maintenance/generateSitemap.php

   1 <?php
   2 define( 'GS_MAIN', -2 );
   3 define( 'GS_TALK', -1 );
   4 /**
   5  * Creates a sitemap for the site
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with this program; if not, write to the Free Software Foundation, Inc.,
  19  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  20  * http://www.gnu.org/copyleft/gpl.html
  21  *
  22  * @ingroup Maintenance
  23  *
  24  * @copyright Copyright © 2005, Ævar Arnfjörð Bjarmason
  25  * @copyright Copyright © 2005, Jens Frank <jeluf@gmx.de>
  26  * @copyright Copyright © 2005, Brion Vibber <brion@pobox.com>
  27  *
  28  * @see http://www.sitemaps.org/
  29  * @see http://www.sitemaps.org/schemas/sitemap/0.9/sitemap.xsd
  30  *
  31  * @license http://www.gnu.org/copyleft/gpl.html GNU General Public License 2.0 or later
  32  */
  33
  34 require_once( dirname(__FILE__) . '/Maintenance.php' );
  35
  36 class GenerateSitemap extends Maintenance {
  37         /**
  38          * The maximum amount of urls in a sitemap file
  39          *
  40          * @link http://www.sitemaps.org/schemas/sitemap/0.9/sitemap.xsd
  41          *
  42          * @var int
  43          */
  44         var $url_limit;
  45
  46         /**
  47          * The maximum size of a sitemap file
  48          *
  49          * @link http://www.sitemaps.org/faq.php#faq_sitemap_size
  50          *
  51          * @var int
  52          */
  53         var $size_limit;
  54
  55         /**
  56          * The path to prepend to the filename
  57          *
  58          * @var string
  59          */
  60         var $fspath;
  61
  62         /**
  63          * The path to append to the domain name
  64          *
  65          * @var string
  66          */
  67         var $path;
  68
  69         /**
  70          * Whether or not to use compression
  71          *
  72          * @var bool
  73          */
  74         var $compress;
  75
  76         /**
  77          * The number of entries to save in each sitemap file
  78          *
  79          * @var array
  80          */
  81         var $limit = array();
  82
  83         /**
  84          * Key => value entries of namespaces and their priorities
  85          *
  86          * @var array
  87          */
  88         var $priorities = array();
  89
  90         /**
  91          * A one-dimensional array of namespaces in the wiki
  92          *
  93          * @var array
  94          */
  95         var $namespaces = array();
  96
  97         /**
  98          * When this sitemap batch was generated
  99          *
 100          * @var string
 101          */
 102         var $timestamp;
 103
 104         /**
 105          * A database slave object
 106          *
 107          * @var object
 108          */
 109         var $dbr;
 110
 111         /**
 112          * A resource pointing to the sitemap index file
 113          *
 114          * @var resource
 115          */
 116         var $findex;
 117
 118
 119         /**
 120          * A resource pointing to a sitemap file
 121          *
 122          * @var resource
 123          */
 124         var $file;
 125
 126         /**
 127          * Constructor
 128          */
 129         public function __construct() {
 130                 parent::__construct();
 131                 $this->mDescription = "Creates a sitemap for the site";
 132                 $this->addOption( 'fspath', 'The file system path to save to, e.g. /tmp/sitemap' .
 133                                                                         "\n\t\tdefaults to current directory", false, true );
 134                 $this->addOption( 'compress', 'Compress the sitemap files, can take value yes|no, default yes', false, true );
 135         }
 136
 137         /**
 138          * Execute
 139          */
 140         public function execute() {
 141                 global $wgScriptPath;
 142                 $this->setNamespacePriorities();
 143                 $this->url_limit = 50000;
 144                 $this->size_limit = pow( 2, 20 ) * 10;
 145                 $this->fspath = self::init_path( $this->getOption( 'fspath', getcwd() ) );
 146                 $this->compress = $this->getOption( 'compress', 'yes' ) !== 'no';
 147                 $this->dbr = wfGetDB( DB_SLAVE );
 148                 $this->generateNamespaces();
 149                 $this->timestamp = wfTimestamp( TS_ISO_8601, wfTimestampNow() );
 150                 $this->findex = fopen( "{$this->fspath}sitemap-index-" . wfWikiID() . ".xml", 'wb' );
 151                 $this->main();
 152         }
 153
 154         private function setNamespacePriorities() {
 155                 // Custom main namespaces
 156                 $this->priorities[GS_MAIN] = '0.5';
 157                 // Custom talk namesspaces
 158                 $this->priorities[GS_TALK] = '0.1';
 159                 // MediaWiki standard namespaces
 160                 $this->priorities[NS_MAIN] = '1.0';
 161                 $this->priorities[NS_TALK] = '0.1';
 162                 $this->priorities[NS_USER] = '0.5';
 163                 $this->priorities[NS_USER_TALK] = '0.1';
 164                 $this->priorities[NS_PROJECT] = '0.5';
 165                 $this->priorities[NS_PROJECT_TALK] = '0.1';
 166                 $this->priorities[NS_FILE] = '0.5';
 167                 $this->priorities[NS_FILE_TALK] = '0.1';
 168                 $this->priorities[NS_MEDIAWIKI] = '0.0';
 169                 $this->priorities[NS_MEDIAWIKI_TALK] = '0.1';
 170                 $this->priorities[NS_TEMPLATE] = '0.0';
 171                 $this->priorities[NS_TEMPLATE_TALK] = '0.1';
 172                 $this->priorities[NS_HELP] = '0.5';
 173                 $this->priorities[NS_HELP_TALK] = '0.1';
 174                 $this->priorities[NS_CATEGORY] = '0.5';
 175                 $this->priorities[NS_CATEGORY_TALK] = '0.1';
 176         }
 177
 178         /**
 179          * Create directory if it does not exist and return pathname with a trailing slash
 180          */
 181         private static function init_path( $fspath ) {
 182                 if( !isset( $fspath ) ) {
 183                         return null;
 184                 }
 185                 # Create directory if needed
 186                 if( $fspath && !is_dir( $fspath ) ) {
 187                         wfMkdirParents( $fspath ) or die("Can not create directory $fspath.\n");
 188                 }
 189
 190                 return realpath( $fspath ). DIRECTORY_SEPARATOR ;
 191         }
 192
 193         /**
 194          * Generate a one-dimensional array of existing namespaces
 195          */
 196         function generateNamespaces() {
 197                 // Only generate for specific namespaces if $wgSitemapNamespaces is an array.
 198                 global $wgSitemapNamespaces;
 199                 if( is_array( $wgSitemapNamespaces ) ) {
 200                         $this->namespaces = $wgSitemapNamespaces;
 201                         return;
 202                 }
 203
 204                 $res = $this->dbr->select( 'page',
 205                         array( 'page_namespace' ),
 206                         array(),
 207                         __METHOD__,
 208                         array(
 209                                 'GROUP BY' => 'page_namespace',
 210                                 'ORDER BY' => 'page_namespace',
 211                         )
 212                 );
 213
 214                 foreach ( $res as $row )
 215                         $this->namespaces[] = $row->page_namespace;
 216         }
 217
 218         /**
 219          * Get the priority of a given namespace
 220          *
 221          * @param int $namespace The namespace to get the priority for
 222          +
 223          * @return string
 224          */
 225
 226         function priority( $namespace ) {
 227                 return isset( $this->priorities[$namespace] ) ? $this->priorities[$namespace] : $this->guessPriority( $namespace );
 228         }
 229
 230         /**
 231          * If the namespace isn't listed on the priority list return the
 232          * default priority for the namespace, varies depending on whether it's
 233          * a talkpage or not.
 234          *
 235          * @param int $namespace The namespace to get the priority for
 236          *
 237          * @return string
 238          */
 239         function guessPriority( $namespace ) {
 240                 return MWNamespace::isMain( $namespace ) ? $this->priorities[GS_MAIN] : $this->priorities[GS_TALK];
 241         }
 242
 243         /**
 244          * Return a database resolution of all the pages in a given namespace
 245          *
 246          * @param int $namespace Limit the query to this namespace
 247          *
 248          * @return resource
 249          */
 250         function getPageRes( $namespace ) {
 251                 return $this->dbr->select( 'page',
 252                         array(
 253                                 'page_namespace',
 254                                 'page_title',
 255                                 'page_touched',
 256                         ),
 257                         array( 'page_namespace' => $namespace ),
 258                         __METHOD__
 259                 );
 260         }
 261
 262         /**
 263          * Main loop
 264          *
 265          * @access public
 266          */
 267         function main() {
 268                 global $wgContLang;
 269
 270                 fwrite( $this->findex, $this->openIndex() );
 271
 272                 foreach ( $this->namespaces as $namespace ) {
 273                         $res = $this->getPageRes( $namespace );
 274                         $this->file = false;
 275                         $this->generateLimit( $namespace );
 276                         $length = $this->limit[0];
 277                         $i = $smcount = 0;
 278
 279                         $fns = $wgContLang->getFormattedNsText( $namespace );
 280                         $this->output( "$namespace ($fns)" );
 281                         foreach ( $res as $row ) {
 282                                 if ( $i++ === 0 || $i === $this->url_limit + 1 || $length + $this->limit[1] + $this->limit[2] > $this->size_limit ) {
 283                                         if ( $this->file !== false ) {
 284                                                 $this->write( $this->file, $this->closeFile() );
 285                                                 $this->close( $this->file );
 286                                         }
 287                                         $filename = $this->sitemapFilename( $namespace, $smcount++ );
 288                                         $this->file = $this->open( $this->fspath . $filename, 'wb' );
 289                                         $this->write( $this->file, $this->openFile() );
 290                                         fwrite( $this->findex, $this->indexEntry( $filename ) );
 291                                         $this->output( "\t$this->fspath$filename\n" );
 292                                         $length = $this->limit[0];
 293                                         $i = 1;
 294                                 }
 295                                 $title = Title::makeTitle( $row->page_namespace, $row->page_title );
 296                                 $date = wfTimestamp( TS_ISO_8601, $row->page_touched );
 297                                 $entry = $this->fileEntry( $title->getFullURL(), $date, $this->priority( $namespace ) );
 298                                 $length += strlen( $entry );
 299                                 $this->write( $this->file, $entry );
 300                                 // generate pages for language variants
 301                                 if($wgContLang->hasVariants()){
 302                                         $variants = $wgContLang->getVariants();
 303                                         foreach($variants as $vCode){
 304                                                 if($vCode==$wgContLang->getCode()) continue; // we don't want default variant
 305                                                 $entry = $this->fileEntry( $title->getFullURL('',$vCode), $date, $this->priority( $namespace ) );
 306                                                 $length += strlen( $entry );
 307                                                 $this->write( $this->file, $entry );
 308                                         }
 309                                 }
 310                         }
 311                         if ( $this->file ) {
 312                                 $this->write( $this->file, $this->closeFile() );
 313                                 $this->close( $this->file );
 314                         }
 315                 }
 316                 fwrite( $this->findex, $this->closeIndex() );
 317                 fclose( $this->findex );
 318         }
 319
 320         /**
 321          * gzopen() / fopen() wrapper
 322          *
 323          * @return resource
 324          */
 325         function open( $file, $flags ) {
 326                 return $this->compress ? gzopen( $file, $flags ) : fopen( $file, $flags );
 327         }
 328
 329         /**
 330          * gzwrite() / fwrite() wrapper
 331          */
 332         function write( &$handle, $str ) {
 333                 if ( $this->compress )
 334                         gzwrite( $handle, $str );
 335                 else
 336                         fwrite( $handle, $str );
 337         }
 338
 339         /**
 340          * gzclose() / fclose() wrapper
 341          */
 342         function close( &$handle ) {
 343                 if ( $this->compress )
 344                         gzclose( $handle );
 345                 else
 346                         fclose( $handle );
 347         }
 348
 349         /**
 350          * Get a sitemap filename
 351          *
 352          * @static
 353          *
 354          * @param int $namespace The namespace
 355          * @param int $count The count
 356          *
 357          * @return string
 358          */
 359         function sitemapFilename( $namespace, $count ) {
 360                 $ext = $this->compress ? '.gz' : '';
 361                 return "sitemap-".wfWikiID()."-NS_$namespace-$count.xml$ext";
 362         }
 363
 364         /**
 365          * Return the XML required to open an XML file
 366          *
 367          * @static
 368          *
 369          * @return string
 370          */
 371         function xmlHead() {
 372                 return '<?xml version="1.0" encoding="UTF-8"?>' . "\n";
 373         }
 374
 375         /**
 376          * Return the XML schema being used
 377          *
 378          * @static
 379          *
 380          * @returns string
 381          */
 382         function xmlSchema() {
 383                 return 'http://www.sitemaps.org/schemas/sitemap/0.9';
 384         }
 385
 386         /**
 387          * Return the XML required to open a sitemap index file
 388          *
 389          * @return string
 390          */
 391         function openIndex() {
 392                 return $this->xmlHead() . '<sitemapindex xmlns="' . $this->xmlSchema() . '">' . "\n";
 393         }
 394
 395         /**
 396          * Return the XML for a single sitemap indexfile entry
 397          *
 398          * @static
 399          *
 400          * @param string $filename The filename of the sitemap file
 401          *
 402          * @return string
 403          */
 404         function indexEntry( $filename ) {
 405                 return
 406                         "\t<sitemap>\n" .
 407                         "\t\t<loc>$filename</loc>\n" .
 408                         "\t\t<lastmod>{$this->timestamp}</lastmod>\n" .
 409                         "\t</sitemap>\n";
 410         }
 411
 412         /**
 413          * Return the XML required to close a sitemap index file
 414          *
 415          * @static
 416          *
 417          * @return string
 418          */
 419         function closeIndex() {
 420                 return "</sitemapindex>\n";
 421         }
 422
 423         /**
 424          * Return the XML required to open a sitemap file
 425          *
 426          * @return string
 427          */
 428         function openFile() {
 429                 return $this->xmlHead() . '<urlset xmlns="' . $this->xmlSchema() . '">' . "\n";
 430         }
 431
 432         /**
 433          * Return the XML for a single sitemap entry
 434          *
 435          * @static
 436          *
 437          * @param string $url An RFC 2396 compliant URL
 438          * @param string $date A ISO 8601 date
 439          * @param string $priority A priority indicator, 0.0 - 1.0 inclusive with a 0.1 stepsize
 440          *
 441          * @return string
 442          */
 443         function fileEntry( $url, $date, $priority ) {
 444                 return
 445                         "\t<url>\n" .
 446                         "\t\t<loc>$url</loc>\n" .
 447                         "\t\t<lastmod>$date</lastmod>\n" .
 448                         "\t\t<priority>$priority</priority>\n" .
 449                         "\t</url>\n";
 450         }
 451
 452         /**
 453          * Return the XML required to close sitemap file
 454          *
 455          * @static
 456          * @return string
 457          */
 458         function closeFile() {
 459                 return "</urlset>\n";
 460         }
 461
 462         /**
 463          * Populate $this->limit
 464          */
 465         function generateLimit( $namespace ) {
 466                 $title = Title::makeTitle( $namespace, str_repeat( "\xf0\xa8\xae\x81", 63 ) . "\xe5\x96\x83" );
 467
 468                 $this->limit = array(
 469                         strlen( $this->openFile() ),
 470                         strlen( $this->fileEntry( $title->getFullUrl(), wfTimestamp( TS_ISO_8601, wfTimestamp() ), $this->priority( $namespace ) ) ),
 471                         strlen( $this->closeFile() )
 472                 );
 473         }
 474 }
 475
 476 $maintClass = "GenerateSitemap";
 477 require_once( DO_MAINTENANCE );