5 * Copyright (C) 2004 Brion Vibber <brion@pobox.com>
6 * https://www.mediawiki.org/
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
28 * Search engine hook for MySQL 4+
31 class SearchMySQL
extends SearchDatabase
{
32 protected $strictMatching = true;
34 private static $mMinSearchLength;
37 * Parse the user's query and transform it into an SQL fragment which will
38 * become part of a WHERE clause
40 * @param string $filteredText
41 * @param string $fulltext
45 function parseQuery( $filteredText, $fulltext ) {
48 $lc = $this->legalSearchChars( self
::CHARS_NO_SYNTAX
); // Minus syntax chars (" and *)
50 $this->searchTerms
= [];
52 # @todo FIXME: This doesn't handle parenthetical expressions.
54 if ( preg_match_all( '/([-+<>~]?)(([' . $lc . ']+)(\*?)|"[^"]*")/',
55 $filteredText, $m, PREG_SET_ORDER
) ) {
56 foreach ( $m as $bits ) {
57 MediaWiki\
suppressWarnings();
58 list( /* all */, $modifier, $term, $nonQuoted, $wildcard ) = $bits;
59 MediaWiki\restoreWarnings
();
61 if ( $nonQuoted != '' ) {
65 $term = str_replace( '"', '', $term );
69 if ( $searchon !== '' ) {
72 if ( $this->strictMatching
&& ( $modifier == '' ) ) {
73 // If we leave this out, boolean op defaults to OR which is rarely helpful.
77 // Some languages such as Serbian store the input form in the search index,
78 // so we may need to search for matches in multiple writing system variants.
79 $convertedVariants = $wgContLang->autoConvertToAllVariants( $term );
80 if ( is_array( $convertedVariants ) ) {
81 $variants = array_unique( array_values( $convertedVariants ) );
83 $variants = [ $term ];
86 // The low-level search index does some processing on input to work
87 // around problems with minimum lengths and encoding in MySQL's
89 // For Chinese this also inserts spaces between adjacent Han characters.
90 $strippedVariants = array_map(
91 [ $wgContLang, 'normalizeForSearch' ],
94 // Some languages such as Chinese force all variants to a canonical
95 // form when stripping to the low-level search index, so to be sure
96 // let's check our variants list for unique items after stripping.
97 $strippedVariants = array_unique( $strippedVariants );
99 $searchon .= $modifier;
100 if ( count( $strippedVariants ) > 1 ) {
103 foreach ( $strippedVariants as $stripped ) {
104 $stripped = $this->normalizeText( $stripped );
105 if ( $nonQuoted && strpos( $stripped, ' ' ) !== false ) {
106 // Hack for Chinese: we need to toss in quotes for
107 // multiple-character phrases since normalizeForSearch()
108 // added spaces between them to make word breaks.
109 $stripped = '"' . trim( $stripped ) . '"';
111 $searchon .= "$quote$stripped$quote$wildcard ";
113 if ( count( $strippedVariants ) > 1 ) {
117 // Match individual terms or quoted phrase in result highlighting...
118 // Note that variants will be introduced in a later stage for highlighting!
119 $regexp = $this->regexTerm( $term, $wildcard );
120 $this->searchTerms
[] = $regexp;
122 wfDebug( __METHOD__
. ": Would search with '$searchon'\n" );
123 wfDebug( __METHOD__
. ': Match with /' . implode( '|', $this->searchTerms
) . "/\n" );
125 wfDebug( __METHOD__
. ": Can't understand search query '{$filteredText}'\n" );
128 $searchon = $this->db
->addQuotes( $searchon );
129 $field = $this->getIndexField( $fulltext );
130 return " MATCH($field) AGAINST($searchon IN BOOLEAN MODE) ";
133 function regexTerm( $string, $wildcard ) {
136 $regex = preg_quote( $string, '/' );
137 if ( $wgContLang->hasWordBreaks() ) {
139 // Don't cut off the final bit!
142 $regex = "\b$regex\b";
145 // For Chinese, words may legitimately abut other words in the text literal.
146 // Don't add \b boundary checks... note this could cause false positives
152 public static function legalSearchChars( $type = self
::CHARS_ALL
) {
153 $searchChars = parent
::legalSearchChars( $type );
154 if ( $type === self
::CHARS_ALL
) {
155 // " for phrase, * for wildcard
156 $searchChars = "\"*" . $searchChars;
162 * Perform a full text search query and return a result set.
164 * @param string $term Raw search term
165 * @return SqlSearchResultSet
167 function searchText( $term ) {
168 return $this->searchInternal( $term, true );
172 * Perform a title-only search query and return a result set.
174 * @param string $term Raw search term
175 * @return SqlSearchResultSet
177 function searchTitle( $term ) {
178 return $this->searchInternal( $term, false );
181 protected function searchInternal( $term, $fulltext ) {
182 // This seems out of place, why is this called with empty term?
183 if ( trim( $term ) === '' ) {
187 $filteredTerm = $this->filter( $term );
188 $query = $this->getQuery( $filteredTerm, $fulltext );
189 $resultSet = $this->db
->select(
190 $query['tables'], $query['fields'], $query['conds'],
191 __METHOD__
, $query['options'], $query['joins']
195 $query = $this->getCountQuery( $filteredTerm, $fulltext );
196 $totalResult = $this->db
->select(
197 $query['tables'], $query['fields'], $query['conds'],
198 __METHOD__
, $query['options'], $query['joins']
201 $row = $totalResult->fetchObject();
203 $total = intval( $row->c
);
205 $totalResult->free();
207 return new SqlSearchResultSet( $resultSet, $this->searchTerms
, $total );
210 public function supports( $feature ) {
211 switch ( $feature ) {
212 case 'title-suffix-filter':
215 return parent
::supports( $feature );
220 * Add special conditions
221 * @param array &$query
224 protected function queryFeatures( &$query ) {
225 foreach ( $this->features
as $feature => $value ) {
226 if ( $feature === 'title-suffix-filter' && $value ) {
227 $query['conds'][] = 'page_title' . $this->db
->buildLike( $this->db
->anyString(), $value );
233 * Add namespace conditions
234 * @param array &$query
235 * @since 1.18 (changed)
237 function queryNamespaces( &$query ) {
238 if ( is_array( $this->namespaces
) ) {
239 if ( count( $this->namespaces
) === 0 ) {
240 $this->namespaces
[] = '0';
242 $query['conds']['page_namespace'] = $this->namespaces
;
248 * @param array &$query
251 protected function limitResult( &$query ) {
252 $query['options']['LIMIT'] = $this->limit
;
253 $query['options']['OFFSET'] = $this->offset
;
257 * Construct the SQL query to do the search.
258 * The guts shoulds be constructed in queryMain()
259 * @param string $filteredTerm
260 * @param bool $fulltext
262 * @since 1.18 (changed)
264 function getQuery( $filteredTerm, $fulltext ) {
273 $this->queryMain( $query, $filteredTerm, $fulltext );
274 $this->queryFeatures( $query );
275 $this->queryNamespaces( $query );
276 $this->limitResult( $query );
282 * Picks which field to index on, depending on what type of query.
283 * @param bool $fulltext
286 function getIndexField( $fulltext ) {
287 return $fulltext ?
'si_text' : 'si_title';
291 * Get the base part of the search query.
293 * @param array &$query Search query array
294 * @param string $filteredTerm
295 * @param bool $fulltext
296 * @since 1.18 (changed)
298 function queryMain( &$query, $filteredTerm, $fulltext ) {
299 $match = $this->parseQuery( $filteredTerm, $fulltext );
300 $query['tables'][] = 'page';
301 $query['tables'][] = 'searchindex';
302 $query['fields'][] = 'page_id';
303 $query['fields'][] = 'page_namespace';
304 $query['fields'][] = 'page_title';
305 $query['conds'][] = 'page_id=si_page';
306 $query['conds'][] = $match;
310 * @since 1.18 (changed)
311 * @param string $filteredTerm
312 * @param bool $fulltext
315 function getCountQuery( $filteredTerm, $fulltext ) {
316 $match = $this->parseQuery( $filteredTerm, $fulltext );
319 'tables' => [ 'page', 'searchindex' ],
320 'fields' => [ 'COUNT(*) as c' ],
321 'conds' => [ 'page_id=si_page', $match ],
326 $this->queryFeatures( $query );
327 $this->queryNamespaces( $query );
333 * Create or update the search index record for the given page.
334 * Title and text should be pre-processed.
337 * @param string $title
338 * @param string $text
340 function update( $id, $title, $text ) {
341 $dbw = wfGetDB( DB_MASTER
);
342 $dbw->replace( 'searchindex',
346 'si_title' => $this->normalizeText( $title ),
347 'si_text' => $this->normalizeText( $text )
352 * Update a search index record's title only.
353 * Title should be pre-processed.
356 * @param string $title
358 function updateTitle( $id, $title ) {
359 $dbw = wfGetDB( DB_MASTER
);
361 $dbw->update( 'searchindex',
362 [ 'si_title' => $this->normalizeText( $title ) ],
363 [ 'si_page' => $id ],
365 [ $dbw->lowPriorityOption() ] );
369 * Delete an indexed page
370 * Title should be pre-processed.
372 * @param int $id Page id that was deleted
373 * @param string $title Title of page that was deleted
375 function delete( $id, $title ) {
376 $dbw = wfGetDB( DB_MASTER
);
378 $dbw->delete( 'searchindex', [ 'si_page' => $id ], __METHOD__
);
382 * Converts some characters for MySQL's indexing to grok it correctly,
383 * and pads short words to overcome limitations.
384 * @param string $string
385 * @return mixed|string
387 function normalizeText( $string ) {
390 $out = parent
::normalizeText( $string );
392 // MySQL fulltext index doesn't grok utf-8, so we
393 // need to fold cases and convert to hex
394 $out = preg_replace_callback(
395 "/([\\xc0-\\xff][\\x80-\\xbf]*)/",
396 [ $this, 'stripForSearchCallback' ],
397 $wgContLang->lc( $out ) );
399 // And to add insult to injury, the default indexing
400 // ignores short words... Pad them so we can pass them
401 // through without reconfiguring the server...
402 $minLength = $this->minSearchLength();
403 if ( $minLength > 1 ) {
411 // Periods within things like hostnames and IP addresses
412 // are also important -- we want a search for "example.com"
413 // or "192.168.1.1" to work sanely.
414 // MySQL's search seems to ignore them, so you'd match on
415 // "example.wikipedia.com" and "192.168.83.1" as well.
425 * Armor a case-folded UTF-8 string to get through MySQL's
426 * fulltext search without being mucked up by funny charset
427 * settings or anything else of the sort.
428 * @param array $matches
431 protected function stripForSearchCallback( $matches ) {
432 return 'u8' . bin2hex( $matches[1] );
436 * Check MySQL server's ft_min_word_len setting so we know
437 * if we need to pad short words...
441 protected function minSearchLength() {
442 if ( is_null( self
::$mMinSearchLength ) ) {
443 $sql = "SHOW GLOBAL VARIABLES LIKE 'ft\\_min\\_word\\_len'";
445 $dbr = wfGetDB( DB_REPLICA
);
446 $result = $dbr->query( $sql, __METHOD__
);
447 $row = $result->fetchObject();
450 if ( $row && $row->Variable_name
== 'ft_min_word_len' ) {
451 self
::$mMinSearchLength = intval( $row->Value
);
453 self
::$mMinSearchLength = 0;
456 return self
::$mMinSearchLength;