How to replace/remove 4(+)-byte characters from a UTF-8 string in PHP?

Question

It seems like MySQL does not support characters with more than 3 bytes in its default UTF-8 charset.

So, in PHP, how can I get rid of all 4(-and-more)-byte characters in a string and replace them with something like by some other character?

Answer 1

NOTE: you should not just strip, but replace with replacement character U+FFFD to avoid unicode attacks, mostly XSS:

http://unicode.org/reports/tr36/#Deletion_of_Noncharacters

preg_replace('/[\x{10000}-\x{10FFFF}]/u', "\xEF\xBF\xBD", $value);

Answer 2

Since 4-byte UTF-8 sequences always start with the bytes 0xF0-0xF7 , the following should work:

$str = preg_replace('/[\xF0-\xF7].../s', '', $str);

Alternatively, you could use preg_replace in UTF-8 mode but this will probably be slower:

$str = preg_replace('/[\x{10000}-\x{10FFFF}]/u', '', $str);

This works because 4-byte UTF-8 sequences are used for code points in the supplementary Unicode planes starting from 0x10000 .

Answer 3

Here's an example:

<?php 

 mb_internal_encoding("UTF-8");

 //utf8 string,  13 bytes, 9 utf8 chars, 7 ASCII, 1 in latin1, 1 outside the BMP
 $str = "qué \xF0\x9D\x92\xB3 tal"; 
 $array = mbStringToArray($str);
 print "str: [$str]  strlen:" . strlen($str) . " chars:" . count($array) . "\n";
 $str1 = "";
 foreach($array as $c) {
   //  print "$c : " .  strlen($c)  ."\n";
   $str1 .= strlen($c)<=3? $c : '?';
 }
 print "[$str1]\n";


 function mbStringToArray ($str) {
    if (empty($str)) return false;
    $len = mb_strlen($str);
    $array = array();
    for ($i = 0; $i < $len; $i++) {
        $array[] = mb_substr($str, $i, 1);
    }
    return $array;
 }

Or, a little more compact and efficient:

<?php /// 

 mb_internal_encoding("UTF-8");

 //utf8 string,  13 bytes, 9 utf8 chars, 7 ASCII, 1 in latin1, 1 outside the BMP
 $str = "qué \xF0\x9D\x92\xB3 tal";
 $str1 = trimOutsideBMP($str);
 print "original: [$str]\n";
 print "trimmed:  [$str1]\n";


 // Replaces non-BMP characters in the UTF-8 string by a '?' character 
 // Assumes UTF-8 default encoding ( if not sure, call first mb_internal_encoding("UTF-8"); )
 function trimOutsideBMP($str) {
    if (empty($str)) return $str;
    $len = mb_strlen($str);
    $str1 = '';
    for ($i = 0; $i < $len; $i++) {
        $c = mb_substr($str, $i, 1);
        $str1 .= strlen($c) <= 3 ? $c : '?';
    }
    return $str1;
 }

Answer 4

Came across this question when trying to solve my own issue (Facebook spits out certain emoticons as 4-byte characters, Amazon Mechanical Turk does not accept 4-byte characters).

I ended up using this, doesn't require mbstring extension:

function remove_4_byte($string) {
    $char_array = preg_split('/(?<!^)(?!$)/u', $string );
    for($x=0;$x<sizeof($char_array);$x++) {
        if(strlen($char_array[$x])>3) {
            $char_array[$x] = "";
        }
    }
    return implode($char_array, "");
}

Answer 5

Below function change 3 and 4 bytes characters from utf8 string to '#':

function remove3and4bytesCharFromUtf8Str($str) {
        return preg_replace('/([\xF0-\xF7]...)|([\xE0-\xEF]..)/s', '#', $str);
    }

Answer 6

Here is my implementation to filter out 4-byte chars

$string = preg_replace_callback(
    '/./u',
    function (array $match) {
        return strlen($match[0]) >= 4 ? null : $match[0];
    },
    $string
);

you could tweak it and replace null (which removes the char) with some substitute string. You can also replace >= 4 with some other byte-length check.

Answer 7

Another filter implementation, more complex.

It try transliterate to ASCII characters, otherwise iserts unicode replacement character to avoid XSS, eg.: <a href='java\script:alert("XSS")'>

$tr = preg_replace_callback('/([\x{10000}-\x{10FFFF}])/u', function($m){
    $c = iconv('ISO-8859-2', 'UTF-8',iconv('utf-8','ISO-8859-2//TRANSLIT//IGNORE', $m[1]));
    if($c == '')
        return '�';
    return $c;

}, $s);