1f41bbe4cSAndreas Gohr<?php 2f41bbe4cSAndreas Gohr 3f41bbe4cSAndreas Gohrnamespace dokuwiki\Utf8; 4f41bbe4cSAndreas Gohr 5f41bbe4cSAndreas Gohr/** 6f41bbe4cSAndreas Gohr * Methods to convert from and to UTF-8 strings 7f41bbe4cSAndreas Gohr */ 8f41bbe4cSAndreas Gohrclass Conversion 9f41bbe4cSAndreas Gohr{ 10f41bbe4cSAndreas Gohr 11f41bbe4cSAndreas Gohr /** 12f41bbe4cSAndreas Gohr * Encodes UTF-8 characters to HTML entities 13f41bbe4cSAndreas Gohr * 14f41bbe4cSAndreas Gohr * @author Tom N Harris <tnharris@whoopdedo.org> 15f41bbe4cSAndreas Gohr * @author <vpribish at shopping dot com> 16f41bbe4cSAndreas Gohr * @link http://php.net/manual/en/function.utf8-decode.php 17f41bbe4cSAndreas Gohr * 18f41bbe4cSAndreas Gohr * @param string $str 19c0c77cd2SAndreas Gohr * @param bool $all Encode non-utf8 char to HTML as well 20f41bbe4cSAndreas Gohr * @return string 21f41bbe4cSAndreas Gohr */ 22c0c77cd2SAndreas Gohr public static function toHtml($str, $all = false) 23f41bbe4cSAndreas Gohr { 24f41bbe4cSAndreas Gohr $ret = ''; 25f41bbe4cSAndreas Gohr foreach (Unicode::fromUtf8($str) as $cp) { 26c0c77cd2SAndreas Gohr if ($cp < 0x80 && !$all) { 27f41bbe4cSAndreas Gohr $ret .= chr($cp); 28f41bbe4cSAndreas Gohr } elseif ($cp < 0x100) { 29f41bbe4cSAndreas Gohr $ret .= "&#$cp;"; 30f41bbe4cSAndreas Gohr } else { 31f41bbe4cSAndreas Gohr $ret .= '&#x' . dechex($cp) . ';'; 32f41bbe4cSAndreas Gohr } 33f41bbe4cSAndreas Gohr } 34f41bbe4cSAndreas Gohr return $ret; 35f41bbe4cSAndreas Gohr } 36f41bbe4cSAndreas Gohr 37f41bbe4cSAndreas Gohr /** 38f41bbe4cSAndreas Gohr * Decodes HTML entities to UTF-8 characters 39f41bbe4cSAndreas Gohr * 40f41bbe4cSAndreas Gohr * Convert any &#..; entity to a codepoint, 41f41bbe4cSAndreas Gohr * The entities flag defaults to only decoding numeric entities. 42f41bbe4cSAndreas Gohr * Pass HTML_ENTITIES and named entities, including & < etc. 43f41bbe4cSAndreas Gohr * are handled as well. Avoids the problem that would occur if you 44f41bbe4cSAndreas Gohr * had to decode "&#38;&amp;#38;" 45f41bbe4cSAndreas Gohr * 468cbc5ee8SAndreas Gohr * unhtmlspecialchars(\dokuwiki\Utf8\Conversion::fromHtml($s)) -> "&&" 478cbc5ee8SAndreas Gohr * \dokuwiki\Utf8\Conversion::fromHtml(unhtmlspecialchars($s)) -> "&&#38;" 48f41bbe4cSAndreas Gohr * what it should be -> "&&#38;" 49f41bbe4cSAndreas Gohr * 50f41bbe4cSAndreas Gohr * @author Tom N Harris <tnharris@whoopdedo.org> 51f41bbe4cSAndreas Gohr * 52f41bbe4cSAndreas Gohr * @param string $str UTF-8 encoded string 53f41bbe4cSAndreas Gohr * @param boolean $entities decode name entities in addtition to numeric ones 54f41bbe4cSAndreas Gohr * @return string UTF-8 encoded string with numeric (and named) entities replaced. 55f41bbe4cSAndreas Gohr */ 56f41bbe4cSAndreas Gohr public static function fromHtml($str, $entities = false) 57f41bbe4cSAndreas Gohr { 58f41bbe4cSAndreas Gohr if (!$entities) { 59f41bbe4cSAndreas Gohr return preg_replace_callback( 60f41bbe4cSAndreas Gohr '/(&#([Xx])?([0-9A-Za-z]+);)/m', 61*e025be72SAndreas Gohr [self::class, 'decodeNumericEntity'], 62f41bbe4cSAndreas Gohr $str 63f41bbe4cSAndreas Gohr ); 64f41bbe4cSAndreas Gohr } 65f41bbe4cSAndreas Gohr 66f41bbe4cSAndreas Gohr return preg_replace_callback( 67f41bbe4cSAndreas Gohr '/&(#)?([Xx])?([0-9A-Za-z]+);/m', 68*e025be72SAndreas Gohr [self::class, 'decodeAnyEntity'], 69f41bbe4cSAndreas Gohr $str 70f41bbe4cSAndreas Gohr ); 71f41bbe4cSAndreas Gohr } 72f41bbe4cSAndreas Gohr 73f41bbe4cSAndreas Gohr /** 74f41bbe4cSAndreas Gohr * Decodes any HTML entity to it's correct UTF-8 char equivalent 75f41bbe4cSAndreas Gohr * 76f41bbe4cSAndreas Gohr * @param string $ent An entity 77f41bbe4cSAndreas Gohr * @return string 78f41bbe4cSAndreas Gohr */ 79f41bbe4cSAndreas Gohr protected static function decodeAnyEntity($ent) 80f41bbe4cSAndreas Gohr { 81f41bbe4cSAndreas Gohr // create the named entity lookup table 82f41bbe4cSAndreas Gohr static $table = null; 83f41bbe4cSAndreas Gohr if ($table === null) { 84f41bbe4cSAndreas Gohr $table = get_html_translation_table(HTML_ENTITIES); 85f41bbe4cSAndreas Gohr $table = array_flip($table); 86f41bbe4cSAndreas Gohr $table = array_map( 87*e025be72SAndreas Gohr static fn($c) => Unicode::toUtf8([ord($c)]), 88f41bbe4cSAndreas Gohr $table 89f41bbe4cSAndreas Gohr ); 90f41bbe4cSAndreas Gohr } 91f41bbe4cSAndreas Gohr 92f41bbe4cSAndreas Gohr if ($ent[1] === '#') { 93f41bbe4cSAndreas Gohr return self::decodeNumericEntity($ent); 94f41bbe4cSAndreas Gohr } 95f41bbe4cSAndreas Gohr 96f41bbe4cSAndreas Gohr if (array_key_exists($ent[0], $table)) { 97f41bbe4cSAndreas Gohr return $table[$ent[0]]; 98f41bbe4cSAndreas Gohr } 99f41bbe4cSAndreas Gohr 100f41bbe4cSAndreas Gohr return $ent[0]; 101f41bbe4cSAndreas Gohr } 102f41bbe4cSAndreas Gohr 103f41bbe4cSAndreas Gohr /** 104f41bbe4cSAndreas Gohr * Decodes numeric HTML entities to their correct UTF-8 characters 105f41bbe4cSAndreas Gohr * 106f41bbe4cSAndreas Gohr * @param $ent string A numeric entity 107f41bbe4cSAndreas Gohr * @return string|false 108f41bbe4cSAndreas Gohr */ 109f41bbe4cSAndreas Gohr protected static function decodeNumericEntity($ent) 110f41bbe4cSAndreas Gohr { 111f41bbe4cSAndreas Gohr switch ($ent[2]) { 112f41bbe4cSAndreas Gohr case 'X': 113f41bbe4cSAndreas Gohr case 'x': 114f41bbe4cSAndreas Gohr $cp = hexdec($ent[3]); 115f41bbe4cSAndreas Gohr break; 116f41bbe4cSAndreas Gohr default: 117*e025be72SAndreas Gohr $cp = (int) $ent[3]; 118f41bbe4cSAndreas Gohr break; 119f41bbe4cSAndreas Gohr } 120*e025be72SAndreas Gohr return Unicode::toUtf8([$cp]); 121f41bbe4cSAndreas Gohr } 122f41bbe4cSAndreas Gohr 123f41bbe4cSAndreas Gohr /** 124f41bbe4cSAndreas Gohr * UTF-8 to UTF-16BE conversion. 125f41bbe4cSAndreas Gohr * 126f41bbe4cSAndreas Gohr * Maybe really UCS-2 without mb_string due to utf8_to_unicode limits 127f41bbe4cSAndreas Gohr * 128f41bbe4cSAndreas Gohr * @param string $str 129f41bbe4cSAndreas Gohr * @param bool $bom 130f41bbe4cSAndreas Gohr * @return string 131f41bbe4cSAndreas Gohr */ 132f41bbe4cSAndreas Gohr public static function toUtf16be($str, $bom = false) 133f41bbe4cSAndreas Gohr { 134f41bbe4cSAndreas Gohr $out = $bom ? "\xFE\xFF" : ''; 135f41bbe4cSAndreas Gohr if (UTF8_MBSTRING) { 136f41bbe4cSAndreas Gohr return $out . mb_convert_encoding($str, 'UTF-16BE', 'UTF-8'); 137f41bbe4cSAndreas Gohr } 138f41bbe4cSAndreas Gohr 139f41bbe4cSAndreas Gohr $uni = Unicode::fromUtf8($str); 140f41bbe4cSAndreas Gohr foreach ($uni as $cp) { 141f41bbe4cSAndreas Gohr $out .= pack('n', $cp); 142f41bbe4cSAndreas Gohr } 143f41bbe4cSAndreas Gohr return $out; 144f41bbe4cSAndreas Gohr } 145f41bbe4cSAndreas Gohr 146f41bbe4cSAndreas Gohr /** 147f41bbe4cSAndreas Gohr * UTF-8 to UTF-16BE conversion. 148f41bbe4cSAndreas Gohr * 149f41bbe4cSAndreas Gohr * Maybe really UCS-2 without mb_string due to utf8_to_unicode limits 150f41bbe4cSAndreas Gohr * 151f41bbe4cSAndreas Gohr * @param string $str 152f41bbe4cSAndreas Gohr * @return false|string 153f41bbe4cSAndreas Gohr */ 154f41bbe4cSAndreas Gohr public static function fromUtf16be($str) 155f41bbe4cSAndreas Gohr { 156f41bbe4cSAndreas Gohr $uni = unpack('n*', $str); 157f41bbe4cSAndreas Gohr return Unicode::toUtf8($uni); 158f41bbe4cSAndreas Gohr } 159f41bbe4cSAndreas Gohr 160f41bbe4cSAndreas Gohr} 161