README

PHP Serialized string array and scalar unserialization using pure PHP.

Runtime PHP 7.3+ (aligned with PHPUnit 9 for development). The composer.lock is resolved for PHP 7.3 via config.platform so installs stay consistent on older runtimes.

Usage

Install with composer:

composer require lyte/serial

Serial

Serial is a simplified interface that attempts to work well in a legacy code base.

Load the namespace:

use Lyte\Serial\Serial;
// unserialize statically
$unserialized = Serial::unserialize($someSerializedString);
// or with an instance
$serial = new Serial;
$unserialized = $serial->unserialize($someSerializedString);

// check if a string appears to be serialized
if (Serial::isSerialized($someUnknownString)) {
	$unserialized = Serial::unserialize($someUnknownString);
}

// or rely on exceptions
try {
	$unserialized = Serial::unserialize($someUnknownString);
} catch (\Exception $e) {
	// ...
}

Unserializer

Unserializer is the internal work horse.

use Lyte\Serial\Unserializer;
$serial = new Unserializer($someSerializedString);
$unserialized = $serial->unserialize();

Wrong s: byte lengths (nested payloads)

If a string declares s:N: but the next " is not exactly N bytes later (common when counts used UTF-8 lengths, wrong newlines, or copy-paste errors), strict parsing fails. Enable inference and optional wire repair:

$u = new Unserializer($blob, ['repairLengths' => true]);
$data = $u->unserialize();
// $u->lengthIssues lists declared vs actual lengths and digit span offsets

list($fixedBlob, $issues) = Unserializer::repairSerializedString($blob);
// or Serial::repairSerializedString($blob)

repairLengths is read with filter_var(..., FILTER_VALIDATE_BOOLEAN, FILTER_NULL_ON_FAILURE) and must coerce to strict true (e.g. true, 1, "1", "true", "on", "yes"). Values such as false, 0, "0", "false", "no", "", null, or unset leave repair off (so string "false" from env/JSON does not accidentally enable it).

For values that look like nested PHP serialization (a:…, s:…, etc.), the actual length is derived by parsing one full value. For other strings, repair uses the first "; sequence with no " inside the payload (unsafe for arbitrary binary that may contain those bytes).

Why?

The standard serialize() and unserialize() calls in PHP are known to be unsafe even if you use the $allowed_classes filter in PHP 7 (there are memory corruption bugs).

The standard answer to this is "use JSON" but some applications were using PHP serialized strings for internal storage long before JSON was a thing (well... popular).

In this case it may be useful to have a safer parser that rejects anything that's not an array or scalar type (i.e what you could safely store in JSON) as a middle ground to harden a code base without having to immediately switch out the underlying storage format.

Note: I'm not advocating letting any strings be unserialized that can in anyway be modified by a user, just that if you use a safer parser and someone compromises some other part of your application this might at least slow them down.

Why can't I use $allowed_classes?

PHP 7 added the $allowed_classes option to the unserialize() function.

In theory you could just set this to null (or a safe set of classes), but unfortunately there's memory corruption bugs meaning if you rely on that behaviour, you are vulnerable.