Unwanted behavior of PHP usort - php

Question

I have a problem with PHP usort(). Let's suppose I have an array like this (that's a simplification, I've not to work with names, and I have an array of objects, not arrays):
$data = array(
array('name' => 'Albert', 'last' => 'Einstein'),
array('name' => 'Lieserl', 'last' => 'Einstein'),
array('name' => 'Alan', 'last' => 'Turing' ),
array('name' => 'Mileva', 'last' => 'Einstein'),
array('name' => 'Hans Albert', 'last' => 'Einstein')
);
As you can see, the array is sorted arbitrarily.
Now, if want to sort it by last, I do:
function sort_some_people($a, $b) { return strcmp($a['last'], $b['last']); }
usort($data, 'sort_some_people');
And I have:
Array (
[0] => Array ( [name] => Mileva [last] => Einstein )
[3] => Array ( [name] => Albert [last] => Einstein )
[1] => Array ( [name] => Lieserl [last] => Einstein )
[2] => Array ( [name] => Hans Albert [last] => Einstein )
[4] => Array ( [name] => Alan [last] => Turing )
)
That is ok, now they are sorted by last. But, as you can see, I've just completely lost the previous sorting. What am I saying? I want to preserve array sorting as it was before, but as a secondary sorting. I hope I was clear. Practically I want to sort data using something as usort() (so, a completely custom sorting), but if sorting field is identical between two items, I want to keep their relative position as it was before. Considering the given example, I want Lieserl Einstein to appear before Mileva Einstein because it was like this at beginning.

Answer 1

The sorting algorithms used in PHP have this property where the order is undefined if the items match.
If you need to keep the order, then you have to roll your own code.
Fortunately some one already has:
http://www.php.net/manual/en/function.usort.php#38827
$data = array(
array('name' => 'Albert', 'last' => 'Einstein'),
array('name' => 'Lieserl', 'last' => 'Einstein'),
array('name' => 'Alan', 'last' => 'Turing' ),
array('name' => 'Mileva', 'last' => 'Einstein'),
array('name' => 'Hans Albert', 'last' => 'Einstein')
);
function sort_some_people($a, $b) {
return strcmp($a['last'], $b['last']);
}
function mergesort(&$array, $cmp_function = 'strcmp') {
// Arrays of size < 2 require no action.
if (count($array) < 2) return;
// Split the array in half
$halfway = count($array) / 2;
$array1 = array_slice($array, 0, $halfway);
$array2 = array_slice($array, $halfway);
// Recurse to sort the two halves
mergesort($array1, $cmp_function);
mergesort($array2, $cmp_function);
// If all of $array1 is <= all of $array2, just append them.
if (call_user_func($cmp_function, end($array1), $array2[0]) < 1) {
$array = array_merge($array1, $array2);
return;
}
// Merge the two sorted arrays into a single sorted array
$array = array();
$ptr1 = $ptr2 = 0;
while ($ptr1 < count($array1) && $ptr2 < count($array2)) {
if (call_user_func($cmp_function, $array1[$ptr1], $array2[$ptr2]) < 1) {
$array[] = $array1[$ptr1++];
}
else {
$array[] = $array2[$ptr2++];
}
}
// Merge the remainder
while ($ptr1 < count($array1)) $array[] = $array1[$ptr1++];
while ($ptr2 < count($array2)) $array[] = $array2[$ptr2++];
return;
}
mergesort($data, 'sort_some_people');
print_r($data);
Output:
Array
(
[0] => Array
(
[name] => Albert
[last] => Einstein
)
[1] => Array
(
[name] => Lieserl
[last] => Einstein
)
[2] => Array
(
[name] => Mileva
[last] => Einstein
)
[3] => Array
(
[name] => Hans Albert
[last] => Einstein
)
[4] => Array
(
[name] => Alan
[last] => Turing
)
)
Voila!

Answer 2

$compare = strcmp($a['last'], $b['last']);
if ($compare == 0)
$compare = strcmp($a['name'], $b['name']);
return $compare

Answer 3

You're asking for a stable sorting algorithm, which php doesn't offer. Beware that even if it may appear stable sometimes, there's no guarantee of it, and so likely to misbehave given certain inputs.
If you know the sorting criteria of the other columns, you can resort it all at once to get the desired behavior. array_multisort does this. The following is a bit more powerful way in that the comparison logic is completely user defined.
// should behave similar to sql "order by last, first"
$comparatorSequence = array(
function($a, $b) {
return strcmp($a['last'], $b['last']);
}
, function($a, $b) {
return strcmp($a['first'], $b['first']);
}
// more functions as needed
);
usort($theArray, function($a, $b) use ($comparatorSequence) {
foreach ($comparatorSequence as $cmpFn) {
$diff = call_user_func($cmpFn, $a, $b);
if ($diff !== 0) {
return $diff;
}
}
return 0;
});
If you really need a stable sort because the existing order of elements isnt well defined, try writing your own sort. bubble sort for example is very easy to write. This is a good solution so long as the number of elements in your lists are relatively small, otherwise look to implement one of the other stable sorting algorithms.

Answer 4

Try this:
function sort_some_people($a, $b)
{
$compareValue = 10 * strcmp($a['last'], $b['last']);
$compareValue += 1 * strcmp($a['name'], $b['name']);
return $compareValue;
}
Sample: http://codepad.org/zkHviVBM
This functions uses each digit of the decimal system for a sort criteria. The one that comes first has the highest digit. Might not be the smartest, but works for me.