PHP: Expose 'get' and 'set' for object with nested associative arrays

I have a class which stores values with a multi-level associative array:

I need to add a way to access and modify nested values. Here is a working solution for my problem, but it is rather slow. Is there a better way of doing this?

Note: The use of get / set functions is not mandatory, but there needs to be an efficient way to define a default value.

class Demo {
    protected $_values = array();

    function __construct(array $values) {
        $this->_values = $values;
    }

    public function get($name, $default = null) {
        $token = strtok($name, '.#');
        $node = $this->_values;
        while ($token !== false) {
            if (!isset($node[$token]))
                return $default;
            $node = $node[$tok开发者_开发知识库en];
            $token = strtok('.#');
        }
        return $node;
    }

    public function set($name, $value) {
        $next_token = strtok($name, '.#');
        $node = &$this->_values;

        while ($next_token !== false) {
            $token = $next_token;
            $next_token = strtok('.#');

            if ($next_token === false) {
                $node[ $token ] = $value;
                break;
            }
            else if (!isset($node[ $token ]))
                $node[ $token ] = array();

            $node = &$node[ $token ];
        }

        unset($node);
    }

}

Which would be used as follows:

$test = new Demo(array(
    'simple'  => 27,
    'general' => array(
        0 => array(
            'something'    => 'Hello World!',
            'message'      => 'Another message',
            'special'      => array(
                'number'       => 27
            )
        ),
        1 => array(
            'something'    => 'Hello World! #2',
            'message'      => 'Another message #2'
        ),
    )
));

$simple = $test->get('simple'); // === 27

$general_0_something = $test->get('general#0.something'); // === 'Hello World!'

$general_0_special_number = $test->get('general#0.special.number'); === 27

Note: 'general.0.something' is the same as 'general#0.something', the alternative punctuation is for the purpose of clarity.

---

Well, the question was interesting enough that I couldn't resist tinkering a bit more. :-)

So, here are my conclusions. Your implementation is probably the most straightforward and clear. And it's working, so I wouldn't really bother about searching for another solution. In fact, how much calls are you gonna get in the end? Is the difference in performance worth the trouble (I mean between "super ultra blazingly fast" and "almost half as fast")?

Put aside though, if performance is really an issue (getting thousands of calls), then there's a way to reduce the execution time if you repetitively lookup the array.

In your version the greatest burden falls on string operations in your get function. Everything that touches string manipulation is doomed to fail in this context. And that was indeed the case with all my initial attempts at solving this problem.

It's hard not to touch strings if we want such a syntax, but we can at least limit how much string operations we do.

If you create a hash map (hash table) so that you can flatten your multidimensional array to a one level deep structure, then most of the computations done are a one time expense. It pays off, because this way you can almost directly lookup your values by the string provided in your get call.

I've come up with something roughly like this:

<?php

class Demo {
    protected $_values = array();
    protected $_valuesByHash = array();

    function createHashMap(&$array, $path = null) {
        foreach ($array as $key => &$value) {
            if (is_array($value)) {
                $this->createHashMap($value, $path.$key.'.');
            } else {
                $this->_valuesByHash[$path.$key] =& $value;
            }
        }
    }

    function __construct(array $values) {
        $this->_values = $values;
        $this->createHashMap($this->_values);

        // Check that references indeed work
        // $this->_values['general'][0]['special']['number'] = 28;
        // print_r($this->_values);
        // print_r($this->_valuesByHash);
        // $this->_valuesByHash['general.0.special.number'] = 29;
        // print_r($this->_values);
        // print_r($this->_valuesByHash);
    }

    public function get($hash, $default = null) {
        return isset($this->_valuesByHash[$hash]) ? $this->_valuesByHash[$hash] : $default;
    }
}


$test = new Demo(array(
    'simple'  => 27,
    'general' => array(
        '0' => array(
            'something'    => 'Hello World!',
            'message'      => 'Another message',
            'special'      => array(
                'number'       => 27
            )
        ),
        '1' => array(
            'something'    => 'Hello World! #2',
            'message'      => 'Another message #2'
        ),
    )
));

$start = microtime(true);

for ($i = 0; $i < 10000; ++$i) {
    $simple = $test->get('simple', 'default');
    $general_0_something = $test->get('general.0.something', 'default');
    $general_0_special_number = $test->get('general.0.special.number', 'default');
}

$stop = microtime(true);

echo $stop-$start;

?>

The setter is not yet implemented, and you would have to modify it for alternative syntax (# separator), but I think it conveys the idea.

At least on my testbed it takes half the time to execute this compared to the original implementation. Still raw array access is faster, but the difference in my case is around 30-40%. At the moment that was the best I could achieve. I hope that your actual case is not big enough that I've hit some memory constraints on the way. :-)

---

Ok, my first approached missed the goal I was aiming for. Here is the solution to using native PHP array syntax (at least for access) and still being able to set a default value.

Update: Added missing functionality for get/set and on the fly converting.

By the way, this is not an approach to take if you are optimizing for performance. This is perhaps 20 times slower than regular array access.

class Demo extends ArrayObject {
    protected $_default;
    public function __construct($array,$default = null) {
        parent::__construct($array);
        $this->_default = $default;
    }
    public function  offsetGet($index) {
        if (!parent::offsetExists($index)) return $this->_default;
        $ret = parent::offsetGet($index);
        if ($ret && is_array($ret)) {
            parent::offsetSet($index, $this->newObject($ret));
            return parent::offsetGet($index);
        }
        return $ret;
    }
    protected function newObject(array $array=null) {
        return new self($array,$this->_default);
    }
}

Init

$test = new Demo(array(
    'general' => array(
        0 => array(
            'something'    => 'Hello World!'
        )
    )
),'Default Value');

Result

$something = $test['general'][0]['something']; // 'Hello World!'
$notfound = $test['general'][0]['notfound']; // 'Default Value'

---

You're looking for something like that? Essentially the get() method uses references to descend into the $values array and breaks out of the method if a requirement could not be met.

class Demo {
    protected $_values = array();

    public function __construct(array $values) {
        $this->_values = $values;
    }

    public function get($name, $default = null) {
        $parts  = preg_split('/[#.]/', $name);
        if (!is_array($parts) || empty($parts)) {
            return null;
        }

        $value  = &$this->_values;
        foreach ($parts as $p) {
            if (array_key_exists($p, $value)) {
                $value  = &$value[$p];
            } else {
                return null;
            }
        }

        return $value;
    }

    /**
     * setter missing
     */
}

$test = new Demo(array(
    'simple'  => 2,
    'general' => array(
        0 => array(
                'something'    => 'Hello World!',
                'message'      => 'Another message',
                'special'      => array(
                    'number'       => 4
                )
            ),
        1 => array(
                'something'    => 'Hello World! #2',
                'message'      => 'Another message #2'
            )
    )
));

$v = $test->get('simple'); 
var_dump($v);

$v = $test->get('general'); 
var_dump($v);

$v = $test->get('general.0'); 
var_dump($v);

$v = $test->get('general#0'); 
var_dump($v);

$v = $test->get('general.0.something'); 
var_dump($v);

$v = $test->get('general#0.something'); 
var_dump($v);

$v = $test->get('general.0.message'); 
var_dump($v);

$v = $test->get('general#0.message'); 
var_dump($v);

$v = $test->get('general.0.special'); 
var_dump($v);

$v = $test->get('general#0.special'); 
var_dump($v);

$v = $test->get('general.0.special.number'); 
var_dump($v);

$v = $test->get('general#0.special.number'); 
var_dump($v);

$v = $test->get('general.1'); 
var_dump($v);

$v = $test->get('general#1'); 
var_dump($v);

$v = $test->get('general.1.something'); 
var_dump($v);

$v = $test->get('general#1.something'); 
var_dump($v);

$v = $test->get('general.1.message'); 
var_dump($v);

$v = $test->get('general#1.message'); 
var_dump($v);

---

This is how multidimensional array work in general in PHP:

$data = array(
    'general' => array(
         0 => array(
             'something'    => 'Hello World!'
         )
    )
);

To receive Hello World:

echo $data['general'][0]['something'];