maintenance/generate-entity-file.php

   1 #!/usr/bin/php
   2 <?php
   3
   4 chdir(dirname(__FILE__));
   5 require_once 'common.php';
   6 assertCli();
   7
   8 /**
   9  * @file
  10  * Parses *.ent files into an entity lookup table, and then serializes and
  11  * writes the whole kaboodle to a file. The resulting file is cached so
  12  * that this script does not need to be run. This script should rarely,
  13  * if ever, be run, since HTML's entities are fairly immutable.
  14  */
  15
  16 // here's where the entity files are located, assuming working directory
  17 // is the same as the location of this PHP file. Needs trailing slash.
  18 $entity_dir = '../docs/entities/';
  19
  20 // defines the output file for the serialized content.
  21 $output_file = '../library/HTMLPurifier/EntityLookup/entities.ser';
  22
  23 // courtesy of a PHP manual comment
  24 function unichr($dec) {
  25     if ($dec < 128) {
  26         $utf  = chr($dec);
  27     } else if ($dec < 2048) {
  28         $utf  = chr(192 + (($dec - ($dec % 64)) / 64));
  29         $utf .= chr(128 + ($dec % 64));
  30     } else {
  31         $utf  = chr(224 + (($dec - ($dec % 4096)) / 4096));
  32         $utf .= chr(128 + ((($dec % 4096) - ($dec % 64)) / 64));
  33         $utf .= chr(128 + ($dec % 64));
  34     }
  35     return $utf;
  36 }
  37
  38 if ( !is_dir($entity_dir) ) exit("Fatal Error: Can't find entity directory.\n");
  39 if ( file_exists($output_file) ) exit("Fatal Error: entity-lookup.txt already exists.\n");
  40
  41 $dh = @opendir($entity_dir);
  42 if ( !$dh ) exit("Fatal Error: Cannot read entity directory.\n");
  43
  44 $entity_files = array();
  45 while (($file = readdir($dh)) !== false) {
  46     if (@$file[0] === '.') continue;
  47     if (substr(strrchr($file, "."), 1) !== 'ent') continue;
  48     $entity_files[] = $file;
  49 }
  50 closedir($dh);
  51
  52 if ( !$entity_files ) exit("Fatal Error: No entity files to parse.\n");
  53
  54 $entity_table = array();
  55 $regexp = '/<!ENTITY\s+([A-Za-z]+)\s+"&#(?:38;#)?([0-9]+);">/';
  56
  57 foreach ( $entity_files as $file ) {
  58     $contents = file_get_contents($entity_dir . $file);
  59     $matches = array();
  60     preg_match_all($regexp, $contents, $matches, PREG_SET_ORDER);
  61     foreach ($matches as $match) {
  62         $entity_table[$match[1]] = unichr($match[2]);
  63     }
  64 }
  65
  66 $output = serialize($entity_table);
  67
  68 $fh = fopen($output_file, 'w');
  69 fwrite($fh, $output);
  70 fclose($fh);
  71
  72 echo "Completed successfully.";
  73