mb_ereg

One of the differences between preg_match() & mb_ereg()about "captured parenthesized subpattern".<?phppreg_match('/(abc)(.*)/', 'abc', $match);var_dump($match);mb_ereg('(abc)(.*)', 'abc', $match);var_dump($match);?>array(3) {  [0]=>  string(3) "abc"  [1]=>  string(3) "abc"  [2]=>  string(0) ""       // <-- "string"(0) "" : preg_match()}array(3) {  [0]=>  string(3) "abc"  [1]=>  string(3) "abc"  [2]=>  bool(false)       // <-- "bool"(false) : mb_ereg()}

Old link to Oniguruma regex syntax is not working anymore, there is a working one:https://github.com/geoffgarside/oniguruma/blob/master/Syntax.txt

Note that mb_ereg() does not support the \uFFFF unicode syntax but uses \x{FFFF} instead:<?PHP$text = 'Peter is a boy.'; // english$text = 'بيتر هو صبي.'; // arabic//$text = 'פיטר הוא ילד.'; // hebrewmb_regex_encoding('UTF-8');if(mb_ereg('[\x{0600}-\x{06FF}]', $text)) // arabic range//if(mb_ereg('[\x{0590}-\x{05FF}]', $text)) // hebrew range{    echo "Text has some arabic/hebrew characters.";}else{    echo "Text doesnt have arabic/hebrew characters.";}?>

mb_ereg() cannot match over 100,000 (100K) characters (not bytes but characters)whereas preg_match() can over 1,000,000,000 (1G, if it's within "memory_limit").Try this.<?phpini_set("memory_limit", "512M"); // <-- must be changed if you try 1G.$length = 100000; // <-- 99999 is OK / 100000 is NG$str = "";for ($i=0; $i<$length; $i++):    $str .= "1"; // <-- same result if it is a multibyte character.endfor;if (mb_ereg('.*', $str)):    echo '<br><span style="background-color:lightgreen">OK!</span><br>memory_limit = '.ini_get("memory_limit").'<br>$length = '.$length;else:    echo '<br><span style="background-color:orange">NG!</span><br>memory_limit = '.ini_get("memory_limit").'<br>$length = '.$length;endif;?>

If adding ".*" at the end of the pattern returns "false" whereas only one "." returns "true",Suspect the string is too long for the pattern matching.In this case, using preg_match() returns "true" when putting ".*" , but adding more "$" or "\z" returns "false" as expected.

mb_ereg() with a named-subpatternnever catches non-named-subpattern.(Oniguruma's restriction)<?php$str = 'abcdefg';$patternA = '\A(abcd)(.*)\z';        // both caught [1]abcd [2]efg$patternB = '\A(abcd)(?<rest>.*)\z'; // non-named 'abcd' never caughtmb_ereg($patternA, $str, $match);echo '<pre>'.print_r($match, true).'</pre>';mb_ereg($patternB, $str, $match);echo '<pre>'.print_r($match, true).'</pre>';?>Array(    [0] => abcdefg    [1] => abcd    [2] => efg)Array(    [0] => abcdefg    [1] => efg    [rest] => efg)

<?php# What mb_ereg() returns & changes $_3rd_argument into# (Just run this script)function dump2str($var) {    ob_start();    var_dump($var);    $output = ob_get_contents();    ob_end_clean();    return $output;}# (PHP7)empty pattern returns bool(false) with Warning# (PHP8)empty pattern throws ValueError    $emp_ptn = '';try{    $emp_ptn.=  dump2str(mb_ereg('', 'abcde'));}catch(Exception | Error $e){    $emp_ptn.=  get_class($e).'<br>';    $emp_ptn.=  $e->getMessage();    $emp_ptn.=  '<pre>'.$e->getTraceAsString().'</pre>';}echo'PHP '.phpversion().'<br><br>'.'# match<br>'.dump2str(mb_ereg("bcd", "abcde")).' : mb_ereg("bcd", "abcde")<br><br>'.'# match with 3rd argument<br>'.dump2str(mb_ereg("bcd", "abcde", $_3rd)).' : mb_ereg("bcd", "abcde", $_3rd)    // '.dump2str($_3rd).'<br><br>'.'# match (0 byte)<br>'.dump2str(mb_ereg("^", "abcde")).' : mb_ereg("^", "abcde")<br><br>'.'# match (0 byte) with 3rd argument<br>'.dump2str(mb_ereg("^", "abcde", $_3rd)).' : mb_ereg("^", "abcde", $_3rd)    // '.dump2str($_3rd).'<br><br>'.'# unmatch<br>'.dump2str(mb_ereg("f", "abcde")).' : mb_ereg("f", "abcde")<br><br>'.'# unmatch with 3rd argument<br>'.dump2str(mb_ereg("f", "abcde", $_3rd)).' : mb_ereg("f", "abcde", $_3rd)    // '.dump2str($_3rd).'<br><br>'.'# empty pattern<br>'.$emp_ptn.' : mb_ereg("", "abcde")<br><br>'.'# empty pattern with 3rd argument<br>'.$emp_ptn.' : mb_ereg("", "abcde", $_3rd)    // '.dump2str($_3rd).'<br><br>';?>

I hope this information is shown somewhere on php.net.According to "https://github.com/php/php-src/tree/PHP-5.6/ext/mbstring/oniguruma",the bundled Oniguruma regex library version seems ... 4.7.1 between PHP 5.3 - 5.4.45, 5.9.2 between PHP 5.5 - 7.1.16, 6.3.0 since PHP 7.2 - .

mb_ereg() seems unable to Use "named subpattern".preg_match() seems a substitute only in UTF-8 encoding.<?php$text = 'multi_byte_string';$pattern = '.*(?<name>string).*';        // "?P" causes "mbregex compile err" in PHP 5.3.5if(mb_ereg($pattern, $text, $matches)){    echo '<pre>'.print_r($matches, true).'</pre>';}else{    echo 'no match';}?>This code ignores "?<name>" in $pattern and displays below.Array(    [0] => multi_byte_string    [1] => string)$pattern = '/.*(?<name>string).*/u';if(preg_match($pattern, $text, $matches)){instead of lines 2 & 3displays below (in UTF-8 encoding).Array(    [0] => multi_byte_string    [name] => string    [1] => string)

<?php// in PHP_VERSION 7.1// WITHOUT $regs (3rd argument)$int = mb_ereg('abcde', '_abcde_'); // [5 bytes match]var_dump($int);                     // int(1)$int = mb_ereg('ab', '_ab_');       // [2 bytes match]var_dump($int);                     // int(1)$int = mb_ereg('^', '_ab_');        // [0 bytes match]var_dump($int);                     // int(1)$int = mb_ereg('ab', '__');         // [not match]var_dump($int);                     // bool(false)$int = mb_ereg('', '_ab_');         // [error : empty pattern]                                    // Warning: mb_ereg(): empty pattern in ...var_dump($int);                     // bool(false)$int = mb_ereg('ab');               // [error : fewer arguments]                                    // Warning: mb_ereg() expects at least 2 parameters, 1 given in ...var_dump($int);                     // bool(false)                    // Without 3rd argument, mb_ereg() returns either int(1) or bool(false).// WITH $regs (3rd argument)$int = mb_ereg('abcde', '_abcde_', $regs);// [5 bytes match]var_dump($int);                           // int(5)var_dump($regs);                          // array(1) { [0]=> string(5) "abcde" }$int = mb_ereg('ab', '_ab_', $regs);      // [2 bytes match]var_dump($int);                           // int(2)var_dump($regs);                          // array(1) { [0]=> string(2) "ab" }$int = mb_ereg('^', '_ab_', $regs);       // [0 bytes match]var_dump($int);                           // int(1)var_dump($regs);                          // array(1) { [0]=> bool(false) }$int = mb_ereg('ab', '__', $regs);        // [not match]var_dump($int);                           // bool(false)var_dump($regs);                          // array(0) { }$int = mb_ereg('', '_ab_', $regs);        // [error : empty pattern]                                          // Warning: mb_ereg(): empty pattern in ...var_dump($int);                           // bool(false)var_dump($regs);                          // array(0) { }$int = mb_ereg('ab');                     // [error : fewer arguments]                                          // Warning: mb_ereg() expects at least 2 parameters, 1 given in ...var_dump($int);                           // bool(false)var_dump($regs);                          // array(0) { }                    // With 3rd argument, mb_ereg() returns either int(how many bytes matched) or bool(false)                    // and 3rd argument is a bit complicated.?>

While hardly mentioned anywhere, it may be useful to note that mb_ereg uses Oniguruma library internally. The syntax for the default mode (ruby) is described here:http://www.geocities.jp/kosako3/oniguruma/doc/RE.txt

Hebrew regex tested on PHP 5, Ubuntu 8.04.
Seems to work fine without the mb_regex_encoding lines (commented out).
Didn't seem to work with \uxxxx (also commented out).

<?php
echo "Line ";
//mb_regex_encoding("ISO-8859-8");
//if(mb_ereg(".*([\u05d0-\u05ea]).*", $this->current_line))
if(mb_ereg(".*([א-ת]).*", $this->current_line))
{
    echo "has";
}
else
{
    echo "doesn't have";
}
echo " Hebrew characters.<br>";    
//mb_regex_encoding("UTF-8");
?>