new modifiers, caching of modifier data.
* Allow matching on the changes made in an edit (added/removed lines, etc)
* Improve display of variables in details view - standardise the table size, add white-space: pre, font-family: monospace.
* Improve modifiers: add htmlencode/htmldecode/urlencode/urldecode.
* Add 'abuse filter tools', currently consisting of a testing ground for the modifiers. In future, may include facilities for testing filters against user contributions, recent changes, or other sources.
* Disallow more than 100 condition-checks per action.
* Disallow filters which are misbehaving (more than 5 matches, constituting more than 5% of recent actions)
* Display statistics for all filters (percentage of times the condition limit is reached, and percentage of times some filter is matched), and individual filters (percentage of times it's matched).
