How to search by key=>value in a multidimensional array in PHP

Question

Code:

function search($array, $key, $value)
{
    $results = array();

    if (is_array($array)) {
        if (isset($array[$key]) && $array[$key] == $value) {
            $results[] = $array;
        }

        foreach ($array as $subarray) {
            $results = array_merge($results, search($subarray, $key, $value));
        }
    }

    return $results;
}

$arr = array(0 => array(id=>1,name=>"cat 1"),
             1 => array(id=>2,name=>"cat 2"),
             2 => array(id=>3,name=>"cat 1"));

print_r(search($arr, 'name', 'cat 1'));

Output:

Array
(
    [0] => Array
        (
            [id] => 1
            [name] => cat 1
        )

    [1] => Array
        (
            [id] => 3
            [name] => cat 1
        )

)

If efficiency is important you could write it so all the recursive calls store their results in the same temporary $results array rather than merging arrays together, like so:

function search($array, $key, $value)
{
    $results = array();
    search_r($array, $key, $value, $results);
    return $results;
}

function search_r($array, $key, $value, &$results)
{
    if (!is_array($array)) {
        return;
    }

    if (isset($array[$key]) && $array[$key] == $value) {
        $results[] = $array;
    }

    foreach ($array as $subarray) {
        search_r($subarray, $key, $value, $results);
    }
}

The key there is that search_r takes its fourth parameter by reference rather than by value; the ampersand & is crucial.

FYI: If you have an older version of PHP then you have to specify the pass-by-reference part in the call to search_r rather than in its declaration. That is, the last line becomes search_r($subarray, $key, $value, &$results).

How about the SPL version instead? It'll save you some typing:

// I changed your input example to make it harder and
// to show it works at lower depths:

$arr = array(0 => array('id'=>1,'name'=>"cat 1"),
             1 => array(array('id'=>3,'name'=>"cat 1")),
             2 => array('id'=>2,'name'=>"cat 2")
);

//here's the code:

    $arrIt = new RecursiveIteratorIterator(new RecursiveArrayIterator($arr));

 foreach ($arrIt as $sub) {
    $subArray = $arrIt->getSubIterator();
    if ($subArray['name'] === 'cat 1') {
        $outputArray[] = iterator_to_array($subArray);
    }
}

What's great is that basically the same code will iterate through a directory for you, by using a RecursiveDirectoryIterator instead of a RecursiveArrayIterator. SPL is the roxor.

The only bummer about SPL is that it's badly documented on the web. But several PHP books go into some useful detail, particularly Pro PHP; and you can probably google for more info, too.

<?php
$arr = array(0 => array("id"=>1,"name"=>"cat 1"),
             1 => array("id"=>2,"name"=>"cat 2"),
             2 => array("id"=>3,"name"=>"cat 1")
);
$arr = array_filter($arr, function($ar) {
   return ($ar['name'] == 'cat 1');
   //return ($ar['name'] == 'cat 1' AND $ar['id'] == '3');// you can add multiple conditions
});

echo "<pre>";
print_r($arr);

?>

Ref: http://php.net/manual/en/function.array-filter.php

Came back to post this update for anyone needing an optimisation tip on these answers, particulary John Kugelman's great answer up above.

His posted function work fine but I had to optimize this scenario for handling a 12 000 row resultset. The function was taking an eternal 8 secs to go through all records, waaaaaay too long.

I simply needed the function to STOP searching and return when match was found. Ie, if searching for a customer_id, we know we only have one in the resultset and once we find the customer_id in the multidimensional array, we want to return.

Here is the speed-optimised ( and much simplified ) version of this function, for anyone in need. Unlike other version, it can only handle only one depth of array, does not recurse and does away with merging multiple results.

// search array for specific key = value
public function searchSubArray(Array $array, $key, $value) {   
    foreach ($array as $subarray){  
        if (isset($subarray[$key]) && $subarray[$key] == $value)
          return $subarray;       
    } 
}

This brought down the the task to match the 12 000 records to a 1.5 secs. Still very costly but much more reasonable.

if (isset($array[$key]) && $array[$key] == $value)

A minor imporvement to the fast version.

Be careful of linear search algorithms (the above are linear) in multiple dimensional arrays as they have compounded complexity as its depth increases the number of iterations required to traverse the entire array. Eg:

array(
    [0] => array ([0] => something, [1] => something_else))
    ...
    [100] => array ([0] => something100, [1] => something_else100))
)

would take at the most 200 iterations to find what you are looking for (if the needle were at [100][1]), with a suitable algorithm.

Linear algorithms in this case perform at O(n) (order total number of elements in entire array), this is poor, a million entries (eg a 1000x100x10 array) would take on average 500,000 iterations to find the needle. Also what would happen if you decided to change the structure of your multidimensional array? And PHP would kick out a recursive algorithm if your depth was more than 100. Computer science can do better:

Where possible, always use objects instead of multiple dimensional arrays:

ArrayObject(
   MyObject(something, something_else))
   ...
   MyObject(something100, something_else100))
)

and apply a custom comparator interface and function to sort and find them:

interface Comparable {
   public function compareTo(Comparable $o);
}

class MyObject implements Comparable {
   public function compareTo(Comparable $o){
      ...
   }
}

function myComp(Comparable $a, Comparable $b){
    return $a->compareTo($b);
}

You can use uasort() to utilize a custom comparator, if you're feeling adventurous you should implement your own collections for your objects that can sort and manage them (I always extend ArrayObject to include a search function at the very least).

$arrayObj->uasort("myComp");

Once they are sorted (uasort is O(n log n), which is as good as it gets over arbitrary data), binary search can do the operation in O(log n) time, ie a million entries only takes ~20 iterations to search. As far as I am aware custom comparator binary search is not implemented in PHP (array_search() uses natural ordering which works on object references not their properties), you would have to implement this your self like I do.

This approach is more efficient (there is no longer a depth) and more importantly universal (assuming you enforce comparability using interfaces) since objects define how they are sorted, so you can recycle the code infinitely. Much better =)

Here is solution:

<?php
$students['e1003']['birthplace'] = ("Mandaluyong <br>");
$students['ter1003']['birthplace'] = ("San Juan <br>");
$students['fgg1003']['birthplace'] = ("Quezon City <br>");
$students['bdf1003']['birthplace'] = ("Manila <br>");

$key = array_search('Delata Jona', array_column($students, 'name'));
echo $key;  

?>