preg_replace_callback()

(PHP 4 >= 4.0.5, PHP 5, PHP 7)

执行一个正则表达式搜索并且使用一个回调进行替换

说明

preg_replace_callback(mixed $pattern,callable $callback, mixed $subject[,int $limit= -1[,int &$count]]): mixed

这个函数的行为除了可以指定一个$callback替代$replacement进行替换字符串的计算，其他方面等同于preg_replace()。

参数

$pattern

要搜索的模式，可以是字符串或一个字符串数组。

$callback

一个回调函数，在每次需要替换时调用，调用时函数得到的参数是从$subject中匹配到的结果。回调函数返回真正参与替换的字符串。这是该回调函数的签名：

handler(array $matches): string

你可能经常会需要$callback函数而仅用于preg_replace_callback()一个地方的调用。在这种情况下，你可以使用匿名函数来定义一个匿名函数作为preg_replace_callback()调用时的回调。这样做你可以保留所有调用信息在同一个位置并且不会因为一个不在任何其他地方使用的回调函数名称而污染函数名称空间。

preg_replace_callback()和匿名函数

<?php
/* 一个unix样式的命令行过滤器，用于将段落开始部分的大写字母转换为小写。 */
$fp = fopen("php://stdin", "r") or die("can't read stdin");
while (!feof($fp)) {
    $line = fgets($fp);
    $line = preg_replace_callback(
        '|<p>\s*\w|',
        function ($matches) {
            return strtolower($matches[0]);
        },
        $line
    );
    echo $line;
}
fclose($fp);
?>

$subject

要搜索替换的目标字符串或字符串数组。

$limit

对于每个模式用于每个$subject字符串的最大可替换次数。默认是-1（无限制）。

$count

如果指定，这个变量将被填充为替换执行的次数。

返回值

如果$subject是一个数组，preg_replace_callback()返回一个数组，其他情况返回字符串。错误发生时返回NULL。

如果查找到了匹配，返回替换后的目标字符串（或字符串数组），其他情况$subject将会无变化返回。

更新日志

范例

preg_replace_callback()示例

<?php
// 将文本中的年份增加一年.
$text = "April fools day is 04/01/2002\n";
$text.= "Last christmas was 12/24/2001\n";
// 回调函数
function next_year($matches)
{
  // 通常: $matches[0]是完成的匹配
  // $matches[1]是第一个捕获子组的匹配
  // 以此类推
  return $matches[1].($matches[2]+1);
}
echo preg_replace_callback(
            "|(\d{2}/\d{2}/)(\d{4})|",
            "next_year",
            $text);
?>

以上例程会输出：

April fools day is 04/01/2003
Last christmas was 12/24/2002

preg_replace_callback()使用递归构造处理BB码的封装

<?php
$input = "plain [indent] deep [indent] deeper [/indent] deep [/indent] plain";
function parseTagsRecursive($input)
{
     /* 译注: 对此正则表达式分段分析
     * 首尾两个#是正则分隔符
     * \[indent] 匹配一个原文的[indent]
     * ((?:[^[]|\[(?!/?indent])|(?R))+)分析:
     *   (?:[^[]|\[(?!/?indent])分析:
     *  首先它是一个非捕获子组
     *   两个可选路径, 一个是非[字符, 另一个是[字符但后面紧跟着不是/indent或indent.
     *   (?R) 正则表达式递归
     *     \[/indent] 匹配结束的[/indent]
     * /
    $regex = '#\[indent]((?:[^[]|\[(?!/?indent])|(?R))+)\[/indent]#';
    if (is_array($input)) {
        $input = '<div style="margin-left: 10px">'.$input[1].'</div>';
    }
    return preg_replace_callback($regex, 'parseTagsRecursive', $input);
}
$output = parseTagsRecursive($input);
echo $output;
?>

参见

• PCRE 模式
• preg_replace_callback_array()Perform a regular expression search and replace using callbacks
• preg_quote()转义正则表达式字符
• preg_replace()执行一个正则表达式的搜索和替换
• preg_last_error()返回最后一个PCRE正则执行产生的错误代码
• 匿名函数
• callback类型的信息

The easiest way to pass more than one parameters to the callback function is with the 'use' keyword. 
[This is better than using global, because it works even when we are already inside a function.]
In this example, the callback function is an anonymous function, which takes one argument, $match, supplied by preg_replace_callback(). The extra 
"use ($ten)" puts the $ten variable into scope for the function.
<?php
$string = "Some numbers: one: 1; two: 2; three: 3 end";
$ten = 10;
$newstring = preg_replace_callback(
  '/(\\d+)/',
  function($match) use ($ten) { return (($match[0] + $ten)); },
  $string
  );
echo $newstring;
#prints "Some numbers: one: 11; two: 12; three: 13 end";
?>

A simple function to replace a list of complete words or terms in a string (for PHP 5.3 or above because of the closure):
<?php
function replace_words($list, $line, $callback) {
  return preg_replace_callback(
    '/(^|[^\\w\\-])(' . implode('|', array_map('preg_quote', $list)) . ')($|[^\\w\\-])/mi',
    function($v) use ($callback) { return $v[1] . $callback($v[2]) . $v[3]; },
    $line
  );
}
?>
Example of usage:
<?php
$list = array('php', 'apache web server');
$str = "php and the apache web server work fine together. php-gtk, for example, won't match. apache web servers shouldn't too.";
echo replace_words($list, $str, function($v) {
  return "<strong>{$v}</strong>";
});
?>

<?php
// Nice little function that convert a string to uppercase by keeping the HTMLentities intact.
public static function strtoupper_entities($str) {
  $patternMajEntities = '/(\&([A-Z])(ACUTE|CEDIL|CARON|CIRC|GRAVE|ORN|RING|SLASH|TH|TILDE|UML)\;)+/';
  $str = preg_replace_callback ($patternMajEntities, 
    function ($matches) {
      return "&" . $matches[2] . strtolower($matches[3]) . ";";
    }, strtoupper($str));
  
  return $str;
}

If you want to call non-static function inside your class, you can do something like this. 
For PHP 5.2 use second argument like array($this, 'replace'):
<?php
class test_preg_callback{
 private function process($text){
  $reg = "/\{([0-9a-zA-Z\- ]+)\:([0-9a-zA-Z\- ]+):?\}/";
  return preg_replace_callback($reg, array($this, 'replace'), $text);
 }
 
 private function replace($matches){
  if (method_exists($this, $matches[1])){
   return @$this->$matches[1]($matches[2]);   
  }
 } 
}
?>
For PHP 5.3 use second argument like "self::replace":
<?php
class test_preg_callback{
 private function process($text){
  $reg = "/\{([0-9a-zA-Z\- ]+)\:([0-9a-zA-Z\- ]+):?\}/";
  return preg_replace_callback($reg, "self::replace", $text);
 }
 
 private function replace($matches){
  if (method_exists($this, $matches[1])){
   return @$this->$matches[1]($matches[2]);   
  }
 } 
}
?>

preg_replace_callback returns NULL when pcre.backtrack_limit is reached; this sometimes occurs faster then you might expect. No error is raised either; so don't forget to check for NULL yourself

Text lines numeration:
<?PHP
// Multieline text:
  $Text = "
Some 
Multieline
text
for
numeration";
// For count:
  $GLOBALS["LineNUMBER"] = 1;
// Replace linestart on number:
  PRINT preg_replace_callback("/^/m",function ()
    {
      return $GLOBALS["LineNUMBER"]++." ";
    },
    $Text);
?>
1
2 Some
3 Multieline
4 text
5 for
6 numeration

Note that, as of PHP 5.3, it seems that named subpatterns are now included in the matches array by their named key as well as their numerical key.
To build off of Chris' previous example, as of PHP 5.3, you *can* do
<?php
preg_replace_callback('/(?<char>[a-z])/', 'callback', 'word');
function callback($matches) {
  var_dump($matches);
}
?>
and expect to get $matches['char'] in your function. * BUT ONLY AS OF PHP 5.3 *
Please be mindful of this if you intend to support PHP 5.2.

This small class allows PHP users to read JSON files with include statements in them. For instance the include {{{ "relative/to/including.json" }}} is replaced by the content of the json file located at "relative/to/including.json".
<?php
  /**
   * Handles JSON files with includes
   * Purpose: handle bigger JSON files by featuring "includes"
   * 
   * @author Florian Arndt
   */
  class JWI {
    /**
     * Parses a JSON file and returns its contents
     * @param String $filename
     */
    static function read($filename) {
      if(!file_exists($filename))
        throw new Exception('<b>JWI Error: JSON file <tt>'.$filename.'</tt> not found!</b>');
      $content = join('', file($filename));
      $dir = dirname($filename);
      /**
       * replace
       *  include statements
       * with
       *  content of the file to include
       * recursively
       */
      $content = preg_replace_callback(
        '/{{{\s*"\s*(.+)\s*"\s*}}}/', // >include file< - pattern
        create_function(
          '$matches', // callback parameter
          sprintf(
            '$fn = "%s/".$matches[1];'.
            'return JWI::read($fn);',
            realpath(dirname($filename))
          )
        ),
        $content
      );
      return $content;
    }
  }

Created this to fetch the link and name of an anchor tag. I use this when cleaning an HTML email to text. Using regex for HTML is not recommended but for this purpose I see no issue with it. This is not designed to work for nested anchors.
A note to keep in mind:
I was primarily concerned with valid HTML so if attributes do no use ' or " to contain the values then this will need to be tweaked.
If you can edit this to work better, please let me know.
<?php
/**
 * Replaces anchor tags with text
 * - Will search string and replace all anchor tags with text (case insensitive)
 * 
 * How it works:
 * - Searches string for an anchor tag, checks to make sure it matches the criteria
 *     Anchor search criteria:
 *       - 1 - <a (must have the start of the anchor tag )
 *       - 2 - Can have any number of spaces or other attributes before and after the href attribute
 *       - 3 - Must close the anchor tag
 * 
 * - Once the check has passed it will then replace the anchor tag with the string replacement
 * - The string replacement can be customized
 * 
 * Know issue:
 * - This will not work for anchors that do not use a ' or " to contain the attributes. 
 *     (i.e.- <a href=http: //php.net>PHP.net</a> will not be replaced)
 */
function replaceAnchorsWithText($data) {
  /**
   * Had to modify $regex so it could post to the site... so I broke it into 6 parts.
   */
  $regex = '/(<a\s*'; // Start of anchor tag
  $regex .= '(.*?)\s*'; // Any attributes or spaces that may or may not exist
  $regex .= 'href=[\'"]+?\s*(?P<link>\S+)\s*[\'"]+?'; // Grab the link
  $regex .= '\s*(.*?)\s*>\s*'; // Any attributes or spaces that may or may not exist before closing tag 
  $regex .= '(?P<name>\S+)'; // Grab the name
  $regex .= '\s*<\/a>)/i'; // Any number of spaces between the closing anchor tag (case insensitive)
  
  if (is_array($data)) {
    // This is what will replace the link (modify to you liking)
    $data = "{$data['name']}({$data['link']})";
  }
  return preg_replace_callback($regex, 'replaceAnchorsWithText', $data);
}
$input = 'Test 1: <a href="http: //php.net1">PHP.NET1</a>.<br />';
$input .= 'Test 2: <A name="test" HREF=\'HTTP: //PHP.NET2\' target="_blank">PHP.NET2</A>.<BR />';
$input .= 'Test 3: <a hRef=http: //php.net3>php.net3</a><br />';
$input .= 'This last line had nothing to do with any of this';
echo replaceAnchorsWithText($input).'<hr/>';
?>
Will output:
Test 1: PHP.NET1(http: //php.net1).
Test 2: PHP.NET2(HTTP: //PHP.NET2).
Test 3: php.net3 (is still an anchor)
This last line had nothing to do with any of this

From PHP 5.3 you can use an anonymous function to pass local variables into the callback.
<?php
public function replace_variables( $subject, $otherVars ) {
  $linkPatterns = array(
    '/(<a .*)href=(")([^"]*)"([^>]*)>/U',
    "/(<a .*)href=(')([^']*)'([^>]*)>/U"
  );
  $callback = function( $matches ) use ( $otherVars ) {
    $this->replace_callback($matches, $otherVars);
  };
  return preg_replace_callback($this->patterns, $callback, $subject);
}
public function replace_callback($matches, $otherVars) {
  return $matches[1] . $otherVars['myVar'];
}
?>

The good version of the class PhpHex2Str
<?php
class PhpHex2Str
{
  private $strings;
  private static function x_hex2str($hex) {
    $hex = substr($hex[0], 1);
    $str = '';
    for($i=0;$i < strlen($hex);$i+=2) {
      $str.=chr(hexdec(substr($hex,$i,2)));
    }
    return $str;
  }
  public function decode($strings = null) {
    $this->strings = (string) $strings;
    return preg_replace_callback('#\%[a-zA-Z0-9]{2}#', 'PhpHex2Str::x_hex2str', $this->strings);
  }
}
// Exemple
$obj = new PhpHex2Str;
$strings = $obj->decode($strings);
var_dump($strings);
?>

The pcre.backtrack_limit option (added in PHP 5.2) can trigger a NULL return, with no errors. The default pcre.backtrack_limit value is 100000. If you have a match that exceeds about half this limit it triggers a NULL response.
e.g. My limit was at 100000 but 500500 triggered a NULL response. I'm not running unicode but I *guess* PCRE runs in utf-16.

<?php
$format = <<<SQL
CREATE DATABASE IF NOT EXISTS :database;
GRANT ALL PRIVILEGES ON :database_name.* TO ':user'@':host';
SET PASSWORD = PASSWORD(':pass');
SQL;
$args = ["database"=>"people", "user"=>"staff", "pass"=>"pass123", "host"=>"localhost"];
preg_replace_callback("/:(\w+)/", function ($matches) use ($args) {
  return @$args[$matches[1]] ?: $matches[0];
}, $format);
/*
Result:
CREATE DATABASE IF NOT EXISTS people;
GRANT ALL PRIVILEGES ON :database_name.* TO 'staff'@'localhost';
SET PASSWORD = PASSWORD('pass123');
The `:database_name` placeholder doesn't exist as a matching key in `$args` so it's returned as is.
This way you know you need to correct the array by adding the "database_name" item.
*/

To use variable from the parent scope of the anonymous callback function passed to preg_replace_callback(), utilize the use() parameter.
$var1 = "one";
$var2 = "two";
$line = preg_replace_callback('/^.*$/',
      function( $matches ) use ( $var1, $var2 ) {
        return( $var1 . " " . $var2 );
      }, $line);
Will replace the entire string with the concatenated values or $var1 and $var2 ("one two") from the parent scope.

Be aware that as of php5.4 you MUST NOT pass variables by reference, as in '[, int &$count ]' - if you do it will result in a fatal error. 
I think the writer is trying to say that the function accepts the parameter by reference, but that is not how it reads. -Manual needs updating/clarifying ?

Decode Hexa to Strings =)
<?php
class PhpHex2Str
{
  private $strings;
  private function x_hex2str($hex) {
    $hex = substr($hex[0], 1);
    $str = '';
    for($i=0;$i < strlen($hex);$i+=2) {
      $str.=chr(hexdec(substr($hex,$i,2)));
    }
    return $str;
  }
  public function decode($strings = null) {
    $this->strings = (string) $strings;
    return preg_replace_callback('#\%[a-zA-Z0-9]{2}#', 'x_hex2str', $this->strings);
  }
}
// Example
$strings = 'a %20 b%0A h %27 h %23';
$obj = new PhpHex2Str;
$strings = $obj->decode($strings);
var_dump($strings);
?>

it is much better on preformance and better practice to use the preg_replace_callback function instead of preg_replace with the e modifier.
function a($text){return($text);}
// 2.76 seconds to run 50000 times
preg_replace("/\{(.*?)\}/e","a('\\1','\\2','\\3',\$b)",$a);
// 0.97 seconds to run 50000 times
preg_replace_callback("/\{(.*?)\}/s","a",$a);

<?php
/**
 * 将中文转为Html实体
 * Turning Chinese into Html entity
 * Author QiangGe
 * Mail 2962051004@qq.com
 *
*/
$str = <<<EOT
你好 world
EOT;
function ChineseToEntity($str) {
 return preg_replace_callback(
    '/[\x{4e00}-\x{9fa5}]/u', // utf-8 
    // '/[\x7f-\xff]+/', // if gb2312
    function ($matches) {
      $json = json_encode(array($matches[0]));
      preg_match('/\[\"(.*)\"\]/', $json, $arr);
      /*
       * 通过json_encode函数将中文转为unicode
       * 然后用正则取出unicode
       * Turn the Chinese into Unicode through the json_encode function, then extract Unicode from regular.
       * I think this idea is seamless.
      */
      return '&#x'. str_replace('\\u', '', $arr[1]). ';';
    }, $str
  );
}
echo ChineseToEntity($str);
// &#x4f60;&#x597d; world

Don't use this function to fetch BBCode, as explained. If you have some text that runs over 5000 chars (average), it will run out of its limit and makes you download the PHP page. 
According to this, you should instead use something more advanced yet complex. You will need a function called "str_replace_once()" (search for it), one called "countWord()", the famous "after()", "before()", "between()". 
str_replace_once does same as str_replace, but only replace first occurence. As for countWord, I guess you know how to count the number of a word occurence. As for after, before and between, this is a function that you may find easily somewhere on the site by a user. Else, you can do it. 
The following function is able to do all blocks, supposing [code] and [/code], you might wish things between parents dont get parsed, including [code] if inside of another [code]. 
<?php
function prepareCode($code, $op, $end)
{
  $ix = 0;
  $iy = 0;
  $nbr_Op = countWord($op, $code);
  while($ix < $nbr_Op)
  {
    if(in_string($op, before($end, $code), false))
    {
      // The following piece of code replace the default [tag] by [tag:#]
      $code = str_replace_once($op, substr($op, 0, -1).':'.$ix.']', $code); 
      $iy++;
    }
    elseif(in_string($end, before($op, $code), false))
    {
      $iy = $iy-1;
      $code = str_replace_once($end, substr($end, 0, -1).':'.($ix-1).']', $code);
      $ix = $ix-2;
    }
    $ix++;
  }
  while(in_string($end, $code))
  {
    $code = str_replace_once($end, substr($end, 0, -1).':'.($iy-1).']', $code);
    $iy=$iy-1;
  }
  $code = preg_replace('#\\'.substr($end, 0, 1).':-[0-9]\]#i', '', $code);
  if(in_string(substr($op, 0, -1).':0]', $code) && !in_string(substr($end, 0, -1).':0]', $code))
  {
    $code .= substr($end, 0, -1).":0]";
  }
  return $code;
}
?>
$code returns the whole text semi-formated. You only need to use it as :
$code = prepareCode($code="Your text", $op="[tag]" , $end="[/tag]");
Then just replace the parent tags :
str_replace("[tag:0]", "<tag>", $code);
str_replace("[/tag:0]", "</tag>", $code);
So at the end something like :
[

For the callback I advise only to use a permanent or anonymous function. 
Depending on the usage you may encounter memory issues when using create_function for the callback possibly due to attempts at being compatible with PHP 5.2 or prior. Some servers refuse to update their PHP for whatever reason.
Please peruse the create_function documentation for more details regarding its memory usage.
Regards.