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         function __construct( $cleanupWikitext = true ) {
  33                 $this->mCleanWikitext = $cleanupWikitext;
  34         }
  35
  36         /**
  37          * Default implementation of wikitext highlighting
  38          *
  39          * @param string $text
  40          * @param array $terms Terms to highlight (unescaped)
  41          * @param int $contextlines
  42          * @param int $contextchars
  43          * @return string
  44          */
  45         public function highlightText( $text, $terms, $contextlines, $contextchars ) {
  46                 global $wgContLang, $wgSearchHighlightBoundaries;
  47
  48                 $fname = __METHOD__;
  49
  50                 if ( $text == '' ) {
  51                         return '';
  52                 }
  53
  54                 // spli text into text + templates/links/tables
  55                 $spat = "/(\\{\\{)|(\\[\\[[^\\]:]+:)|(\n\\{\\|)";
  56                 // first capture group is for detecting nested templates/links/tables/references
  57                 $endPatterns = array(
  58                         1 => '/(\{\{)|(\}\})/', // template
  59                         2 => '/(\[\[)|(\]\])/', // image
  60                         3 => "/(\n\\{\\|)|(\n\\|\\})/" ); // table
  61
  62                 // @todo FIXME: This should prolly be a hook or something
  63                 if ( function_exists( 'wfCite' ) ) {
  64                         $spat .= '|(<ref>)'; // references via cite extension
  65                         $endPatterns[4] = '/(<ref>)|(<\/ref>)/';
  66                 }
  67                 $spat .= '/';
  68                 $textExt = array(); // text extracts
  69                 $otherExt = array(); // other extracts
  70                 wfProfileIn( "$fname-split" );
  71                 $start = 0;
  72                 $textLen = strlen( $text );
  73                 $count = 0; // sequence number to maintain ordering
  74                 while ( $start < $textLen ) {
  75                         // find start of template/image/table
  76                         if ( preg_match( $spat, $text, $matches, PREG_OFFSET_CAPTURE, $start ) ) {
  77                                 $epat = '';
  78                                 foreach ( $matches as $key => $val ) {
  79                                         if ( $key > 0 && $val[1] != - 1 ) {
  80                                                 if ( $key == 2 ) {
  81                                                         // see if this is an image link
  82                                                         $ns = substr( $val[0], 2, - 1 );
  83                                                         if ( $wgContLang->getNsIndex( $ns ) != NS_FILE ) {
  84                                                                 break;
  85                                                         }
  86
  87                                                 }
  88                                                 $epat = $endPatterns[$key];
  89                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, $val[1] - $start ) );
  90                                                 $start = $val[1];
  91                                                 break;
  92                                         }
  93                                 }
  94                                 if ( $epat ) {
  95                                         // find end (and detect any nested elements)
  96                                         $level = 0;
  97                                         $offset = $start + 1;
  98                                         $found = false;
  99                                         while ( preg_match( $epat, $text, $endMatches, PREG_OFFSET_CAPTURE, $offset ) ) {
 100                                                 if ( array_key_exists( 2, $endMatches ) ) {
 101                                                         // found end
 102                                                         if ( $level == 0 ) {
 103                                                                 $len = strlen( $endMatches[2][0] );
 104                                                                 $off = $endMatches[2][1];
 105                                                                 $this->splitAndAdd( $otherExt, $count,
 106                                                                         substr( $text, $start, $off + $len - $start ) );
 107                                                                 $start = $off + $len;
 108                                                                 $found = true;
 109                                                                 break;
 110                                                         } else {
 111                                                                 // end of nested element
 112                                                                 $level -= 1;
 113                                                         }
 114                                                 } else {
 115                                                         // nested
 116                                                         $level += 1;
 117                                                 }
 118                                                 $offset = $endMatches[0][1] + strlen( $endMatches[0][0] );
 119                                         }
 120                                         if ( !$found ) {
 121                                                 // couldn't find appropriate closing tag, skip
 122                                                 $this->splitAndAdd( $textExt, $count, substr( $text, $start, strlen( $matches[0][0] ) ) );
 123                                                 $start += strlen( $matches[0][0] );
 124                                         }
 125                                         continue;
 126                                 }
 127                         }
 128                         // else: add as text extract
 129                         $this->splitAndAdd( $textExt, $count, substr( $text, $start ) );
 130                         break;
 131                 }
 132
 133                 $all = $textExt + $otherExt; // these have disjunct key sets
 134
 135                 wfProfileOut( "$fname-split" );
 136
 137                 // prepare regexps
 138                 foreach ( $terms as $index => $term ) {
 139                         // manually do upper/lowercase stuff for utf-8 since PHP won't do it
 140                         if ( preg_match( '/[\x80-\xff]/', $term ) ) {
 141                                 $terms[$index] = preg_replace_callback(
 142                                         '/./us',
 143                                         array( $this, 'caseCallback' ),
 144                                         $terms[$index]
 145                                 );
 146                         } else {
 147                                 $terms[$index] = $term;
 148                         }
 149                 }
 150                 $anyterm = implode( '|', $terms );
 151                 $phrase = implode( "$wgSearchHighlightBoundaries+", $terms );
 152
 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                 wfProfileIn( "$fname-extract" );
 167
 168                 $left = $contextlines;
 169
 170                 $snippets = array();
 171                 $offsets = array();
 172
 173                 // show beginning only if it contains all words
 174                 $first = 0;
 175                 $firstText = '';
 176                 foreach ( $textExt as $index => $line ) {
 177                         if ( strlen( $line ) > 0 && $line[0] != ';' && $line[0] != ':' ) {
 178                                 $firstText = $this->extract( $line, 0, $contextchars * $contextlines );
 179                                 $first = $index;
 180                                 break;
 181                         }
 182                 }
 183                 if ( $firstText ) {
 184                         $succ = true;
 185                         // check if first text contains all terms
 186                         foreach ( $terms as $term ) {
 187                                 if ( !preg_match( "/$patPre" . $term . "$patPost/ui", $firstText ) ) {
 188                                         $succ = false;
 189                                         break;
 190                                 }
 191                         }
 192                         if ( $succ ) {
 193                                 $snippets[$first] = $firstText;
 194                                 $offsets[$first] = 0;
 195                         }
 196                 }
 197                 if ( !$snippets ) {
 198                         // match whole query on text
 199                         $this->process( $pat1, $textExt, $left, $contextchars, $snippets, $offsets );
 200                         // match whole query on templates/tables/images
 201                         $this->process( $pat1, $otherExt, $left, $contextchars, $snippets, $offsets );
 202                         // match any words on text
 203                         $this->process( $pat2, $textExt, $left, $contextchars, $snippets, $offsets );
 204                         // match any words on templates/tables/images
 205                         $this->process( $pat2, $otherExt, $left, $contextchars, $snippets, $offsets );
 206
 207                         ksort( $snippets );
 208                 }
 209
 210                 // add extra chars to each snippet to make snippets constant size
 211                 $extended = array();
 212                 if ( count( $snippets ) == 0 ) {
 213                         // couldn't find the target words, just show beginning of article
 214                         if ( array_key_exists( $first, $all ) ) {
 215                                 $targetchars = $contextchars * $contextlines;
 216                                 $snippets[$first] = '';
 217                                 $offsets[$first] = 0;
 218                         }
 219                 } else {
 220                         // if begin of the article contains the whole phrase, show only that !!
 221                         if ( array_key_exists( $first, $snippets ) && preg_match( $pat1, $snippets[$first] )
 222                                 && $offsets[$first] < $contextchars * 2 ) {
 223                                 $snippets = array( $first => $snippets[$first] );
 224                         }
 225
 226                         // calc by how much to extend existing snippets
 227                         $targetchars = intval( ( $contextchars * $contextlines ) / count ( $snippets ) );
 228                 }
 229
 230                 foreach ( $snippets as $index => $line ) {
 231                         $extended[$index] = $line;
 232                         $len = strlen( $line );
 233                         if ( $len < $targetchars - 20 ) {
 234                                 // complete this line
 235                                 if ( $len < strlen( $all[$index] ) ) {
 236                                         $extended[$index] = $this->extract(
 237                                                 $all[$index],
 238                                                 $offsets[$index],
 239                                                 $offsets[$index] + $targetchars,
 240                                                 $offsets[$index]
 241                                         );
 242                                         $len = strlen( $extended[$index] );
 243                                 }
 244
 245                                 // add more lines
 246                                 $add = $index + 1;
 247                                 while ( $len < $targetchars - 20
 248                                                 && array_key_exists( $add, $all )
 249                                                 && !array_key_exists( $add, $snippets ) ) {
 250                                         $offsets[$add] = 0;
 251                                         $tt = "\n" . $this->extract( $all[$add], 0, $targetchars - $len, $offsets[$add] );
 252                                         $extended[$add] = $tt;
 253                                         $len += strlen( $tt );
 254                                         $add++;
 255                                 }
 256                         }
 257                 }
 258
 259                 // $snippets = array_map( 'htmlspecialchars', $extended );
 260                 $snippets = $extended;
 261                 $last = - 1;
 262                 $extract = '';
 263                 foreach ( $snippets as $index => $line ) {
 264                         if ( $last == - 1 ) {
 265                                 $extract .= $line; // first line
 266                         } elseif ( $last + 1 == $index
 267                                 && $offsets[$last] + strlen( $snippets[$last] ) >= strlen( $all[$last] )
 268                         ) {
 269                                 $extract .= " " . $line; // continous lines
 270                         } else {
 271                                 $extract .= '<b> ... </b>' . $line;
 272                         }
 273
 274                         $last = $index;
 275                 }
 276                 if ( $extract ) {
 277                         $extract .= '<b> ... </b>';
 278                 }
 279
 280                 $processed = array();
 281                 foreach ( $terms as $term ) {
 282                         if ( !isset( $processed[$term] ) ) {
 283                                 $pat3 = "/$patPre(" . $term . ")$patPost/ui"; // highlight word
 284                                 $extract = preg_replace( $pat3,
 285                                         "\\1<span class='searchmatch'>\\2</span>\\3", $extract );
 286                                 $processed[$term] = true;
 287                         }
 288                 }
 289
 290                 wfProfileOut( "$fname-extract" );
 291
 292                 return $extract;
 293         }
 294
 295         /**
 296          * Split text into lines and add it to extracts array
 297          *
 298          * @param array $extracts Index -> $line
 299          * @param int $count
 300          * @param string $text
 301          */
 302         function splitAndAdd( &$extracts, &$count, $text ) {
 303                 $split = explode( "\n", $this->mCleanWikitext ? $this->removeWiki( $text ) : $text );
 304                 foreach ( $split as $line ) {
 305                         $tt = trim( $line );
 306                         if ( $tt ) {
 307                                 $extracts[$count++] = $tt;
 308                         }
 309                 }
 310         }
 311
 312         /**
 313          * Do manual case conversion for non-ascii chars
 314          *
 315          * @param array $matches
 316          * @return string
 317          */
 318         function caseCallback( $matches ) {
 319                 global $wgContLang;
 320                 if ( strlen( $matches[0] ) > 1 ) {
 321                         return '[' . $wgContLang->lc( $matches[0] ) . $wgContLang->uc( $matches[0] ) . ']';
 322                 } else {
 323                         return $matches[0];
 324                 }
 325         }
 326
 327         /**
 328          * Extract part of the text from start to end, but by
 329          * not chopping up words
 330          * @param string $text
 331          * @param int $start
 332          * @param int $end
 333          * @param int $posStart (out) actual start position
 334          * @param int $posEnd (out) actual end position
 335          * @return string
 336          */
 337         function extract( $text, $start, $end, &$posStart = null, &$posEnd = null ) {
 338                 if ( $start != 0 ) {
 339                         $start = $this->position( $text, $start, 1 );
 340                 }
 341                 if ( $end >= strlen( $text ) ) {
 342                         $end = strlen( $text );
 343                 } else {
 344                         $end = $this->position( $text, $end );
 345                 }
 346
 347                 if ( !is_null( $posStart ) ) {
 348                         $posStart = $start;
 349                 }
 350                 if ( !is_null( $posEnd ) ) {
 351                         $posEnd = $end;
 352                 }
 353
 354                 if ( $end > $start ) {
 355                         return substr( $text, $start, $end - $start );
 356                 } else {
 357                         return '';
 358                 }
 359         }
 360
 361         /**
 362          * Find a nonletter near a point (index) in the text
 363          *
 364          * @param string $text
 365          * @param int $point
 366          * @param int $offset Offset to found index
 367          * @return int Nearest nonletter index, or beginning of utf8 char if none
 368          */
 369         function position( $text, $point, $offset = 0 ) {
 370                 $tolerance = 10;
 371                 $s = max( 0, $point - $tolerance );
 372                 $l = min( strlen( $text ), $point + $tolerance ) - $s;
 373                 $m = array();
 374
 375                 if ( preg_match(
 376                         '/[ ,.!?~!@#$%^&*\(\)+=\-\\\|\[\]"\'<>]/',
 377                         substr( $text, $s, $l ),
 378                         $m,
 379                         PREG_OFFSET_CAPTURE
 380                 ) ) {
 381                         return $m[0][1] + $s + $offset;
 382                 } else {
 383                         // check if point is on a valid first UTF8 char
 384                         $char = ord( $text[$point] );
 385                         while ( $char >= 0x80 && $char < 0xc0 ) {
 386                                 // skip trailing bytes
 387                                 $point++;
 388                                 if ( $point >= strlen( $text ) ) {
 389                                         return strlen( $text );
 390                                 }
 391                                 $char = ord( $text[$point] );
 392                         }
 393
 394                         return $point;
 395
 396                 }
 397         }
 398
 399         /**
 400          * Search extracts for a pattern, and return snippets
 401          *
 402          * @param string $pattern Regexp for matching lines
 403          * @param array $extracts Extracts to search
 404          * @param int $linesleft Number of extracts to make
 405          * @param int $contextchars Length of snippet
 406          * @param array $out Map for highlighted snippets
 407          * @param array $offsets Map of starting points of snippets
 408          * @protected
 409          */
 410         function process( $pattern, $extracts, &$linesleft, &$contextchars, &$out, &$offsets ) {
 411                 if ( $linesleft == 0 ) {
 412                         return; // nothing to do
 413                 }
 414                 foreach ( $extracts as $index => $line ) {
 415                         if ( array_key_exists( $index, $out ) ) {
 416                                 continue; // this line already highlighted
 417                         }
 418
 419                         $m = array();
 420                         if ( !preg_match( $pattern, $line, $m, PREG_OFFSET_CAPTURE ) ) {
 421                                 continue;
 422                         }
 423
 424                         $offset = $m[0][1];
 425                         $len = strlen( $m[0][0] );
 426                         if ( $offset + $len < $contextchars ) {
 427                                 $begin = 0;
 428                         } elseif ( $len > $contextchars ) {
 429                                 $begin = $offset;
 430                         } else {
 431                                 $begin = $offset + intval( ( $len - $contextchars ) / 2 );
 432                         }
 433
 434                         $end = $begin + $contextchars;
 435
 436                         $posBegin = $begin;
 437                         // basic snippet from this line
 438                         $out[$index] = $this->extract( $line, $begin, $end, $posBegin );
 439                         $offsets[$index] = $posBegin;
 440                         $linesleft--;
 441                         if ( $linesleft == 0 ) {
 442                                 return;
 443                         }
 444                 }
 445         }
 446
 447         /**
 448          * Basic wikitext removal
 449          * @protected
 450          * @param string $text
 451          * @return mixed
 452          */
 453         function removeWiki( $text ) {
 454                 $fname = __METHOD__;
 455                 wfProfileIn( $fname );
 456
 457                 // $text = preg_replace( "/'{2,5}/", "", $text );
 458                 // $text = preg_replace( "/\[[a-z]+:\/\/[^ ]+ ([^]]+)\]/", "\\2", $text );
 459                 // $text = preg_replace( "/\[\[([^]|]+)\]\]/", "\\1", $text );
 460                 // $text = preg_replace( "/\[\[([^]]+\|)?([^|]]+)\]\]/", "\\2", $text );
 461                 // $text = preg_replace( "/\\{\\|(.*?)\\|\\}/", "", $text );
 462                 // $text = preg_replace( "/\\[\\[[A-Za-z_-]+:([^|]+?)\\]\\]/", "", $text );
 463                 $text = preg_replace( "/\\{\\{([^|]+?)\\}\\}/", "", $text );
 464                 $text = preg_replace( "/\\{\\{([^|]+\\|)(.*?)\\}\\}/", "\\2", $text );
 465                 $text = preg_replace( "/\\[\\[([^|]+?)\\]\\]/", "\\1", $text );
 466                 $text = preg_replace_callback(
 467                         "/\\[\\[([^|]+\\|)(.*?)\\]\\]/",
 468                         array( $this, 'linkReplace' ),
 469                         $text
 470                 );
 471                 // $text = preg_replace("/\\[\\[([^|]+\\|)(.*?)\\]\\]/", "\\2", $text);
 472                 $text = preg_replace( "/<\/?[^>]+>/", "", $text );
 473                 $text = preg_replace( "/'''''/", "", $text );
 474                 $text = preg_replace( "/('''|<\/?[iIuUbB]>)/", "", $text );
 475                 $text = preg_replace( "/''/", "", $text );
 476
 477                 wfProfileOut( $fname );
 478                 return $text;
 479         }
 480
 481         /**
 482          * callback to replace [[target|caption]] kind of links, if
 483          * the target is category or image, leave it
 484          *
 485          * @param array $matches
 486          */
 487         function linkReplace( $matches ) {
 488                 $colon = strpos( $matches[1], ':' );
 489                 if ( $colon === false ) {
 490                         return $matches[2]; // replace with caption
 491                 }
 492                 global $wgContLang;
 493                 $ns = substr( $matches[1], 0, $colon );
 494                 $index = $wgContLang->getNsIndex( $ns );
 495                 if ( $index !== false && ( $index == NS_FILE || $index == NS_CATEGORY ) ) {
 496                         return $matches[0]; // return the whole thing
 497                 } else {
 498                         return $matches[2];
 499                 }
 500         }
 501
 502         /**
 503          * Simple & fast snippet extraction, but gives completely unrelevant
 504          * snippets
 505          *
 506          * @param string $text
 507          * @param array $terms
 508          * @param int $contextlines
 509          * @param int $contextchars
 510          * @return string
 511          */
 512         public function highlightSimple( $text, $terms, $contextlines, $contextchars ) {
 513                 global $wgContLang;
 514                 $fname = __METHOD__;
 515
 516                 $lines = explode( "\n", $text );
 517
 518                 $terms = implode( '|', $terms );
 519                 $max = intval( $contextchars ) + 1;
 520                 $pat1 = "/(.*)($terms)(.{0,$max})/i";
 521
 522                 $lineno = 0;
 523
 524                 $extract = "";
 525                 wfProfileIn( "$fname-extract" );
 526                 foreach ( $lines as $line ) {
 527                         if ( 0 == $contextlines ) {
 528                                 break;
 529                         }
 530                         ++$lineno;
 531                         $m = array();
 532                         if ( !preg_match( $pat1, $line, $m ) ) {
 533                                 continue;
 534                         }
 535                         --$contextlines;
 536                         // truncate function changes ... to relevant i18n message.
 537                         $pre = $wgContLang->truncate( $m[1], - $contextchars, '...', false );
 538
 539                         if ( count( $m ) < 3 ) {
 540                                 $post = '';
 541                         } else {
 542                                 $post = $wgContLang->truncate( $m[3], $contextchars, '...', false );
 543                         }
 544
 545                         $found = $m[2];
 546
 547                         $line = htmlspecialchars( $pre . $found . $post );
 548                         $pat2 = '/(' . $terms . ")/i";
 549                         $line = preg_replace( $pat2, "<span class='searchmatch'>\\1</span>", $line );
 550
 551                         $extract .= "${line}\n";
 552                 }
 553                 wfProfileOut( "$fname-extract" );
 554
 555                 return $extract;
 556         }
 557
 558         /**
 559          * Returns the first few lines of the text
 560          *
 561          * @param string $text
 562          * @param int $contextlines max number of returned lines
 563          * @param int $contextchars average number of characters per line
 564          * @return string
 565          */
 566         public function highlightNone( $text, $contextlines, $contextchars) {
 567                 $match = array();
 568                 $text = ltrim( $text ) . "\n"; // make sure the preg_match may find the last line
 569                 $text = str_replace( "\n\n", "\n", $text); // remove empty lines
 570                 preg_match( "/^(.*\n){0,$contextlines}/", $text , $match);
 571                 $text = htmlspecialchars( substr( trim( $match[0] ), 0, $contextlines * $contextchars ) ); // trim and limit to max number of chars
 572                 return str_replace( "\n", '<br>', $text );
 573         }
 574 }