php 'preg_match_all' and 'str_replace': regular expression to replace constants with array keys

If you want to wrap any valid identifiers inside square brackets, you can use preg_replace directly:

$regex = '/(["\'])(?:(?=(\\\\?))\2.)*?\1(*SKIP)(*F)|\[([a-zA-Z_\x7f-\xff][a-zA-Z0-9_\x7f-\xff]*)]/s';
$ouptut = preg_replace($regex, '$3', $text);

See the regex demo. Details:

• (["'])(?:(?=(\\?))\2.)*?\1 - matches a string between single or double quotation marks (contains two capturing groups)
• (*SKIP)(*F) - discards the matched text and fails the match starting a new search from the failure location
• | - or
• \[ - [ char
• ([a-zA-Z_\x7f-\xff][a-zA-Z0-9_\x7f-\xff]*) - Group 3: a letter, underscore, or any char from the \x7f-\xff range and then any alphanumeric, underscore or any char from the \x7f-\xff range
• ] - a ] char.

See the PHP demo:

$regex = '/(["\'])(?:(?=(\\\\?))\2.)*?\1(*SKIP)(*F)|\[([a-zA-Z_\x7f-\xff][a-zA-Z0-9_\x7f-\xff]*)]/s';
$str = '$output = "<input name=\'variable[key]\' has to be preserved as it is.";
$output = \'Even this string variable[key] has to be preserved as it is.\';

$variable[key] = "WhatElse";
$result = $wso->RSLA("7050", $vegalot, "600", "WFID_OK_WEB","1300", $_POST[username]);
if ($result[ECD] != 0) {
if ($line=="AAAA" && in_array(substr($wso->lot,0,7),$lot_aaaa_list) && $lot[wafer][25]) {';
echo preg_replace($regex, "['\$3']", $str);

Output:

$output = "<input name='variable[key]' has to be preserved as it is.";
$output = 'Even this string variable[key] has to be preserved as it is.';

$variable['key'] = "WhatElse";
$result = $wso->RSLA("7050", $vegalot, "600", "WFID_OK_WEB","1300", $_POST['username']);
if ($result['ECD'] != 0) {
if ($line=="AAAA" && in_array(substr($wso->lot,0,7),$lot_aaaa_list) && $lot['wafer'][25]) {

[I know this isn't regexp, but since you asked for 'suggestion to better solve such issue' I give you my 2 cents]

How about simply parsing the code ;):

$source = file_get_contents('/tmp/test.php'); // Change this
$tokens = token_get_all($source);

$history = [];
foreach ($tokens as $token) {
    if (is_string($token)) { // simple 1-character token       
        array_push($history, str_replace(["\r", "\n"], '', $token));
        $history = array_slice($history, -2);

        echo $token;
    } else {
        list($id, $text) = $token;

        switch ($id) {
            case T_STRING:
                if ($history == [T_VARIABLE, '[']) {
                    // Token sequence is [T_VARIABLE, '[', T_STRING]
                    echo "'$text'";
                }
                else {
                    echo $text;
                }
                break;

            default:
                // anything else -> output "as is"
                echo $text;
                break;
        }

        array_push($history, $id);
        $history = array_slice($history, -2);
    }
}

Of course, the $source needs to be changed to whatever suits you. token_get_all() then loads the PHP code and parses it into a list of tokens. That list is then processed item by item and possibly changed before being output again, according to our needs.

1-char tokens like [ ("[" and "]" in f.ex $myVariable[1] both get to be tokens) are a special case which has to be handled in the loop. Otherwise $token is an array with an ID for the type of token and the token itself.

"Unfortunately" T_STRING is kind of a general case, so to pinpoint only the strings being used as constants in array indexing we store the 2 items preceding the current in $history. ("$myVariable" and "[")

..and..that's it, really. The code is read from a file, processed and output to stdout. Everything but the "constants as array index" case should be output as is.

If you like I can rewrite it as a function or something. The above should be kind of the general solution, though.

Edit Version 2, support for $myObject->myProp[key]:

<?php
$source = file_get_contents('/tmp/test.php'); // Change this
$tokens = token_get_all($source);

//print_r($tokens); exit();

$history = [];
foreach ($tokens as $token) {
    if (is_string($token)) { // simple 1-character token       
        array_push($history, str_replace(["\r", "\n"], '', $token));

        echo $token;
    } else {
        list($id, $text) = $token;

        switch ($id) {
            case T_STRING:
                if (array_slice($history, -2) == [T_VARIABLE, '[']) {
                    // Token sequence is [T_VARIABLE, '[', T_STRING]
                    echo "'$text'";
                }
                else if (array_slice($history, -4) == [T_VARIABLE, T_OBJECT_OPERATOR, T_STRING, '[']) {
                    echo "'$text'";
                }
                else {
                    echo $text;
                }
                break;

            default:
                // anything else -> output "as is"
                echo $text;
                break;
        }

        array_push($history, $id);
    }

    // This has to be at least as large as the largest chunk
    // checked anywhere above
    $history = array_slice($history, -5); 
}

As can be seen, the tough part about introducing more cases is that $history won't be as uniform anymore. At first I thought about fetching things directly from $tokens, but they aren't sanitized so I stuck to $history. It's possible that the "pruning line" at the bottom isn't needed, it's just there for memory usage. Maybe it's cleaner to skip $history, sanitize all $tokens items before the foreach() and then fetch things directly from it(adding the index to the foreach(), of course). I think I feel version 3 coming up ;-j..

Edit Version 3: This should be as simple as it gets. Simply look for brackets with unquoted strings inside.

$source = file_get_contents('/tmp/test.php'); // Change this
$tokens = token_get_all($source);

$history = [];
foreach ($tokens as $token) {
    if (is_string($token)) { // simple 1-character token       
        array_push($history, str_replace(["\r", "\n"], '', $token));

        echo $token;
    } else {
        list($id, $text) = $token;

        switch ($id) {
            case T_STRING:
                if (array_slice($history, -1) == ['[']) {
                    echo "'$text'";
                }
                else {
                    echo $text;
                }
                break;

            default:
                // anything else -> output "as is"
                echo $text;
                break;
        }

        array_push($history, $id);
    }
}

Test input(/tmp/test.php):

<?php
$variable[key] = "WhatElse";
$result = $wso->RSLA("7050", $vegalot, "600", "WFID_OK_WEB","1300", $_POST[username]);
if ($result[ECD] != 0) {
if ($line=="AAAA" && in_array(substr($wso->lot,0,7),$lot_aaaa_list) && $lot[wafer][25]) {
    $object->method[key];

    $variable[test] = 'one';
    $variable[one][two] = 'three';
    $variable->property[three]['four'] = 5;

I solved the issue with a double loop:

<?
$script = file("/tmp/test.php");
/*
 * Search any row containing a php variable $(...)
 */
$scanstr = preg_grep("/\\$([\w\-\>]+)(\[.*\])/", $script);
foreach($scanstr as $k => $str) {
   unset($matchvar, $match);
   /* Get php variable */
   preg_match_all('/\$(\w|-|>|\[|\]|\'|")+/', $str, $matchvar, PREG_SET_ORDER, 0);
   /* Get array key name */
   preg_match_all('/(\[(\w+)\])/', $matchvar[0][0], $match, PREG_SET_ORDER, 0);
   $replace_str = $str;
   foreach($match as $m) {
      /*
       * if key is not defined or a number, then skip conversion
       */
      if (!$m[2]) continue;
      if (is_numeric($m[2])) continue;
      $r = str_replace($m[1], "['" . $m[2] . "']", $matchvar[0][0]);
      $replace_str = str_replace($matchvar[0][0], $r, $replace_str);
      $matchvar[0][0] = $r;
      $local_changeflag = true;
    }
  ?>

It works for any of the following cases:

 $variable[test] = 'one';
 $variable[one][two] = 'three';
 $variable->property[three]['four'] = 5;

I know, it's not very clean ;)