array()); $file = $dir . '/x' . sprintf('%02x', $bank) . '.php'; if (is_file($file)) { include($file); } $base = $UTF8_TO_ASCII[$bank]; // For unknown characters, these files have '[?]' in them. Replace with // NULL for compatibility with our data. $base = array_map('_replace_question_with_null', $base); $out[$bank] = $base; } return $out; } /** * Reads in the CPAN Text::Unidecode data set. * * The data is expected to be in files xNN.pm in directory 'Unidecode' under * this file's directory. It can be downloaded from * http://search.cpan.org/~sburke/Text-Unidecode-0.04/lib/Text/Unidecode.pm. * * @return array * Nested array of transliteration data. Outer keys are the first two * bytes of Unicode characters (or 0 for base ASCII characters). The next * level is the other two bytes, and the values are the transliterations. */ function read_cpan_data() { $dir = __DIR__ . '/Unidecode'; $out = array(); // Read data files. for ($bank = 0; $bank < 256; $bank++) { $base = array(); $file = $dir . '/x' . sprintf('%02x', $bank) . '.pm'; if (is_file($file)) { $base = _cpan_read_file($file); } $out[$bank] = $base; } return $out; } /** * Reads in the data in a single file from the Text::Unidecode CPAN project. * * @param string $file * File to read from. * * @return array * Data read from the file. * * @see read_cpan_data() */ function _cpan_read_file($file) { $contents = file($file); $save = ''; foreach ($contents as $line) { // Discard lines starting with # or $. The first line seems to have a // comment starting with #, the second has a Perl line like // $Text::Unidecode::Char[0x04] = [, -- and we do not want either. if (preg_match('|^\s*[#\$]|', $line)) { continue; } // Discard lines ending with semi-colons, which we also don't want // (there seem to be two of these lines at the end of the files). if (preg_match('|;\s*$|', $line)) { continue; } // Replace '[?]' with nothing (that means "don't know how to // transliterate"). In some files, this is encoded as qq{[?]} or // qq{[?] } instead. $line = str_replace('qq{[?]}', 'NULL', $line); $line = str_replace('qq{[?] }', 'NULL', $line); $line = str_replace("'[?]'", 'NULL', $line); // Replace qq{} with either "" or '' or nothing, depending on what is // inside it. $line = str_replace('qq{\{}', "'{'", $line); $line = str_replace('qq{\}}', "'}'", $line); $line = str_replace('qq{\} }', "'} '", $line); $line = str_replace("qq{\\\\}", '"\\\\"', $line); $line = str_replace("qq{\\", "qq{'", $line); $line = str_replace("qq{\"'}", "\"\\\"'\"", $line); $line = preg_replace('|qq\{([^\'\}]+)\}|', "'$1'", $line); $line = preg_replace('|qq\{([^\}]+)\}|', '"$1"', $line); $save .= $line; } // Now we should have a string that looks like: // 'a', 'b', ... // Evaluate as an array. $save = 'return array(' . $save . ');'; $data = @eval($save); if (isset($data) && is_array($data)) { $data = array_map('_replace_hex_with_character', $data); } else { // There was a problem, so throw an error and exit. print "Problem in evaluating $file\n"; print $save; eval($save); exit(); } // For unknown characters, these files may still have '[?]' in them. Replace // with NULL for compatibility with our data. $data = array_map('_replace_question_with_null', $data); return $data; } /** * Reads in the Node.js transliteration data. * * The data is expected to be in files xNN.yml in directory unidecoder_data * under the directory where this file resides. It can be downloaded from * https://github.com/bitwalker/stringex/downloads. You also need the PECL * 'yaml' extension installed for this function to work. * * @return array * Nested array of transliteration data. Outer keys are the first two * bytes of Unicode characters (or 0 for base ASCII characters). The next * level is the other two bytes, and the values are the transliterations. */ function read_nodejs_data() { $dir = __DIR__ . '/unidecoder_data'; $out = array(); // Read data files. for ($bank = 0; $bank < 256; $bank++) { $base = array(); $file = $dir . '/x' . sprintf('%02x', $bank) . '.yml'; if (is_file($file)) { $base = yaml_parse_file($file); // For unknown characters, these files have '[?]' in them. Replace with // NULL for compatibility with our data. $base = array_map('_replace_question_with_null', $base); } $out[$bank] = $base; } return $out; } /** * Loads the PECL 'intl' Transliterator class's transliteration data. * * You need to have the PECL 'intl' package installed for this to work. * * @return array * Nested array of transliteration data. Outer keys are the first two * bytes of Unicode characters (or 0 for base ASCII characters). The next * level is the other two bytes, and the values are the transliterations. */ function read_intl_data() { // In order to transliterate, you first have to create a transliterator // object. This needs a list of transliteration operations. You can get a // list of available operations with: // print_r(Transliterator::listIDs()); exit(); // And a few of these are documented on // http://userguide.icu-project.org/transforms/general and // http://www.unicode.org/reports/tr15/ (for normalizations). // There are also maps to the Unicode characters at: // http://www.unicode.org/roadmaps/bmp/ // http://www.unicode.org/charts/nameslist/ $ops = ''; // The first step in any transform: separate out accents and remove them. $ops .= 'NFD; [:Nonspacing Mark:] Remove; NFC;'; // Then you need to do a bunch of language-specific or script-specific // transliterations. Here is hopefully a representative set. There are // quite a few scripts that don't appear to have rules currently, such // as Etheopian. $ops .= 'Greek-Latin; '; $ops .= 'Cyrillic-Latin; '; $ops .= 'Armenian-Latin; '; $ops .= 'Hebrew-Latin; '; $ops .= 'Arabic-Latin; '; $ops .= 'Syriac-Latin; '; $ops .= 'Thaana-Latin; '; $ops .= 'Devanagari-Latin; '; $ops .= 'Bengali-Latin; '; $ops .= 'Gurmukhi-Latin; '; $ops .= 'Gujarati-Latin; '; $ops .= 'Oriya-Latin; '; $ops .= 'Tamil-Latin; '; $ops .= 'Telugu-Latin; '; $ops .= 'Kannada-Latin; '; $ops .= 'Malayalam-Latin; '; $ops .= 'Thai-Latin; '; $ops .= 'Georgian-Latin; '; $ops .= 'Hangul-Latin; '; $ops .= 'Mongolian-Latin/BGN; '; $ops .= 'Jamo-Latin; '; $ops .= 'Katakana-Latin; '; $ops .= 'Any-Latin; '; // Finally, after transforming to Latin, transform to ASCII. $ops .= 'Latin-ASCII; '; // Remove any remaining accents and recompose. $ops .= 'NFD; [:Nonspacing Mark:] Remove; NFC;'; $trans = Transliterator::create($ops); $out = array(); // Transliterate all possible characters. for ($bank = 0; $bank < 256; $bank++) { $data = array(); for ($chr = 0; $chr < 256; $chr++) { // Skip the UTF-16 and "private use" ranges completely. $OK = ($bank <= 0xd8 || $bank > 0xf8); $result = $OK ? $trans->transliterate(mb_convert_encoding(pack('n', 256 * $bank + $chr), 'UTF-8', 'UTF-16BE')) : ''; // See if we have managed to transliterate this to ASCII or not. If not, // return NULL instead of this character. $max = chr(127); foreach (preg_split('//u', $result, 0, PREG_SPLIT_NO_EMPTY) as $character) { if ($character > $max) { $OK = $OK && FALSE; break; } } $data[$chr] = ($OK) ? $result : NULL; } $out[$bank] = $data; } return $out; } /** * Reads in the JUnidecode data set. * * The data is expected to be in files XNN.java in directory 'junidecode' under * this file's directory. It can be downloaded from * http://www.ippatsuman.com/projects/junidecode/index.html * * @return array * Nested array of transliteration data. Outer keys are the first two * bytes of Unicode characters (or 0 for base ASCII characters). The next * level is the other two bytes, and the values are the transliterations. */ function read_junidecode_data() { $dir = __DIR__ . '/junidecode'; $out = array(); // Read data files. for ($bank = 0; $bank < 256; $bank++) { $base = array(); $file = $dir . '/X' . sprintf('%02x', $bank) . '.java'; if (is_file($file)) { $base = _junidecode_read_file($file); } $out[$bank] = $base; } return $out; } /** * Reads in the data in a single file from the JUnidecode project. * * @param string $file * File to read from. * * @return array * Data read from the file. * * @see read_junidecode_data() */ function _junidecode_read_file($file) { $contents = file($file); $save = ''; foreach ($contents as $line) { // Discard lines starting with * or / or package or class or public or }, // to get rid of comments and Java code. if (preg_match('|^\s*[\*/\}]|', $line)) { continue; } if (preg_match('/^\s*package|public|class/', $line)) { continue; } // Some of the lines look like this: // new String("" + (char) 0x00), // 0x00 // Transform to be '0x00,' $line = preg_replace('|^\s*new\s+String\s*\(\s*""\s*\+\s*\(char\)\s+0x([0-9]+).*$|', '0x$1,', $line); // Strings are in double quotes, yet many have \' in them. $line = str_replace("\'", "'", $line); // Everything else should probably be OK -- the lines are like: // "Ie", // 0x00 $save .= $line; } // Evaluate as an array. $save = 'return array(' . $save . ');'; $data = @eval($save); if (isset($data) && is_array($data)) { $data = array_map('_replace_hex_with_character', $data); $data = array_map('_replace_question_with_null', $data); } else { // There was a problem, so throw an error and exit. print "Problem in evaluating $file\n"; print $save; eval($save); exit(); } return $data; } /** * Callback for array_map(): Returns $data, with '[?]' replaced with NULL. */ function _replace_question_with_null($data) { return ($data == '[?]' || $data == '[?] ') ? NULL : $data; } /** * Callback for array_map(): Replaces '\xNN' with the actual character. */ function _replace_hex_with_character($item) { if (strpos($item, '\x') === 0) { $item = eval($item); } return $item; } /** * Writes a data file out in the standard Drupal Core data format. * * @param array $data * Array of data to write out. * @param string $bank * Bank of characters it belongs to. * @param string $dir * Output directory. */ function write_data_file($data, $bank, $outdir) { $dir = __DIR__ . '/' . $outdir; $file = $dir . '/x' . sprintf('%02x', $bank) . '.php'; $out = ''; $out .= "'; $elems = array_values(array_slice($data, $line, 16)); for ($i = 0; $i < 16; $i++ ) { if (isset($elems[$i])) { $out .= " '" . addcslashes($elems[$i], "'\\") . "',"; } else { $out .= ' NULL,'; } } $out .= "\n"; } $out .= ");\n"; file_put_contents($file, $out); }