preg_match_all

preg_match_all() and other preg_*() functions doesn't work well with very long strings, at least longer that 1Mb.
In this case case function returns FALSE and $matchers value is unpredictable, may contain some values, may be empty.
In this case workaround is pre-split long string onto parts, for instance explode() long string by some criteria and then apply preg_match_all() on each part.
Typical scenario for this case is log analysis by regular expressions.
Tested on PHP 7.2.0

PREG_OFFSET_CAPTURE always seems to provide byte offsets, rather than character position offsets, even when you are using the unicode /u modifier.

Here is a awesome online regex editor https://regex101.com/
which helps you test your regular expressions (prce, js, python) with real-time highlighting of regex match on data input.

if you want to extract all {token}s from a string:

<?php
$pattern = "/{[^}]*}/";
$subject = "{token1} foo {token2} bar";
preg_match_all($pattern, $subject, $matches);
print_r($matches);
?>

output:

Array
(
    [0] => Array
        (
            [0] => {token1}
            [1] => {token2}
        )

)

The code that john at mccarthy dot net posted is not necessary. If you want your results grouped by individual match simply use:

<?
preg_match_all($pattern, $string, $matches, PREG_SET_ORDER);
?>

E.g.

<?
preg_match_all('/([GH])([12])([!?])/', 'G1? H2!', $matches); // Default PREG_PATTERN_ORDER
// $matches = array(0 => array(0 => 'G1?', 1 => 'H2!'),
//                  1 => array(0 => 'G', 1 => 'H'),
//                  2 => array(0 => '1', 1 => '2'),
//                  3 => array(0 => '?', 1 => '!'))

preg_match_all('/([GH])([12])([!?])/', 'G1? H2!', $matches, PREG_SET_ORDER);
// $matches = array(0 => array(0 => 'G1?', 1 => 'G', 2 => '1', 3 => '?'),
//                  1 => array(0 => 'H2!', 1 => 'H', 2 => '2', 3 => '!'))
?>

This is a function to convert byte offsets into (UTF-8) character offsets (this is reagardless of whether you use /u modifier:

<?php

function mb_preg_match_all($ps_pattern, $ps_subject, &$pa_matches, $pn_flags = PREG_PATTERN_ORDER, $pn_offset = 0, $ps_encoding = NULL) {
  // WARNING! - All this function does is to correct offsets, nothing else:
  //
  if (is_null($ps_encoding))
    $ps_encoding = mb_internal_encoding();

  $pn_offset = strlen(mb_substr($ps_subject, 0, $pn_offset, $ps_encoding));
  $ret = preg_match_all($ps_pattern, $ps_subject, $pa_matches, $pn_flags, $pn_offset);

  if ($ret && ($pn_flags & PREG_OFFSET_CAPTURE))
    foreach($pa_matches as &$ha_match)
      foreach($ha_match as &$ha_match)
        $ha_match[1] = mb_strlen(substr($ps_subject, 0, $ha_match[1]), $ps_encoding);
    //
    // (code is independent of PREG_PATTER_ORDER / PREG_SET_ORDER)

  return $ret;
  }

?>

To count str_length in UTF-8 string i use

$count = preg_match_all("/[[:print:]\pL]/u", $str, $pockets);

where
[:print:] - printing characters, including space
\pL - UTF-8 Letter
/u - UTF-8 string
other unicode character properties on http://www.pcre.org/pcre.txt

Here is a function that replaces all occurrences of a number in a string by the number--

<?php
function decremente_chaine($chaine)
    {
        //récupérer toutes les occurrences de nombres et leurs indices
        preg_match_all("/[0-9]+/",$chaine,$out,PREG_OFFSET_CAPTURE);
            //parcourir les occurrences 
            for($i=0;$i<sizeof($out[0]);$i++)
            {
                $longueurnombre = strlen((string)$out[0][$i][0]);
                $taillechaine = strlen($chaine);
                // découper la chaine en 3 morceaux
                $debut = substr($chaine,0,$out[0][$i][1]);
                $milieu = ($out[0][$i][0])-1;
                $fin = substr($chaine,$out[0][$i][1]+$longueurnombre,$taillechaine);
                 // si c'est 10,100,1000 etc. on décale tout de 1 car le résultat comporte un chiffre de moins
                 if(preg_match('#[1][0]+$#', $out[0][$i][0]))
                 {
                    for($j = $i+1;$j<sizeof($out[0]);$j++)
                    {
                        $out[0][$j][1] = $out[0][$j][1] -1;
                    }
                 }
                $chaine = $debut.$milieu.$fin;
            }
        return $chaine;
    }
?>

Here's some fleecy code to 1. validate RCF2822 conformity of address lists and 2. to extract the address specification (the part commonly known as 'email'). I wouldn't suggest using it for input form email checking, but it might be just what you want for other email applications. I know it can be optimized further, but that part I'll leave up to you nutcrackers. The total length of the resulting Regex is about 30000 bytes. That because it accepts comments. You can remove that by setting $cfws to $fws and it shrinks to about 6000 bytes. Conformity checking is absolutely and strictly referring to RFC2822. Have fun and email me if you have any enhancements!

<?php
function mime_extract_rfc2822_address($string)
{
        //rfc2822 token setup
        $crlf           = "(?:\r\n)";
        $wsp            = "[\t ]";
        $text           = "[\\x01-\\x09\\x0B\\x0C\\x0E-\\x7F]";
        $quoted_pair    = "(?:\\\\$text)";
        $fws            = "(?:(?:$wsp*$crlf)?$wsp+)";
        $ctext          = "[\\x01-\\x08\\x0B\\x0C\\x0E-\\x1F" .
                          "!-'*-[\\]-\\x7F]";
        $comment        = "(\\((?:$fws?(?:$ctext|$quoted_pair|(?1)))*" .
                          "$fws?\\))";
        $cfws           = "(?:(?:$fws?$comment)*(?:(?:$fws?$comment)|$fws))";
        //$cfws           = $fws; //an alternative to comments
        $atext          = "[!#-'*+\\-\\/0-9=?A-Z\\^-~]";
        $atom           = "(?:$cfws?$atext+$cfws?)";
        $dot_atom_text  = "(?:$atext+(?:\\.$atext+)*)";
        $dot_atom       = "(?:$cfws?$dot_atom_text$cfws?)";
        $qtext          = "[\\x01-\\x08\\x0B\\x0C\\x0E-\\x1F!#-[\\]-\\x7F]";
        $qcontent       = "(?:$qtext|$quoted_pair)";
        $quoted_string  = "(?:$cfws?\"(?:$fws?$qcontent)*$fws?\"$cfws?)";
        $dtext          = "[\\x01-\\x08\\x0B\\x0C\\x0E-\\x1F!-Z\\^-\\x7F]";
        $dcontent       = "(?:$dtext|$quoted_pair)";
        $domain_literal = "(?:$cfws?\\[(?:$fws?$dcontent)*$fws?]$cfws?)";
        $domain         = "(?:$dot_atom|$domain_literal)";
        $local_part     = "(?:$dot_atom|$quoted_string)";
        $addr_spec      = "($local_part@$domain)";
        $display_name   = "(?:(?:$atom|$quoted_string)+)";
        $angle_addr     = "(?:$cfws?<$addr_spec>$cfws?)";
        $name_addr      = "(?:$display_name?$angle_addr)";
        $mailbox        = "(?:$name_addr|$addr_spec)";
        $mailbox_list   = "(?:(?:(?:(?<=:)|,)$mailbox)+)";
        $group          = "(?:$display_name:(?:$mailbox_list|$cfws)?;$cfws?)";
        $address        = "(?:$mailbox|$group)";
        $address_list   = "(?:(?:^|,)$address)+";

        //output length of string (just so you see how f**king long it is)
        echo(strlen($address_list) . " ");

        //apply expression
        preg_match_all("/^$address_list$/", $string, $array, PREG_SET_ORDER);

        return $array;
};
?>

Recently I had to write search engine in hebrew and ran into huge amount of problems. My data was stored in MySQL table with utf8_bin encoding.

So, to be able to write hebrew in utf8 table you need to do
<?php
$prepared_text = addslashes(urf8_encode($text));
?>

But then I had to find if some word exists in stored text. This is the place I got stuck. Simple preg_match would not find text since hebrew doesnt work that easy. I've tried with /u and who kows what else.

Solution was somewhat logical and simple... 
<?php
$db_text = bin2hex(stripslashes(utf8_decode($db_text)));
$word = bin2hex($word);

$found = preg_match_all("/($word)+/i", $db_text, $matches);
?>

I've used preg_match_all since it returns number of occurences. So I could sort search results acording to that.

Hope someone finds this useful!

I needed a function to rotate the results of a preg_match_all query, and made this. Not sure if it exists.

<?php
function turn_array($m)
{
    for ($z = 0;$z < count($m);$z++)
    {
        for ($x = 0;$x < count($m[$z]);$x++)
        {
            $rt[$x][$z] = $m[$z][$x];
        }
    }    
    
    return $rt;
}
?>

Example - Take results of some preg_match_all query:

Array
(
    [0] => Array
        (
            [1] => Banff 
            [2] => Canmore
            [3] => Invermere
        )
 
    [1] => Array
        (
            [1] => AB 
            [2] => AB
            [3] => BC
        )
 
    [2] => Array
        (
            [1] => 51.1746254 
            [2] => 51.0938416
            [3] => 50.5065193
        )
 
    [3] => Array
        (
            [1] => -115.5719757 
            [2] => -115.3517761
            [3] => -116.0321884
        )
 
    [4] => Array
        (
            [1] => T1L 1B3 
            [2] => T1W 1N2
            [3] => V0B 2G0
        )

)

Rotate it 90 degrees to group results as records:

Array
(
    [0] => Array
        (
            [1] => Banff 
            [2] => AB
            [3] => 51.1746254
            [4] => -115.5719757
            [5] => T1L 1B3
        )
 
    [1] => Array
        (
            [1] => Canmore
            [2] => AB
            [3] => 51.0938416
            [4] => -115.3517761
            [5] => T1W 1N2
        )
 
    [2] => Array
        (
            [1] => Invermere
            [2] => BC
            [3] => 50.5065193
            [4] => -116.0321884
            [5] => V0B 2G0
        )
)

Be careful with this pattern match and large input buffer on preg_match_* functions.

<?php
$pattern = '/\{(?:[^{}]|(?R))*\}/';

preg_match_all($pattern, $buffer, $matches); 
?>

if $buffer is 80+ KB in size, you'll end up with segfault! 

[89396.588854] php[4384]: segfault at 7ffd6e2bdeb0 ip 00007fa20c8d67ed sp 00007ffd6e2bde70 error 6 in libpcre.so.3.13.1[7fa20c8c3000+3c000]

This is due to the PCRE recursion. This is a known bug in PHP since 2008, but it's source is not PHP itself but PCRE library. 

Rasmus Lerdorf has the answer: https://bugs.php.net/bug.php?id=45735#1365812629

"The problem here is that there is no way to detect run-away regular expressions 
here without huge performance and memory penalties. Yes, we could build PCRE in a 
way that it wouldn't segfault and we could crank up the default backtrack limit 
to something huge, but it would slow every regex call down by a lot. If PCRE 
provided a way to handle this in a more graceful manner without the performance 
hit we would of course use it."

I found simpleXML to be useful only in cases where the XML was extremely small, otherwise the server would run out of memory (I suspect there is a memory leak or something?). So while searching for alternative parsers, I decided to try a simpler approach. I don't know how this compares with cpu usage, but I know it works with large XML structures. This is more a manual method, but it works for me since I always know what structure of data I will be receiving. 

Essentially I just preg_match() unique nodes to find the values I am looking for, or I preg_match_all to find multiple nodes. This puts the results in an array and I can then process this data as I please.

I was unhappy though, that preg_match_all() stores the data twice (requiring twice the memory), one array for all the full pattern matches, and one array for all the sub pattern matches. You could probably write your own function that overcame this. But for now this works for me, and I hope it saves someone else some time as well.

// SAMPLE XML
<RETS ReplyCode="0" ReplyText="Operation Successful">
  <COUNT Records="14" />
  <DELIMITER value="09" />
  <COLUMNS>PropertyID</COLUMNS>
  <DATA>521897</DATA>
  <DATA>677208</DATA>
  <DATA>686037</DATA>
</RETS>

<?PHP

// SAMPLE FUNCTION
function parse_xml($xml) {
    
    
    // GET DELIMITER (single instance)
    $match_res = preg_match('/<DELIMITER value ?= ?"(.*)" ?\/>/', $xml, $matches);
    if(!empty($matches[1])) {
        $results["delimiter"] = chr($matches[1]);
    } else {
        // DEFAULT DELIMITER
        $results["delimiter"] = "\t";
    }
    unset($match_res, $matches);
    
    
    // GET MULTIPLE DATA NODES (multiple instances)
    $results["data_count"] = preg_match_all("/<DATA>(.*)<\/DATA>/", $xml, $matches);
    // GET MATCHES OF SUB PATTERN, DISCARD THE REST
    $results["data"]=$matches[1];
    unset($match_res, $matches);
    
    // UNSET XML TO SAVE MEMORY (should unset outside the function as well)
    unset($xml);

    // RETURN RESULTS ARRAY
    return $results;
    
    
}

?>

Better use preg_replace to convert text in a clickable link with tag <a> 

$html = preg_replace('"\b(http://\S+)"', '<a href="$1">$1</a>', $text);

For parsing queries with entities use:

<?php
preg_match_all("/(?:^|(?<=\&(?![a-z]+\;)))([^\=]+)=(.*?)(?:$|\&(?![a-z]+\;))/i",
  $s, $m, PREG_SET_ORDER );
?>

Perhaps you want to find the positions of all anchor tags.  This will return a two dimensional array of which the starting and ending positions will be returned.

<?php
function getTagPositions($strBody)
{
    define(DEBUG, false);
    define(DEBUG_FILE_PREFIX, "/tmp/findlinks_");
    
    preg_match_all("/<[^>]+>(.*)<\/[^>]+>/U", $strBody, $strTag, PREG_PATTERN_ORDER);
    $intOffset = 0;
    $intIndex = 0;
    $intTagPositions = array();

    foreach($strTag[0] as $strFullTag) {
        if(DEBUG == true) {
            $fhDebug = fopen(DEBUG_FILE_PREFIX.time(), "a");
            fwrite($fhDebug, $fulltag."\n");
            fwrite($fhDebug, "Starting position: ".strpos($strBody, $strFullTag, $intOffset)."\n");
            fwrite($fhDebug, "Ending position: ".(strpos($strBody, $strFullTag, $intOffset) + strlen($strFullTag))."\n");
            fwrite($fhDebug, "Length: ".strlen($strFullTag)."\n\n");
            fclose($fhDebug);
        }
        $intTagPositions[$intIndex] = array('start' => (strpos($strBody, $strFullTag, $intOffset)), 'end' => (strpos($strBody, $strFullTag, $intOffset) + strlen($strFullTag)));
        $intOffset += strlen($strFullTag);
        $intIndex++;
    }
    return $intTagPositions;
}

$strBody = 'I have lots of <a href="http://my.site.com">links</a> on this <a href="http://my.site.com">page</a> that I want to <a href="http://my.site.com">find</a> the positions.';

$strBody = strip_tags(html_entity_decode($strBody), '<a>');
$intTagPositions = getTagPositions($strBody);
print_r($intTagPositions);

/*****
Output:

Array ( 
    [0] => Array ( 
        [start] => 15 
        [end] => 53 ) 
    [1] => Array ( 
        [start] => 62 
        [end] => 99 ) 
    [2] => Array ( 
        [start] => 115 
        [end] => 152 )
 ) 
*****/
?>

i have made up a simple function to extract a number from a string..

I am not sure how good it is, but it works.

It gets only the numbers 0-9, the "-", " ", "(", ")", "."

characters.. This is as far as I know the most widely used characters for a Phone number.

<?php
function clean_phone_number($phone) {
       if (!empty($phone)) {
               //var_dump($phone);
               preg_match_all('/[0-9\(\)+.\- ]/s', $phone, $cleaned);
               foreach($cleaned[0] as $k=>$v) {
                       $ready .= $v;
               }
               var_dump($ready);
               die;
               if (mb_strlen($cleaned) > 4 && mb_strlen($cleaned) <=25) {
                       return $cleaned;
               }
               else {
                       return false;
               }
       }
       return false;
}
?>

A multi-byte safe preg_match_all that fixes capture offsets when using PREG_OFFSET_CAPTURE on utf-8 strings
 
<?php 
function mb_preg_match_all($pattern, $subject, &$matches = null, $flags = 0, $offset = 0) {
    $out=preg_match_all($pattern, $subject, $matches, $flags, $offset);
    if($flags & PREG_OFFSET_CAPTURE && is_array($matches) && count($matches)>0) {
        foreach ($matches[0] as &$match) {
            $match[1] = mb_strlen(substr($subject, 0, $match[1]));
        }
    }
    return $out;
}
?>

Sometimes you don't just want to cherry-pick the matches but need that the entire subject is made up from matching substrings, so every character of the subject is a member of a match. None of the existing preg_* function is easily applicable for this task, so I made the preg_match_entire() function.
It uses the (*MARK) syntax which is documented here: https://pcre.org/original/doc/html/pcrepattern.html#SEC27

<?php 

// returns: the array of matches
// null if the string is not a repetition of the pattern
// false on error
function preg_match_entire( string $pattern, string $subject, int $flags = 0 ){
  // Rebuild and wrap the pattern
  $delimiter = $pattern[0];
  $ldp       = strrpos( $pattern, $delimiter );
  $pattern   = substr( $pattern, 1, $ldp - 1 );
  $modifiers = substr( $pattern,    $ldp + 1 );
  $pattern   = "{$delimiter}   \G\z (*MARK:END)   |   \G (?:{$pattern})   {$delimiter}x{$modifiers}";
  $r = preg_match_all( $pattern, $subject, $m, PREG_SET_ORDER | $flags );
  if( $r === false )  return false;  // error
  $end = array_pop( $m );
  if( $end === null || ! isset( $end['MARK']) || $end['MARK'] !== 'END')
    return null;  // end of string not reached
  return $m;  // return actual matches, may be an empty array
}

// Same results:
test('#{\d+}#', '');              // []
test('#{\d+}#', '{11}{22}{33}');  // {11},{22},{33}

// Different results: preg_match_entire won't match this:
test('#{\d+}#', '{11}{}{aa}{22},{{33}}');
// preg_match_entire: null
// preg_match_all:    {11},{22},{33}

function test( $pattern, $subject ){
  echo "pattern:           $pattern\n";
  echo "subject:           $subject\n";
  print_matches('preg_match_entire: ', preg_match_entire( $pattern, $subject ));
  preg_match_all( $pattern, $subject, $matches, PREG_SET_ORDER );
  print_matches('preg_match_all:    ', $matches );
  echo "\n";
}
function print_matches( $t, $m ){
  echo $t, is_array( $m ) && $m ? implode(',', array_column( $m, 0 )) : json_encode( $m ), "\n";
} ?>

<?php
//Allow limited span formatting in html text

$str='<span style="text-decoration-line: underline; font-weight: bold; font-style: italic;">White</span>
<span style="text-decoration-line: underline;">RED</span><span style="color:blue">blue</span>';

function next_format($str)
{
     $array=array("text-decoration-line"=>"underline","font-weight"=>"bold","font-style"=>"italic");
    foreach ($array as $key=>$val)
    {
          if($str[1]==$key && $str[2]==$val)
        {
              return $str[1].': '.$str[2].";";
        }
     }
          return '';
  
}
function next_span($matches)
{
  $needFormat=preg_replace_callback('/([a-z\-]+):\s*([^;]+)(;|)/ism',"next_format",$matches[2]);
  return $matches[1].$needFormat.$matches[3];
  
}
 echo preg_replace_callback(
            "/(\<span\s+style\=\")([^\"]+)(\">)/ism",
            "next_span",
            $str);
?>

Why <?php preg_match_all('/(?:^|\s)(ABC|XYZ)(?:\s|$)/i', 'ABC  XYZ', $match) ?> finds only 'ABC'?

Because the first full match is 'ABC ' - containing the trailing space. And that space is not available for further processing.

Use lookbehind and lookahead to solve this problem: <?php preg_match_all('/(?<=^|\s)(ABC|XYZ)(?=\s|$)/i', 'ABC XYZ', $match)  ?>

when regex is for longer and shorter version of a string,
only one of that long and short versions is catched.
when regex match occurs in one position of string,
only one match is saved in matches[0] for that position.
if ? is used, regex is greedy, and catches more long version,
if | is used, most first matching variant is catched:
<?php
preg_match_all('/ab|abc/','abc',$m);
var_dump($m);
preg_match_all('/abc?/','abc',$m);
var_dump($m);
?>
['ab', 'abc'] in $m[0] for both can be expected, but it is not so,
actually they output [['ab']] and [['abc']]:
array(1) {
  [0]=>
  array(1) {
    [0]=>
    string(2) "ab"
  }
}
array(1) {
  [0]=>
  array(1) {
    [0]=>
    string(3) "abc"
  }
}

EXAMPLE:
$file = file_get_contents('file');
if(preg_match_all('#Task To Run(.*)#s', $file, $m)) {
var_dump($m);
}

No output...

preg_match_all  not work if file exist BOM bytes (FF FE) :

╰─$ head -n1 file | hexdump -C
00000000  ff fe 48 00 6f 00 73 00  74 00 4e 00 61 00 6d 00  |..H.o.s.t.N.a.m.|

clear BOM via dos2unix:

╰─$ dos2unix file 
dos2unix: converting UTF-16LE file file to UTF-8 Unix format...

Check again:

╰─$ head -n1 file | hexdump -C
00000000  48 6f 73 74 4e 61 6d 65  3a 20 20 20 20 20 20 20  |HostName:       |

Great! Now preg_match_all works fine.

Using preg_match_all I made a pretty handy function.

<?php

function reg_smart_replace($pattern, $replacement, $subject, $replacementChar = "$$$", $limit = -1)
{
    if (! $pattern || ! $subject || ! $replacement ) { return false; }
    
    $replacementChar = preg_quote($replacementChar);
    
    preg_match_all ( $pattern, $subject, $matches);
    
    if ($limit > -1) {
        foreach ($matches as $count => $value )
        {
            if ($count + 1 > $limit ) { unset($matches[$count]); }
        }
    }
    foreach ($matches[0] as $match) {
        $rep = ereg_replace($replacementChar, $match, $replacement);
        $subject = ereg_replace($match, $rep, $subject);
    }
    
    return $subject;
}
?>

This function can turn blocks of text into clickable links or whatever.  Example:

<?php
reg_smart_replace(EMAIL_REGEX, '<a href="mailto:$$$">$$$</a>', $description)
?>
will turn all email addresses into actual links.

Just substitute $$$ with the text that will be found by the regex.  If you can't use $$$ then use the 4th parameter $replacementChar

Here is a way to match everything on the page, performing an action for each match as you go. I had used this idiom in other languages, where its use is customary, but in PHP it seems to be not quite as common.

<?php
function custom_preg_match_all($pattern, $subject)
{
    $offset = 0;
    $match_count = 0;
    while(preg_match($pattern, $subject, $matches, PREG_OFFSET_CAPTURE, $offset))
    {
        // Increment counter
        $match_count++;
    
        // Get byte offset and byte length (assuming single byte encoded)
        $match_start = $matches[0][1];
        $match_length = strlen(matches[0][0]);

        // (Optional) Transform $matches to the format it is usually set as (without PREG_OFFSET_CAPTURE set)
        foreach($matches as $k => $match) $newmatches[$k] = $match[0];
        $matches = $new_matches;
    
        // Your code here
        echo "Match number $match_count, at byte offset $match_start, $match_length bytes long: ".$matches[0]."\r\n";
            
        // Update offset to the end of the match
        $offset = $match_start + $match_length;
    }

    return $match_count;
}
?>

Note that the offsets returned are byte values (not necessarily number of characters) so you'll have to make sure the data is single-byte encoded. (Or have a look at paolo mosna's strByte function on the strlen manual page).
I'd be interested to know how this method performs speedwise against using preg_match_all and then recursing through the results.

If PREG_OFFSET_CAPTURE is set then unmatched captures (i.e. ones with '?') will not be present in the result array. This is presumably because there is no offset, and thus the original PHP dev decided best to just leave it out.

I had been crafting and testing some regexp patterns online using the tools Regex101 and a `preg_match_all()` tester and found that the regexp patterns I wrote worked fine on them, just not in my code.

My problem was not double-escaping backslash characters:

<?php
// Input test
$input = "\"something\",\"something here\",\"some\nnew\nlines\",\"this is the end\"";

// Work with online regexp testers, doesn't work in PHP
preg_match_all( "/(?:,|^)(?<!\\)\".*?(?<!\\)\"(?:(?=,)|$)/s", $input, $matches );

/*
Outputs: NULL
*/

// Works with online regexp testers, does work in PHP
preg_match_all( "/(?:,|^)(?<!\\\\)\".*?(?<!\\\\)\"(?:(?=,)|$)/s", $input, $matches );

/*
Outputs:
array(2) {
  [0]=>
  array(4) {
    [0]=>
    string(11) ""something""
    [1]=>
    string(17) ","something here""
    [2]=>
    string(17) ","some
new
lines""
    [3]=>
    string(18) ","this is the end""
  }
  [1]=>
  array(4) {
    [0]=>
    string(9) "something"
    [1]=>
    string(14) "something here"
    [2]=>
    string(14) "some
new
lines"
    [3]=>
    string(15) "this is the end"
  }
}
*/
?>

If you'd like to include DOUBLE QUOTES on a regular expression for use with preg_match_all, try ESCAPING THRICE, as in: \\\"

For example, the pattern:
'/<table>[\s\w\/<>=\\\"]*<\/table>/'

Should be able to match:
<table>
<row>
<col align="left" valign="top">a</col>
<col align="right" valign="bottom">b</col>
</row>
</table>
.. with all there is under those table tags.

I'm not really sure why this is so, but I tried just the double quote and one or even two escape characters and it won't work. In my frustration I added another one and then it's cool.

I have received complains, that my html2a() code (see below) doesn't work in some cases. 
It is however not the problem with algorithm or procedure, but with PCRE recursive stack limits.

If you use recursive PCRE (?R) you should remember to increase those two ini settings:

ini_set('pcre.backtrack_limit', 10000000);
ini_set('pcre.recursion_limit', 10000000);

But be warned: (from php.ini)

;Please note that if you set this value to a high number you may consume all
;the available process stack and eventually crash PHP (due to reaching the
;stack size limit imposed by the Operating System).

I have written this example mainly to demonstrate the power of PCRE LANGUAGE, not the power of it's implementation  :) 

But if you like it, use it, of course on your own risk.

As I intended to create for my own purpose a clean PHP class to act on XML files, combining the use of DOM and simplexml functions, I had that small problem, but very annoying, that the offsets in a path is not numbered the same in both. 

That is to say, for example, if i get a DOM xpath object it appears like:
/ANODE/ANOTHERNODE/SOMENODE[9]/NODE[2]
and as a simplexml object would be equivalent to:
ANODE->ANOTHERNODE->SOMENODE[8]->NODE[1]

So u see what I mean? I used preg_match_all to solve that problem, and finally I got this after some hours of headlock (as I'm french the names of variables are in French sorry), hoping it could be useful to some of you:

<?php
function decrease_string($string)
    {
        /* retrieve all occurrences AND offsets of numbers in the original string: */

        preg_match_all("/[0-9]+/",$chaine,$out,PREG_OFFSET_CAPTURE);
            for($i=0;$i<sizeof($out[0]);$i++)
            {
                $longueurnombre = strlen((string)$out[0][$i][0]);
                $taillechaine = strlen($chaine);
                // cut the string in 3 pieces
                $debut = substr($chaine,0,$out[0][$i][1]);
                $milieu = ($out[0][$i][0])-1;
                $fin = substr($chaine,$out[0][$i][1]+$longueurnombre,$taillechaine);
                 /* if it's 10,100,1000, the problem is that the string gets shorter and it shifts all the offsets, so we have to decrease them of 1 */
                 if(preg_match('#[1][0]+$#', $out[0][$i][0]))
                 {
                    for($j = $i+1;$j<sizeof($out[0]);$j++)
                    {
                        $out[0][$j][1] = $out[0][$j][1] -1;
                    }
                 }
                $chaine = $debut.$milieu.$fin;
            }
        return $chaine;
    }
?>

please note, that the function of "mail at SPAMBUSTER at milianw dot de" can result in invalid xhtml in some cases. think i used it in the right way but my result is sth like this:

<img src="./img.jpg" alt="nice picture" />foo foo foo foo </img>

correct me if i'm wrong. 
i'll see when there's time to fix that. -.-

Here http://tryphpregex.com/ is a  php based online regex editor which helps you test your regular expressions with real-time highlighting of regex match on data input.

This is very useful to combine matches:
$a = array_combine($matches[1], $matches[2]);

Extract fields out of csv string : ( since before php5.3 you can't use str_getcsv function ) 
Here is the regex :

<?php

$csvData = <<<EOF
10,'20',"30","'40","'50'","\"60","70,80","09\\/18,/\"2011",'a,sdfcd'
EOF

$reg = <<<EOF
/
    (
        (
            ([\'\"])
            (
               ( 
                [^\'\"]
                |
                (\\\\.)
               )*
            )
            (\\3)
            |
            (
                [^,]
                |
                (\\\\.)
            )*
    ),)
    /x
EOF;

preg_match_all($reg,$csvData,$matches);

// to extract csv fields
print_r($matches[2]);
?>

<?php
// Returns an array of strings where the start and end are found
    function findinside($start, $end, $string) {
        preg_match_all('/' . preg_quote($start, '/') . '([^\.)]+)'. preg_quote($end, '/').'/i', $string, $m);
        return $m[1];
    }
    
    $start = "mary has";
    $end = "lambs.";
    $string = "mary has 6 lambs. phil has 13 lambs. mary stole phil's lambs. now mary has all the lambs.";

    $out = findinside($start, $end, $string);

    print_r ($out);

/* Results in 
(
    [0] =>  6 
    [1] =>  all the 
)
*/ 
?>

The power of pregs is limited only by your *imagination* :)
I wrote this html2a() function using preg recursive match (?R) which provides quite safe and bulletproof html/xml extraction:
<?php
function html2a ( $html ) {
  if ( !preg_match_all( '
@
\<\s*?(\w+)((?:\b(?:\'[^\']*\'|"[^"]*"|[^\>])*)?)\>
((?:(?>[^\<]*)|(?R))*)
\<\/\s*?\\1(?:\b[^\>]*)?\>
|\<\s*(\w+)(\b(?:\'[^\']*\'|"[^"]*"|[^\>])*)?\/?\>
@uxis', $html = trim($html), $m, PREG_OFFSET_CAPTURE | PREG_SET_ORDER) )
    return $html;
  $i = 0;
  $ret = array();
  foreach ($m as $set) {
    if ( strlen( $val = trim( substr($html, $i, $set[0][1] - $i) ) ) )
      $ret[] = $val;
    $val = $set[1][1] < 0 
      ? array( 'tag' => strtolower($set[4][0]) )
      : array( 'tag' => strtolower($set[1][0]), 'val' => html2a($set[3][0]) );
    if ( preg_match_all( '
/(\w+)\s*(?:=\s*(?:"([^"]*)"|\'([^\']*)\'|(\w+)))?/usix
', isset($set[5]) && $set[2][1] < 0
  ? $set[5][0]
  : $set[2][0]
  ,$attrs, PREG_SET_ORDER ) ) {
      foreach ($attrs as $a) {
        $val['attr'][$a[1]]=$a[count($a)-1];
      }
    }
    $ret[] = $val;
    $i = $set[0][1]+strlen( $set[0][0] );
  }
  $l = strlen($html);
  if ( $i < $l )
    if ( strlen( $val = trim( substr( $html, $i, $l - $i ) ) ) )
      $ret[] = $val;
  return $ret;
}
?>

Now let's try it with this example: (there are some really nasty xhtml compliant bugs, but ... we shouldn't worry)

<?php
$html = <<<EOT
some leftover text...
     < DIV class=noCompliant style = "text-align:left;" >
... and some other ...
< dIv > < empty>  </ empty>
  <p> This is yet another text <br  >
     that wasn't <b>compliant</b> too... <br   />
     </p>
 <div class="noClass" > this one is better but we don't care anyway </div ><P>
    <input   type= "text"  name ='my "name' value  = "nothin really." readonly>
end of paragraph </p> </Div>   </div>   some trailing text 
EOT;

$a = html2a($html);
//now we will make some neat html out of it
echo a2html($a);

function a2html ( $a, $in = "" ) {
  if ( is_array($a) ) {
    $s = "";
    foreach ($a as $t)
      if ( is_array($t) ) {
        $attrs=""; 
        if ( isset($t['attr']) )
          foreach( $t['attr'] as $k => $v )
            $attrs.=" ${k}=".( strpos( $v, '"' )!==false ? "'$v'" : "\"$v\"" );
        $s.= $in."<".$t['tag'].$attrs.( isset( $t['val'] ) ? ">\n".a2html( $t['val'], $in."  " ).$in."</".$t['tag'] : "/" ).">\n";
      } else
        $s.= $in.$t."\n";
  } else {
    $s = empty($a) ? "" : $in.$a."\n";
  }
  return $s;
}
?>
This produces:
some leftover text...
<div class="noCompliant" style="text-align:left;">
  ... and some other ...
  <div>
    <empty>
    </empty>
    <p>
      This is yet another text
      <br/>
      that wasn't
      <b>
        compliant
      </b>
      too...
      <br/>
    </p>
    <div class="noClass">
      this one is better but we don't care anyway
    </div>
    <p>
      <input type="text" name='my "name' value="nothin really." readonly="readonly"/>
      end of paragraph
    </p>
  </div>
</div>
some trailing text

The next function works with almost any complex xml/xhtml string

<?php
/**
* Find and close unclosed xml tags
**/
function close_tags($text) {
    $patt_open    = "%((?<!</)(?<=<)[\s]*[^/!>\s]+(?=>|[\s]+[^>]*[^/]>)(?!/>))%";
    $patt_close    = "%((?<=</)([^>]+)(?=>))%";
    if (preg_match_all($patt_open,$text,$matches))
    {
        $m_open = $matches[1];
        if(!empty($m_open))
        {
            preg_match_all($patt_close,$text,$matches2);
            $m_close = $matches2[1];
            if (count($m_open) > count($m_close))
            {
                $m_open = array_reverse($m_open);
                foreach ($m_close as $tag) $c_tags[$tag]++;
                foreach ($m_open as $k => $tag)    if ($c_tags[$tag]--<=0) $text.='</'.$tag.'>';
            }
        }
    }
    return $text;
}
?>

// Here is function that allows you to preg_match_all array of patters

function getMatches($pattern, $subject) {
    $matches = array();

    if (is_array($pattern)) {
        foreach ($pattern as $p) {
            $m = getMatches($p, $subject);

            foreach ($m as $key => $match) {
                if (isset($matches[$key])) {
                    $matches[$key] = array_merge($matches[$key], $m[$key]);    
                } else {
                    $matches[$key] = $m[$key];
                }
            }
        }
    } else {
        preg_match_all($pattern, $subject, $matches);
    }

    return $matches;
}

$patterns = array(
    '/<span>(.*?)<\/span>/',
    '/<a href=".*?">(.*?)<\/a>/'
);

$html = '<span>some text</span>';
$html .= '<span>some text in another span</span>';
$html .= '<a href="path/">here is the link</a>';
$html .= '<address>address is here</address>';
$html .= '<span>here is one more span</span>';

$matches = getMatches($patterns, $html);

print_r($matches); // result is below

/*
Array
(
    [0] => Array
        (
            [0] => <span>some text</span>
            [1] => <span>some text in another span</span>
            [2] => <span>here is one more span</span>
            [3] => <a href="path/">here is the link</a>
        )

    [1] => Array
        (
            [0] => some text
            [1] => some text in another span
            [2] => here is one more span
            [3] => here is the link
        )

)
*/