MediaWiki  master
SearchMySQL.php
Go to the documentation of this file.
1 <?php
28 
33 class SearchMySQL extends SearchDatabase {
34  protected $strictMatching = true;
35 
36  private static $mMinSearchLength;
37 
47  private function parseQuery( $filteredText, $fulltext ) {
48  $lc = $this->legalSearchChars( self::CHARS_NO_SYNTAX ); // Minus syntax chars (" and *)
49  $searchon = '';
50  $this->searchTerms = [];
51 
52  # @todo FIXME: This doesn't handle parenthetical expressions.
53  $m = [];
54  if ( preg_match_all( '/([-+<>~]?)(([' . $lc . ']+)(\*?)|"[^"]*")/',
55  $filteredText, $m, PREG_SET_ORDER ) ) {
56  foreach ( $m as $bits ) {
57  Wikimedia\suppressWarnings();
58  list( /* all */, $modifier, $term, $nonQuoted, $wildcard ) = $bits;
59  Wikimedia\restoreWarnings();
60 
61  if ( $nonQuoted != '' ) {
62  $term = $nonQuoted;
63  $quote = '';
64  } else {
65  $term = str_replace( '"', '', $term );
66  $quote = '"';
67  }
68 
69  if ( $searchon !== '' ) {
70  $searchon .= ' ';
71  }
72  if ( $this->strictMatching && ( $modifier == '' ) ) {
73  // If we leave this out, boolean op defaults to OR which is rarely helpful.
74  $modifier = '+';
75  }
76 
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  $contLang = MediaWikiServices::getInstance()->getContentLanguage();
80  $convertedVariants = $contLang->autoConvertToAllVariants( $term );
81  if ( is_array( $convertedVariants ) ) {
82  $variants = array_unique( array_values( $convertedVariants ) );
83  } else {
84  $variants = [ $term ];
85  }
86 
87  // The low-level search index does some processing on input to work
88  // around problems with minimum lengths and encoding in MySQL's
89  // fulltext engine.
90  // For Chinese this also inserts spaces between adjacent Han characters.
91  $strippedVariants = array_map( [ $contLang, 'normalizeForSearch' ], $variants );
92 
93  // Some languages such as Chinese force all variants to a canonical
94  // form when stripping to the low-level search index, so to be sure
95  // let's check our variants list for unique items after stripping.
96  $strippedVariants = array_unique( $strippedVariants );
97 
98  $searchon .= $modifier;
99  if ( count( $strippedVariants ) > 1 ) {
100  $searchon .= '(';
101  }
102  foreach ( $strippedVariants as $stripped ) {
103  $stripped = $this->normalizeText( $stripped );
104  if ( $nonQuoted && strpos( $stripped, ' ' ) !== false ) {
105  // Hack for Chinese: we need to toss in quotes for
106  // multiple-character phrases since normalizeForSearch()
107  // added spaces between them to make word breaks.
108  $stripped = '"' . trim( $stripped ) . '"';
109  }
110  $searchon .= "$quote$stripped$quote$wildcard ";
111  }
112  if ( count( $strippedVariants ) > 1 ) {
113  $searchon .= ')';
114  }
115 
116  // Match individual terms or quoted phrase in result highlighting...
117  // Note that variants will be introduced in a later stage for highlighting!
118  $regexp = $this->regexTerm( $term, $wildcard );
119  $this->searchTerms[] = $regexp;
120  }
121  wfDebug( __METHOD__ . ": Would search with '$searchon'\n" );
122  wfDebug( __METHOD__ . ': Match with /' . implode( '|', $this->searchTerms ) . "/\n" );
123  } else {
124  wfDebug( __METHOD__ . ": Can't understand search query '{$filteredText}'\n" );
125  }
126 
127  $dbr = $this->lb->getConnectionRef( DB_REPLICA );
128  $searchon = $dbr->addQuotes( $searchon );
129  $field = $this->getIndexField( $fulltext );
130  return [
131  " MATCH($field) AGAINST($searchon IN BOOLEAN MODE) ",
132  " MATCH($field) AGAINST($searchon IN NATURAL LANGUAGE MODE) DESC "
133  ];
134  }
135 
136  private function regexTerm( $string, $wildcard ) {
137  $regex = preg_quote( $string, '/' );
138  if ( MediaWikiServices::getInstance()->getContentLanguage()->hasWordBreaks() ) {
139  if ( $wildcard ) {
140  // Don't cut off the final bit!
141  $regex = "\b$regex";
142  } else {
143  $regex = "\b$regex\b";
144  }
145  } else {
146  // For Chinese, words may legitimately abut other words in the text literal.
147  // Don't add \b boundary checks... note this could cause false positives
148  // for Latin chars.
149  }
150  return $regex;
151  }
152 
153  public function legalSearchChars( $type = self::CHARS_ALL ) {
154  $searchChars = parent::legalSearchChars( $type );
155  if ( $type === self::CHARS_ALL ) {
156  // " for phrase, * for wildcard
157  $searchChars = "\"*" . $searchChars;
158  }
159  return $searchChars;
160  }
161 
168  protected function doSearchTextInDB( $term ) {
169  return $this->searchInternal( $term, true );
170  }
171 
178  protected function doSearchTitleInDB( $term ) {
179  return $this->searchInternal( $term, false );
180  }
181 
182  protected function searchInternal( $term, $fulltext ) {
183  // This seems out of place, why is this called with empty term?
184  if ( trim( $term ) === '' ) {
185  return null;
186  }
187 
188  $filteredTerm = $this->filter( $term );
189  $query = $this->getQuery( $filteredTerm, $fulltext );
190  $dbr = $this->lb->getConnectionRef( DB_REPLICA );
191  $resultSet = $dbr->select(
192  $query['tables'], $query['fields'], $query['conds'],
193  __METHOD__, $query['options'], $query['joins']
194  );
195 
196  $total = null;
197  $query = $this->getCountQuery( $filteredTerm, $fulltext );
198  $totalResult = $dbr->select(
199  $query['tables'], $query['fields'], $query['conds'],
200  __METHOD__, $query['options'], $query['joins']
201  );
202 
203  $row = $totalResult->fetchObject();
204  if ( $row ) {
205  $total = intval( $row->c );
206  }
207  $totalResult->free();
208 
209  return new SqlSearchResultSet( $resultSet, $this->searchTerms, $total );
210  }
211 
212  public function supports( $feature ) {
213  switch ( $feature ) {
214  case 'title-suffix-filter':
215  return true;
216  default:
217  return parent::supports( $feature );
218  }
219  }
220 
226  protected function queryFeatures( &$query ) {
227  foreach ( $this->features as $feature => $value ) {
228  if ( $feature === 'title-suffix-filter' && $value ) {
229  $dbr = $this->lb->getConnectionRef( DB_REPLICA );
230  $query['conds'][] = 'page_title' . $dbr->buildLike( $dbr->anyString(), $value );
231  }
232  }
233  }
234 
240  function queryNamespaces( &$query ) {
241  if ( is_array( $this->namespaces ) ) {
242  if ( count( $this->namespaces ) === 0 ) {
243  $this->namespaces[] = '0';
244  }
245  $query['conds']['page_namespace'] = $this->namespaces;
246  }
247  }
248 
254  protected function limitResult( &$query ) {
255  $query['options']['LIMIT'] = $this->limit;
256  $query['options']['OFFSET'] = $this->offset;
257  }
258 
267  private function getQuery( $filteredTerm, $fulltext ) {
268  $query = [
269  'tables' => [],
270  'fields' => [],
271  'conds' => [],
272  'options' => [],
273  'joins' => [],
274  ];
275 
276  $this->queryMain( $query, $filteredTerm, $fulltext );
277  $this->queryFeatures( $query );
278  $this->queryNamespaces( $query );
279  $this->limitResult( $query );
280 
281  return $query;
282  }
283 
289  private function getIndexField( $fulltext ) {
290  return $fulltext ? 'si_text' : 'si_title';
291  }
292 
301  private function queryMain( &$query, $filteredTerm, $fulltext ) {
302  $match = $this->parseQuery( $filteredTerm, $fulltext );
303  $query['tables'][] = 'page';
304  $query['tables'][] = 'searchindex';
305  $query['fields'][] = 'page_id';
306  $query['fields'][] = 'page_namespace';
307  $query['fields'][] = 'page_title';
308  $query['conds'][] = 'page_id=si_page';
309  $query['conds'][] = $match[0];
310  $query['options']['ORDER BY'] = $match[1];
311  }
312 
319  private function getCountQuery( $filteredTerm, $fulltext ) {
320  $match = $this->parseQuery( $filteredTerm, $fulltext );
321 
322  $query = [
323  'tables' => [ 'page', 'searchindex' ],
324  'fields' => [ 'COUNT(*) as c' ],
325  'conds' => [ 'page_id=si_page', $match[0] ],
326  'options' => [],
327  'joins' => [],
328  ];
329 
330  $this->queryFeatures( $query );
331  $this->queryNamespaces( $query );
332 
333  return $query;
334  }
335 
344  function update( $id, $title, $text ) {
345  $dbw = $this->lb->getConnectionRef( DB_MASTER );
346  $dbw->replace( 'searchindex',
347  [ 'si_page' ],
348  [
349  'si_page' => $id,
350  'si_title' => $this->normalizeText( $title ),
351  'si_text' => $this->normalizeText( $text )
352  ], __METHOD__ );
353  }
354 
362  function updateTitle( $id, $title ) {
363  $dbw = $this->lb->getConnectionRef( DB_MASTER );
364  $dbw->update( 'searchindex',
365  [ 'si_title' => $this->normalizeText( $title ) ],
366  [ 'si_page' => $id ],
367  __METHOD__
368  );
369  }
370 
378  function delete( $id, $title ) {
379  $dbw = $this->lb->getConnectionRef( DB_MASTER );
380  $dbw->delete( 'searchindex', [ 'si_page' => $id ], __METHOD__ );
381  }
382 
389  function normalizeText( $string ) {
390  $out = parent::normalizeText( $string );
391 
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  MediaWikiServices::getInstance()->getContentLanguage()->lc( $out ) );
398 
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 ) {
404  $n = $minLength - 1;
405  $out = preg_replace(
406  "/\b(\w{1,$n})\b/",
407  "$1u800",
408  $out );
409  }
410 
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.
416  $out = preg_replace(
417  "/(\w)\.(\w|\*)/u",
418  "$1u82e$2",
419  $out );
420 
421  return $out;
422  }
423 
431  protected function stripForSearchCallback( $matches ) {
432  return 'u8' . bin2hex( $matches[1] );
433  }
434 
441  protected function minSearchLength() {
442  if ( is_null( self::$mMinSearchLength ) ) {
443  $sql = "SHOW GLOBAL VARIABLES LIKE 'ft\\_min\\_word\\_len'";
444 
445  $dbr = $this->lb->getConnectionRef( DB_REPLICA );
446  $result = $dbr->query( $sql, __METHOD__ );
447  $row = $result->fetchObject();
448  $result->free();
449 
450  if ( $row && $row->Variable_name == 'ft_min_word_len' ) {
451  self::$mMinSearchLength = intval( $row->Value );
452  } else {
453  self::$mMinSearchLength = 0;
454  }
455  }
456  return self::$mMinSearchLength;
457  }
458 }
deferred txt A few of the database updates required by various functions here can be deferred until after the result page is displayed to the user For updating the view updating the linked to tables after a etc PHP does not yet have any way to tell the server to actually return and disconnect while still running these but it might have such a feature in the future We handle these by creating a deferred update object and putting those objects on a global list
Definition: deferred.txt:11
return true to allow those checks to and false if checking is done remove or add to the links of a group of changes in EnhancedChangesList Hook subscribers can return false to omit this line from recentchanges use this to change the tables headers change it to an object instance and return false override the list derivative used $groups Array of ChangesListFilterGroup objects(added in 1.34) 'FileDeleteComplete' null for the local wiki Added should default to null in handler for backwards compatibility add a value to it if you want to add a cookie that have to vary cache options can modify $query
Definition: hooks.txt:1535
whereas SearchGetNearMatch runs after $term
Definition: hooks.txt:2874
to move a page</td >< td > &*You are moving the page across namespaces
minSearchLength()
Check MySQL server&#39;s ft_min_word_len setting so we know if we need to pad short words...
Apache License January AND DISTRIBUTION Definitions License shall mean the terms and conditions for use
limitResult(&$query)
Add limit options.
queryMain(&$query, $filteredTerm, $fulltext)
Get the base part of the search query.
$value
getIndexField( $fulltext)
Picks which field to index on, depending on what type of query.
injection txt This is an overview of how MediaWiki makes use of dependency injection The design described here grew from the discussion of RFC T384 The term dependency this means that anything an object needs to operate should be injected from the the object itself should only know narrow no concrete implementation of the logic it relies on The requirement to inject everything typically results in an architecture that based on two main types of and essentially stateless service objects that use other service objects to operate on the value objects As of the beginning MediaWiki is only starting to use the DI approach Much of the code still relies on global state or direct resulting in a highly cyclical dependency MediaWikiServices
Definition: injection.txt:23
normalizeText( $string)
Converts some characters for MySQL&#39;s indexing to grok it correctly, and pads short words to overcome ...
const DB_MASTER
Definition: defines.php:26
The index of the header message $result[1]=The index of the body text message $result[2 through n]=Parameters passed to body text message. Please note the header message cannot receive/use parameters. 'ImgAuthModifyHeaders':Executed just before a file is streamed to a user via img_auth.php, allowing headers to be modified beforehand. $title:LinkTarget object & $headers:HTTP headers(name=> value, names are case insensitive). Two headers get special handling:If-Modified-Since(value must be a valid HTTP date) and Range(must be of the form "bytes=(\*-\*)") will be honored when streaming the file. 'ImportHandleLogItemXMLTag':When parsing a XML tag in a log item. Return false to stop further processing of the tag $reader:XMLReader object $logInfo:Array of information 'ImportHandlePageXMLTag':When parsing a XML tag in a page. Return false to stop further processing of the tag $reader:XMLReader object & $pageInfo:Array of information 'ImportHandleRevisionXMLTag':When parsing a XML tag in a page revision. Return false to stop further processing of the tag $reader:XMLReader object $pageInfo:Array of page information $revisionInfo:Array of revision information 'ImportHandleToplevelXMLTag':When parsing a top level XML tag. Return false to stop further processing of the tag $reader:XMLReader object 'ImportHandleUnknownUser':When a user doesn 't exist locally, this hook is called to give extensions an opportunity to auto-create it. If the auto-creation is successful, return false. $name:User name 'ImportHandleUploadXMLTag':When parsing a XML tag in a file upload. Return false to stop further processing of the tag $reader:XMLReader object $revisionInfo:Array of information 'ImportLogInterwikiLink':Hook to change the interwiki link used in log entries and edit summaries for transwiki imports. & $fullInterwikiPrefix:Interwiki prefix, may contain colons. & $pageTitle:String that contains page title. 'ImportSources':Called when reading from the $wgImportSources configuration variable. Can be used to lazy-load the import sources list. & $importSources:The value of $wgImportSources. Modify as necessary. See the comment in DefaultSettings.php for the detail of how to structure this array. 'InfoAction':When building information to display on the action=info page. $context:IContextSource object & $pageInfo:Array of information 'InitializeArticleMaybeRedirect':MediaWiki check to see if title is a redirect. & $title:Title object for the current page & $request:WebRequest & $ignoreRedirect:boolean to skip redirect check & $target:Title/string of redirect target & $article:Article object 'InternalParseBeforeLinks':during Parser 's internalParse method before links but after nowiki/noinclude/includeonly/onlyinclude and other processings. & $parser:Parser object & $text:string containing partially parsed text & $stripState:Parser 's internal StripState object 'InternalParseBeforeSanitize':during Parser 's internalParse method just before the parser removes unwanted/dangerous HTML tags and after nowiki/noinclude/includeonly/onlyinclude and other processings. Ideal for syntax-extensions after template/parser function execution which respect nowiki and HTML-comments. & $parser:Parser object & $text:string containing partially parsed text & $stripState:Parser 's internal StripState object 'InterwikiLoadPrefix':When resolving if a given prefix is an interwiki or not. Return true without providing an interwiki to continue interwiki search. $prefix:interwiki prefix we are looking for. & $iwData:output array describing the interwiki with keys iw_url, iw_local, iw_trans and optionally iw_api and iw_wikiid. 'InvalidateEmailComplete':Called after a user 's email has been invalidated successfully. $user:user(object) whose email is being invalidated 'IRCLineURL':When constructing the URL to use in an IRC notification. Callee may modify $url and $query, URL will be constructed as $url . $query & $url:URL to index.php & $query:Query string $rc:RecentChange object that triggered url generation 'IsFileCacheable':Override the result of Article::isFileCacheable()(if true) & $article:article(object) being checked 'IsTrustedProxy':Override the result of IP::isTrustedProxy() & $ip:IP being check & $result:Change this value to override the result of IP::isTrustedProxy() 'IsUploadAllowedFromUrl':Override the result of UploadFromUrl::isAllowedUrl() $url:URL used to upload from & $allowed:Boolean indicating if uploading is allowed for given URL 'isValidEmailAddr':Override the result of Sanitizer::validateEmail(), for instance to return false if the domain name doesn 't match your organization. $addr:The e-mail address entered by the user & $result:Set this and return false to override the internal checks 'isValidPassword':Override the result of User::isValidPassword() $password:The password entered by the user & $result:Set this and return false to override the internal checks $user:User the password is being validated for 'Language::getMessagesFileName':$code:The language code or the language we 're looking for a messages file for & $file:The messages file path, you can override this to change the location. 'LanguageGetNamespaces':Provide custom ordering for namespaces or remove namespaces. Do not use this hook to add namespaces. Use CanonicalNamespaces for that. & $namespaces:Array of namespaces indexed by their numbers 'LanguageGetTranslatedLanguageNames':Provide translated language names. & $names:array of language code=> language name $code:language of the preferred translations 'LanguageLinks':Manipulate a page 's language links. This is called in various places to allow extensions to define the effective language links for a page. $title:The page 's Title. & $links:Array with elements of the form "language:title" in the order that they will be output. & $linkFlags:Associative array mapping prefixed links to arrays of flags. Currently unused, but planned to provide support for marking individual language links in the UI, e.g. for featured articles. 'LanguageSelector':Hook to change the language selector available on a page. $out:The output page. $cssClassName:CSS class name of the language selector. 'LinkBegin':DEPRECATED since 1.28! Use HtmlPageLinkRendererBegin instead. Used when generating internal and interwiki links in Linker::link(), before processing starts. Return false to skip default processing and return $ret. See documentation for Linker::link() for details on the expected meanings of parameters. $skin:the Skin object $target:the Title that the link is pointing to & $html:the contents that the< a > tag should have(raw HTML) $result
Definition: hooks.txt:1981
updateTitle( $id, $title)
Update a search index record&#39;s title only.
this hook is for auditing only or null if authentication failed before getting that far or null if we can t even determine that When $user is not it can be in the form of< username >< more info > e g for bot passwords intended to be added to log contexts Fields it might only if the login was with a bot password it is not rendered in wiki pages or galleries in category pages allow injecting custom HTML after the section Any uses of the hook need to handle escaping see BaseTemplate::getToolbox and BaseTemplate::makeListItem for details on the format of individual items inside of this array or by returning and letting standard HTTP rendering take place modifiable or by returning false and taking over the output $out
Definition: hooks.txt:773
doSearchTitleInDB( $term)
Perform a title-only search query and return a result set.
regexTerm( $string, $wildcard)
Search engine hook for MySQL 4+.
Definition: SearchMySQL.php:33
queryFeatures(&$query)
Add special conditions.
wfDebug( $text, $dest='all', array $context=[])
Sends a line to the debug log if enabled or, optionally, to a comment in output.
getQuery( $filteredTerm, $fulltext)
Construct the SQL query to do the search.
supports( $feature)
This class is used for different SQL-based search engines shipped with MediaWiki. ...
stripForSearchCallback( $matches)
Armor a case-folded UTF-8 string to get through MySQL&#39;s fulltext search without being mucked up by fu...
this hook is for auditing only or null if authentication failed before getting that far or null if we can t even determine that When $user is not null
Definition: hooks.txt:773
getCountQuery( $filteredTerm, $fulltext)
namespace and then decline to actually register it file or subcat img or subcat $title
Definition: hooks.txt:918
This document is intended to provide useful advice for parties seeking to redistribute MediaWiki to end users It s targeted particularly at maintainers for Linux since it s been observed that distribution packages of MediaWiki often break We ve consistently had to recommend that users seeking support use official tarballs instead of their distribution s and this often solves whatever problem the user is having It would be nice if this could such as
Definition: distributors.txt:9
static $mMinSearchLength
Definition: SearchMySQL.php:36
parseQuery( $filteredText, $fulltext)
Parse the user&#39;s query and transform it into two SQL fragments: a WHERE condition and an ORDER BY exp...
Definition: SearchMySQL.php:47
injection txt This is an overview of how MediaWiki makes use of dependency injection The design described here grew from the discussion of RFC T384 The term dependency this means that anything an object needs to operate should be injected from the the object itself should only know narrow no concrete implementation of the logic it relies on The requirement to inject everything typically results in an architecture that based on two main types of and essentially stateless service objects that use other service objects to operate on the value objects As of the beginning MediaWiki is only starting to use the DI approach Much of the code still relies on global state or direct resulting in a highly cyclical dependency which acts as the top level factory for services in MediaWiki which can be used to gain access to default instances of various services MediaWikiServices however also allows new services to be defined and default services to be redefined Services are defined or redefined by providing a callback the instantiator that will return a new instance of the service When it will create an instance of MediaWikiServices and populate it with the services defined in the files listed by thereby bootstrapping the DI framework Per $wgServiceWiringFiles lists includes ServiceWiring php
Definition: injection.txt:35
int [] null $namespaces
filter( $text)
Return a &#39;cleaned up&#39; search string.
legalSearchChars( $type=self::CHARS_ALL)
update( $id, $title, $text)
Create or update the search index record for the given page.
Base search engine base class for database-backed searches.
searchInternal( $term, $fulltext)
doSearchTextInDB( $term)
Perform a full text search query and return a result set.
const DB_REPLICA
Definition: defines.php:25
queryNamespaces(&$query)
Add namespace conditions.
$matches