Add $wgContribScoreIgnoreMinorEdits, bug fixes

A comprehensive list of changes made in this commit is provided in the
README for this fork.

README

@@ -1,2 +1,28 @@
 The Contribution Scores extension polls the wiki database to locate contributors with the highest contribution volume.
 The extension is intended to add a fun metric for contributors to see how much they are helping out.
+
+--- TEAR HERE ---
+
+This is a modified version of Extension:Contribution Scores, the original source
+code for which can be found here: https://www.mediawiki.org/wiki/Extension:Contribution_Scores
+The rest of this document discussed changes made to this copy of the extension.
+
+This fork developed by Eunakria the config option $wgContribScoreIgnoreMinorEdits,
+which is mutually exclusive with $wgContribScoreUseRoughEditCount.
+
+In addition to providing the aforementioned bug fixes, this fork also fixes two
+other unwanted behaviors exhibited by the original extension:
+
+1. When $wgContribScoreUseRoughEditCount was enabled, the extension would query
+   the user_editcount field of the user table but not actually join on the user
+   table, in the parser function {{#cscore}}. This was presumably never noticed
+   since this parser function is not frequently used.
+
+2. On some databases, there was a potential for integer underflow when
+   subtracting unsigned values in both the parser function and contribution
+   score data. This could occur if $wgContribScoreUseRoughEditCount was enabled
+   and edits went uncounted in user_editcount, and was resolved by clamping the
+   right-hand side preimage of subtraction.
+
+There are probably more bugs I don't know about, both in the original upstream
+code *and* my code! In other words, caveat emptor.

extension.json

@@ -1,7 +1,7 @@
 {
 	"name": "ContributionScores",
-	"author": "Tim Laqua",
+	"author": "Tim Laqua (modified by Eunakria)",
-	"url": "https://www.mediawiki.org/wiki/Extension:Contribution_Scores",
+	"url": "https://git.eunakria.com/Eunakria/mediawiki-extensions-ContributionScores",
 	"descriptionmsg": "contributionscores-desc",
 	"version": "1.26.1",
 	"type": "specialpage",
@@ -53,7 +53,11 @@
 		},
 		"ContribScoreUseRoughEditCount": {
 			"value": false,
-			"description": "Set to true to use the rough number of edits in user table, for performance issue."
+			"description": "Set to true to use the rough number of edits in user table, for performance issue. Mutually exclusive with $ContribScoreIgnoreMinorEdits; this option will be disabled if the other is enabled."
+		},
+		"ContribScoreIgnoreMinorEdits": {
+			"value": false,
+			"description": "Set to true to ignore minor edits. Mutually exclusive with $wgContribScoreUseRoughEditCount; if this is enabled, the other option will be disabled."
 		},
 		"ContribScoreCacheTTL": {
 			"value": 30,

src/ContributionScores.php

@@ -16,6 +16,7 @@ use MediaWiki\User\ActorMigration;
  *
  * @ingroup Extensions
  * @author Tim Laqua <t.laqua@gmail.com>
+ * @author Eunakria <eunakria@gmail.com>
  */
 class ContributionScores extends IncludableSpecialPage {
 	const CONTRIBUTIONSCORES_MAXINCLUDELIMIT = 50;
@@ -28,8 +29,19 @@ class ContributionScores extends IncludableSpecialPage {
 		$parser->setFunctionHook( 'cscore', [ self::class, 'efContributionScoresRender' ] );
 	}
 
+	/**
+	 * See extension.json; this function returns whether or not to use the
+	 * "rough edit count" (user_editcount, stored in the user table). Since
+	 * ignoring minor edits is mutually exclusive with this and takes priority,
+	 * we check here.
+	 */
+	private static function shouldUseRoughEditCount(): bool {
+		global $wgContribScoreIgnoreMinorEdits, $wgContribScoreUseRoughEditCount;
+		return $wgContribScoreUseRoughEditCount && !$wgContribScoreIgnoreMinorEdits;
+	}
+
 	public static function efContributionScoresRender( $parser, $usertext, $metric = 'score' ) {
-		global $wgContribScoreDisableCache, $wgContribScoreUseRoughEditCount;
+		global $wgContribScoreDisableCache, $wgContribScoreIgnoreMinorEdits;
 
 		if ( $wgContribScoreDisableCache ) {
 			$parser->getOutput()->updateCacheExpiry( 0 );
@@ -41,37 +53,81 @@ class ContributionScores extends IncludableSpecialPage {
 
 		if ( $user instanceof User && $user->isRegistered() ) {
 			global $wgLang;
-			$revVar = $wgContribScoreUseRoughEditCount ? 'user_editcount' : 'COUNT(rev_id)';
+			$revVar = self::shouldUseRoughEditCount() ? 'user_editcount' : 'COUNT(rev_id)';
 
 			$revWhere = ActorMigration::newMigration()->getWhere( $dbr, 'rev_user', $user );
+			$revConds = (array)$revWhere['conds'];
+			if ( $wgContribScoreIgnoreMinorEdits ) {
+				$revConds['rev_minor_edit'] = 0;
+			}
+
+			# HACK: The original ContributionScores doesn't actually join on
+			# user despite querying on user... again not our fault, but
+			# definitely our problem. What a mess...
+			$revTables = array_merge([ 'revision' ], $revWhere['tables']);
+			$revJoins = $revWhere['joins'];
+			if ( self::shouldUseRoughEditCount() && !in_array( 'user', $revTables ) ) {
+				# Not familiar enough with MediaWiki to know if this is the
+				# right way to do things, but it seems like it. The
+				# ActorMigration class is deprecated and I think I'm supposed
+				# to just join on the bare names?
+				$revTables[] = 'actor';
+				$revTables[] = 'user';
+				$revJoins['actor'] = [ 'JOIN', 'rev_actor = actor_id' ];
+				$revJoins['user'] = [ 'JOIN', 'actor_user = user_id' ];
+			}
+
+			# HACK: This is not our fault (this bug exists in the original
+			# ContributionScores as well!) but in cases where the rough edit
+			# count underestimates the number of revisions, on MariaDB the
+			# unsigned subtraction can underflow, which will throw an error.
+			# So we have to clamp the subtraction rhs.
+			#
+			# Confused by buildLeast's API? Yeah, me too. It eventually
+			# dispatches to buildSuperlative, which looks like this:
+			# https://doc.wikimedia.org/mediawiki-core/master/php/SQLPlatform_8php_source.html#l00127
+			# The way to smuggle in an existing expression is by passing it as
+			# an array element with a non-integer key to the $fields argument,
+			# which is what we do here.
+			#
+			# Of course the use of buildLeast is necessitated by the fact that
+			# MariaDB and Postgres call it "LEAST", but SQLite calls it "MIN"
+			# (a more intuitive name, but the same as the aggregate function).
+			#
+			$clampedRhsSql = $dbr->buildLeast (
+				/* $fields: */ [ 'lhs' => $revVar, 'rhs' => 'COUNT(DISTINCT rev_page)' ],
+				/* $values: */ []
+			);
+			$wikiRankSql = "COUNT(DISTINCT rev_page) + SQRT($revVar-$clampedRhsSql) * 2";
+
 			if ( $metric == 'score' ) {
 				$row = $dbr->selectRow(
-					[ 'revision' ] + $revWhere['tables'],
+					$revTables,
-					[ 'wiki_rank' => "COUNT(DISTINCT rev_page)+SQRT($revVar-COUNT(DISTINCT rev_page))*2" ],
+					[ 'wiki_rank' => $wikiRankSql ],
-					$revWhere['conds'],
+					$revConds,
 					__METHOD__,
 					[],
-					$revWhere['joins']
+					$revJoins
 				);
 				$output = $wgLang->formatNum( round( $row->wiki_rank, 0 ) );
 			} elseif ( $metric == 'changes' ) {
 				$row = $dbr->selectRow(
-					[ 'revision' ] + $revWhere['tables'],
+					$revTables,
 					[ 'rev_count' => $revVar ],
-					$revWhere['conds'],
+					$revConds,
 					__METHOD__,
 					[],
-					$revWhere['joins']
+					$revJoins
 				);
 				$output = $wgLang->formatNum( $row->rev_count );
 			} elseif ( $metric == 'pages' ) {
 				$row = $dbr->selectRow(
-					[ 'revision' ] + $revWhere['tables'],
+					$revTables,
 					[ 'page_count' => 'COUNT(DISTINCT rev_page)' ],
-					$revWhere['conds'],
+					$revConds,
 					__METHOD__,
 					[],
-					$revWhere['joins']
+					$revJoins
 				);
 				$output = $wgLang->formatNum( $row->page_count );
 			} else {
@@ -92,7 +148,7 @@ class ContributionScores extends IncludableSpecialPage {
 	 */
 	public static function getContributionScoreData( $days, $limit ) {
 		global $wgContribScoreIgnoreBots, $wgContribScoreIgnoreBlockedUsers, $wgContribScoreIgnoreUsernames,
-			$wgContribScoreUseRoughEditCount;
+			$wgContribScoreIgnoreMinorEdits;
 
 		$loadBalancer = MediaWikiServices::getInstance()->getDBLoadBalancer();
 		$dbr = $loadBalancer->getConnection( DB_REPLICA );
@@ -105,6 +161,10 @@ class ContributionScores extends IncludableSpecialPage {
 
 		$sqlWhere = [];
 
+		if ( $wgContribScoreIgnoreMinorEdits ) {
+			$sqlWhere[] = 'rev_minor_edit = 0';
+		}
+
 		if ( $days > 0 ) {
 			$date = time() - ( 60 * 60 * 24 * $days );
 			$sqlWhere[] = 'rev_timestamp > ' . $dbr->addQuotes( $dbr->timestamp( $date ) );
@@ -114,7 +174,7 @@ class ContributionScores extends IncludableSpecialPage {
 			'rev_user'   => $revUser,
 			'page_count' => 'COUNT(DISTINCT rev_page)'
 		];
-		if ( $wgContribScoreUseRoughEditCount ) {
+		if ( self::shouldUseRoughEditCount() ) {
 			$revQuery['tables'][] = 'user';
 			$revQuery['joins']['user'] = [ 'LEFT JOIN', [ "$revUser != 0", "user_id = $revUser" ] ];
 			$sqlVars['rev_count'] = 'user_editcount';
@@ -186,6 +246,14 @@ class ContributionScores extends IncludableSpecialPage {
 			$revQuery['joins']
 		);
 
+		# HACK: See previous comment on the use of buildLeast. Not our fault,
+		# but still our problem...
+		$clampedRhsSql = $dbr->buildLeast(
+			/* $fields: */ [ 'rev_count', 'page_count' ],
+			/* $values: */ []
+		);
+		$wikiRankSql = "page_count+SQRT(rev_count-$clampedRhsSql)*2";
+
 		$sqlMostPagesOrRevs = $dbr->unionQueries( [ $sqlMostPages, $sqlMostRevs ], false );
 		$res = $dbr->select(
 			[
@@ -198,7 +266,7 @@ class ContributionScores extends IncludableSpecialPage {
 				'user_real_name',
 				'page_count',
 				'rev_count',
-				'wiki_rank' => 'page_count+SQRT(rev_count-page_count)*2',
+				'wiki_rank' => $wikiRankSql,
 			],
 			[],
 			__METHOD__,