includes/search/SearchHighlighter.php

   1 <?php
   2 /**
   3  * Basic search engine highlighting
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Search
  22  */
  23
  24 /**
  25  * Highlight bits of wikitext
  26  *
  27  * @ingroup Search
  28  */
  29 class SearchHighlighter {
  30         protected $mCleanWikitext = true;
  31
  32         /**
  33          * @warning If you pass false to this constructor, then
  34          *  the caller is responsible for HTML escaping.
  35          */
  36         function __construct( $cleanupWikitext = true ) {
  37                 $this->mCleanWikitext = $cleanupWikitext;
  38         }
  39
  40         /**
  41          * Wikitext highlighting when $wgAdvancedSearchHighlighting = true
  42          *
  43          * @param string $text
  44          * @param array $terms Terms to highlight (not html escaped but
  45          *   regex escaped via SearchDatabase::regexTerm())
  46          * @param int $contextlines
  47          * @param int $contextchars
  48          * @return string
  49          */
  50         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
  51                 global $wgContLang, $wgSearchHighlightBoundaries;
  52
  53                 if ( $text == '' ) {
  54                         return '';
  55                 }
  56
  57                 // spli text into text + templates/links/tables
  58                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
  59                 // first capture group is for detecting nested templates/links/tables/references
  60                 $endPatterns = [
  61                         1 => '/(\{\{)|(\}\})/', // template
  62                         2 => '/(\[\[)|(\]\])/', // image
  63                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ]; // table
  64
  65                 // @todo FIXME: This should prolly be a hook or something
  66                 // instead of hardcoding a class name from the Cite extension
  67                 if ( class_exists( 'Cite' ) ) {
  68                         $spat .= '|(<ref>)'; // references via cite extension
  69                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
  70                 }
  71                 $spat .= '/';
  72                 $textExt = []; // text extracts
  73                 $otherExt = []; // other extracts
  74                 $start = 0;
  75                 $textLen = strlen( $text );
  76                 $count = 0; // sequence number to maintain ordering
  77                 while ( $start < $textLen ) {
  78                         // find start of template/image/table
  79                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
  80                                 $epat = '';
  81                                 foreach ( $matches as $key => $val ) {
  82                                         if ( $key > 0 && $val[1] != -1 ) {
  83                                                 if ( $key == 2 ) {
  84                                                         // see if this is an image link
  85                                                         $ns = substr( $val[0], 2, -1 );
  86                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
  87                                                                 break;
  88                                                         }
  89
  90                                                 }
  91                                                 $epat = $endPatterns[$key];
  92                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
  93                                                 $start = $val[1];
  94                                                 break;
  95                                         }
  96                                 }
  97                                 if ( $epat ) {
  98                                         // find end (and detect any nested elements)
  99                                         $level = 0;
 100                                         $offset = $start + 1;
 101                                         $found = false;
 102                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 103                                                 if ( array_key_exists( 2, $endMatches ) ) {
 104                                                         // found end
 105                                                         if ( $level == 0 ) {
 106                                                                 $len = strlen( $endMatches[2][0] );
 107                                                                 $off = $endMatches[2][1];
 108                                                                 $this->splitAndAdd( $otherExt, $count,
 109                                                                         substr( $text, $start, $off + $len - $start ) );
 110                                                                 $start = $off + $len;
 111                                                                 $found = true;
 112                                                                 break;
 113                                                         } else {
 114                                                                 // end of nested element
 115                                                                 $level -= 1;
 116                                                         }
 117                                                 } else {
 118                                                         // nested
 119                                                         $level += 1;
 120                                                 }
 121                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
 122                                         }
 123                                         if ( !$found ) {
 124                                                 // couldn't find appropriate closing tag, skip
 125                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
 126                                                 $start += strlen( $matches[0][0] );
 127                                         }
 128                                         continue;
 129                                 }
 130                         }
 131                         // else: add as text extract
 132                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
 133                         break;
 134                 }
 135
 136                 $all = $textExt + $otherExt; // these have disjunct key sets
 137
 138                 // prepare regexps
 139                 foreach ( $terms as $index => $term ) {
 140                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 141                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
 142                                 $terms[$index] = preg_replace_callback(
 143                                         '/./us',
 144                                         [ $this, 'caseCallback' ],
 145                                         $terms[$index]
 146                                 );
 147                         } else {
 148                                 $terms[$index] = $term;
 149                         }
 150                 }
 151                 $anyterm = implode( '|', $terms );
 152                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
 153                 // @todo FIXME: A hack to scale contextchars, a correct solution
 154                 // would be to have contextchars actually be char and not byte
 155                 // length, and do proper utf-8 substrings and lengths everywhere,
 156                 // but PHP is making that very hard and unclean to implement :(
 157                 $scale = strlen( $anyterm ) / mb_strlen( $anyterm );
 158                 $contextchars = intval( $contextchars * $scale );
 159
 160                 $patPre = "(^|$wgSearchHighlightBoundaries)";
 161                 $patPost = "($wgSearchHighlightBoundaries|$)";
 162
 163                 $pat1 = "/(" . $phrase . ")/ui";
 164                 $pat2 = "/$patPre(" . $anyterm . ")$patPost/ui";
 165
 166                 $left = $contextlines;
 167
 168                 $snippets = [];
 169                 $offsets = [];
 170
 171                 // show beginning only if it contains all words
 172                 $first = 0;
 173                 $firstText = '';
 174                 foreach ( $textExt as $index => $line ) {
 175                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
 176                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
 177                                 $first = $index;
 178                                 break;
 179                         }
 180                 }
 181                 if ( $firstText ) {
 182                         $succ = true;
 183                         // check if first text contains all terms
 184                         foreach ( $terms as $term ) {
 185                                 if ( !preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
 186                                         $succ = false;
 187                                         break;
 188                                 }
 189                         }
 190                         if ( $succ ) {
 191                                 $snippets[$first] = $firstText;
 192                                 $offsets[$first] = 0;
 193                         }
 194                 }
 195                 if ( !$snippets ) {
 196                         // match whole query on text
 197                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
 198                         // match whole query on templates/tables/images
 199                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
 200                         // match any words on text
 201                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
 202                         // match any words on templates/tables/images
 203                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
 204
 205                         ksort( $snippets );
 206                 }
 207
 208                 // add extra chars to each snippet to make snippets constant size
 209                 $extended = [];
 210                 if ( count( $snippets ) == 0 ) {
 211                         // couldn't find the target words, just show beginning of article
 212                         if ( array_key_exists( $first, $all ) ) {
 213                                 $targetchars = $contextchars * $contextlines;
 214                                 $snippets[$first] = '';
 215                                 $offsets[$first] = 0;
 216                         }
 217                 } else {
 218                         // if begin of the article contains the whole phrase, show only that !!
 219                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
 220                                 && $offsets[$first] < $contextchars * 2 ) {
 221                                 $snippets = [ $first => $snippets[$first] ];
 222                         }
 223
 224                         // calc by how much to extend existing snippets
 225                         $targetchars = intval( ( $contextchars * $contextlines ) / count( $snippets ) );
 226                 }
 227
 228                 foreach ( $snippets as $index => $line ) {
 229                         $extended[$index] = $line;
 230                         $len = strlen( $line );
 231                         if ( $len < $targetchars - 20 ) {
 232                                 // complete this line
 233                                 if ( $len < strlen( $all[$index] ) ) {
 234                                         $extended[$index] = $this->extract(
 235                                                 $all[$index],
 236                                                 $offsets[$index],
 237                                                 $offsets[$index] + $targetchars,
 238                                                 $offsets[$index]
 239                                         );
 240                                         $len = strlen( $extended[$index] );
 241                                 }
 242
 243                                 // add more lines
 244                                 $add = $index + 1;
 245                                 while ( $len < $targetchars - 20
 246                                                 && array_key_exists( $add, $all )
 247                                                 && !array_key_exists( $add, $snippets ) ) {
 248                                         $offsets[$add] = 0;
 249                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
 250                                         $extended[$add] = $tt;
 251                                         $len += strlen( $tt );
 252                                         $add++;
 253                                 }
 254                         }
 255                 }
 256
 257                 // $snippets = array_map( 'htmlspecialchars', $extended );
 258                 $snippets = $extended;
 259                 $last = -1;
 260                 $extract = '';
 261                 foreach ( $snippets as $index => $line ) {
 262                         if ( $last == -1 ) {
 263                                 $extract .= $line; // first line
 264                         } elseif ( $last + 1 == $index
 265                                 && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] )
 266                         ) {
 267                                 $extract .= " " . $line; // continous lines
 268                         } else {
 269                                 $extract .= '<b> ... </b>' . $line;
 270                         }
 271
 272                         $last = $index;
 273                 }
 274                 if ( $extract ) {
 275                         $extract .= '<b> ... </b>';
 276                 }
 277
 278                 $processed = [];
 279                 foreach ( $terms as $term ) {
 280                         if ( !isset( $processed[$term] ) ) {
 281                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
 282                                 $extract = preg_replace( $pat3,
 283                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
 284                                 $processed[$term] = true;
 285                         }
 286                 }
 287
 288                 return $extract;
 289         }
 290
 291         /**
 292          * Split text into lines and add it to extracts array
 293          *
 294          * @param array $extracts Index -> $line
 295          * @param int $count
 296          * @param string $text
 297          */
 298         function splitAndAdd( &$extracts, &$count, $text ) {
 299                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
 300                 foreach ( $split as $line ) {
 301                         $tt = trim( $line );
 302                         if ( $tt ) {
 303                                 $extracts[$count++] = $tt;
 304                         }
 305                 }
 306         }
 307
 308         /**
 309          * Do manual case conversion for non-ascii chars
 310          *
 311          * @param array $matches
 312          * @return string
 313          */
 314         function caseCallback( $matches ) {
 315                 global $wgContLang;
 316                 if ( strlen( $matches[0] ) > 1 ) {
 317                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
 318                 } else {
 319                         return $matches[0];
 320                 }
 321         }
 322
 323         /**
 324          * Extract part of the text from start to end, but by
 325          * not chopping up words
 326          * @param string $text
 327          * @param int $start
 328          * @param int $end
 329          * @param int $posStart (out) actual start position
 330          * @param int $posEnd (out) actual end position
 331          * @return string
 332          */
 333         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
 334                 if ( $start != 0 ) {
 335                         $start = $this->position( $text, $start, 1 );
 336                 }
 337                 if ( $end >= strlen( $text ) ) {
 338                         $end = strlen( $text );
 339                 } else {
 340                         $end = $this->position( $text, $end );
 341                 }
 342
 343                 if ( !is_null( $posStart ) ) {
 344                         $posStart = $start;
 345                 }
 346                 if ( !is_null( $posEnd ) ) {
 347                         $posEnd = $end;
 348                 }
 349
 350                 if ( $end > $start ) {
 351                         return substr( $text, $start, $end - $start );
 352                 } else {
 353                         return '';
 354                 }
 355         }
 356
 357         /**
 358          * Find a nonletter near a point (index) in the text
 359          *
 360          * @param string $text
 361          * @param int $point
 362          * @param int $offset Offset to found index
 363          * @return int Nearest nonletter index, or beginning of utf8 char if none
 364          */
 365         function position( $text, $point, $offset = 0 ) {
 366                 $tolerance = 10;
 367                 $s = max( 0, $point - $tolerance );
 368                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
 369                 $m = [];
 370
 371                 if ( preg_match(
 372                         '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/',
 373                         substr( $text, $s, $l ),
 374                         $m,
 375                         PREG_OFFSET_CAPTURE
 376                 ) ) {
 377                         return $m[0][1] + $s + $offset;
 378                 } else {
 379                         // check if point is on a valid first UTF8 char
 380                         $char = ord( $text[$point] );
 381                         while ( $char >= 0x80 && $char < 0xc0 ) {
 382                                 // skip trailing bytes
 383                                 $point++;
 384                                 if ( $point >= strlen( $text ) ) {
 385                                         return strlen( $text );
 386                                 }
 387                                 $char = ord( $text[$point] );
 388                         }
 389
 390                         return $point;
 391
 392                 }
 393         }
 394
 395         /**
 396          * Search extracts for a pattern, and return snippets
 397          *
 398          * @param string $pattern Regexp for matching lines
 399          * @param array $extracts Extracts to search
 400          * @param int $linesleft Number of extracts to make
 401          * @param int $contextchars Length of snippet
 402          * @param array $out Map for highlighted snippets
 403          * @param array $offsets Map of starting points of snippets
 404          * @protected
 405          */
 406         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
 407                 if ( $linesleft == 0 ) {
 408                         return; // nothing to do
 409                 }
 410                 foreach ( $extracts as $index => $line ) {
 411                         if ( array_key_exists( $index, $out ) ) {
 412                                 continue; // this line already highlighted
 413                         }
 414
 415                         $m = [];
 416                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
 417                                 continue;
 418                         }
 419
 420                         $offset = $m[0][1];
 421                         $len = strlen( $m[0][0] );
 422                         if ( $offset + $len < $contextchars ) {
 423                                 $begin = 0;
 424                         } elseif ( $len > $contextchars ) {
 425                                 $begin = $offset;
 426                         } else {
 427                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
 428                         }
 429
 430                         $end = $begin + $contextchars;
 431
 432                         $posBegin = $begin;
 433                         // basic snippet from this line
 434                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
 435                         $offsets[$index] = $posBegin;
 436                         $linesleft--;
 437                         if ( $linesleft == 0 ) {
 438                                 return;
 439                         }
 440                 }
 441         }
 442
 443         /**
 444          * Basic wikitext removal
 445          * @protected
 446          * @param string $text
 447          * @return mixed
 448          */
 449         function removeWiki( $text ) {
 450                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
 451                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
 452                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
 453                 $text = preg_replace_callback(
 454                         "/\\[\\[([^|]+\\|)(.*?)\\]\\]/",
 455                         [ $this, 'linkReplace' ],
 456                         $text
 457                 );
 458                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
 459                 $text = preg_replace( "/'''''/", "", $text );
 460                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
 461                 $text = preg_replace( "/''/", "", $text );
 462
 463                 // Note, the previous /<\/?[^>]+>/ is insufficient
 464                 // for XSS safety as the HTML tag can span multiple
 465                 // search results (T144845).
 466                 $text = Sanitizer::escapeHtmlAllowEntities( $text );
 467                 return $text;
 468         }
 469
 470         /**
 471          * callback to replace [[target|caption]] kind of links, if
 472          * the target is category or image, leave it
 473          *
 474          * @param array $matches
 475          * @return string
 476          */
 477         function linkReplace( $matches ) {
 478                 $colon = strpos( $matches[1], ':' );
 479                 if ( $colon === false ) {
 480                         return $matches[2]; // replace with caption
 481                 }
 482                 global $wgContLang;
 483                 $ns = substr( $matches[1], 0, $colon );
 484                 $index = $wgContLang->getNsIndex( $ns );
 485                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
 486                         return $matches[0]; // return the whole thing
 487                 } else {
 488                         return $matches[2];
 489                 }
 490         }
 491
 492         /**
 493          * Simple & fast snippet extraction, but gives completely unrelevant
 494          * snippets
 495          *
 496          * Used when $wgAdvancedSearchHighlighting is false.
 497          *
 498          * @param string $text
 499          * @param array $terms Escaped for regex by SearchDatabase::regexTerm()
 500          * @param int $contextlines
 501          * @param int $contextchars
 502          * @return string
 503          */
 504         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
 505                 global $wgContLang;
 506
 507                 $lines = explode( "\n", $text );
 508
 509                 $terms = implode( '|', $terms );
 510                 $max = intval( $contextchars ) + 1;
 511                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
 512
 513                 $lineno = 0;
 514
 515                 $extract = "";
 516                 foreach ( $lines as $line ) {
 517                         if ( 0 == $contextlines ) {
 518                                 break;
 519                         }
 520                         ++$lineno;
 521                         $m = [];
 522                         if ( !preg_match( $pat1, $line, $m ) ) {
 523                                 continue;
 524                         }
 525                         --$contextlines;
 526                         // truncate function changes ... to relevant i18n message.
 527                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
 528
 529                         if ( count( $m ) < 3 ) {
 530                                 $post = '';
 531                         } else {
 532                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
 533                         }
 534
 535                         $found = $m[2];
 536
 537                         $line = htmlspecialchars( $pre . $found . $post );
 538                         $pat2 = '/(' . $terms . ")/i";
 539                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
 540
 541                         $extract .= "${line}\n";
 542                 }
 543
 544                 return $extract;
 545         }
 546
 547         /**
 548          * Returns the first few lines of the text
 549          *
 550          * @param string $text
 551          * @param int $contextlines Max number of returned lines
 552          * @param int $contextchars Average number of characters per line
 553          * @return string
 554          */
 555         public function highlightNone( $text, $contextlines, $contextchars ) {
 556                 $match = [];
 557                 $text = ltrim( $text ) . "\n"; // make sure the preg_match may find the last line
 558                 $text = str_replace( "\n\n", "\n", $text ); // remove empty lines
 559                 preg_match( "/^(.*\n){0,$contextlines}/", $text, $match );
 560
 561                 // Trim and limit to max number of chars
 562                 $text = htmlspecialchars( substr( trim( $match[0] ), 0, $contextlines * $contextchars ) );
 563                 return str_replace( "\n", '<br>', $text );
 564         }
 565 }