PHP: Sorting Arrays of Arrays

PHP provides a range of functions for sorting data based on the key or value of an associative array. Where it gets complicated is when you have an array of associative arrays which you need to sort according to one or more attributes.

This is a very common task for PHP programmers as data returned from database/SQL queries often appears in this format (ref: pg_fetch_assoc and mysql_fetch_assoc).

Note: The array-multisort function provides an equivalent functionality to the code presented here, but is rather convoluted to use.

The Data

The examples below are based on the following data as an associative array. The data contains six records, each of which has a value for "firstname", "lastname" and for "age".

<?PHP $data = array( array("firstname" => "Mary", "lastname" => "Johnson", "age" => 25), array("firstname" => "Amanda", "lastname" => "Miller", "age" => 18), array("firstname" => "James", "lastname" => "Brown", "age" => 31), array("firstname" => "Patricia", "lastname" => "Williams", "age" => 7), array("firstname" => "Michael", "lastname" => "Davis", "age" => 43), array("firstname" => "Sarah", "lastname" => "Miller", "age" => 24), array("firstname" => "Patrick", "lastname" => "Miller", "age" => 27) ); ?>

This dataset can be represented in tabular format as:

	firstname	lastname	age
[0]	Mary	Johnson	25
[1]	Amanda	Miller	18
[2]	James	Brown	31
[3]	Patricia	Williams	7
[4]	Michael	Davis	43
[5]	Sarah	Miller	24
[6]	Patrick	Miller	27

Let's start sorting!

Sorting on a single field

It turns out that sorting associative arrays is really quite simple - IF you know beforehand which field you want to sort by.

For example, to sort by lastname you can use the following function:

<?PHP function compare_lastname($a, $b) { return strnatcmp($a['lastname'], $b['lastname']); } // sort alphabetically by name usort($data, 'compare_lastname'); ?>

The output is as follows:

	firstname	lastname	age
[0]	James	Brown	31
[1]	Michael	Davis	43
[2]	Mary	Johnson	25
[3]	Sarah	Miller	24
[4]	Patrick	Miller	27
[5]	Amanda	Miller	18
[6]	Patricia	Williams	7

You should be able quite easily to modify the compare_lastname function to order by firstname (see below), by age, or any other field you happen to have in the associative array. The strnatcmp function is particularly handy as it can be applied to numbers as well as strings.

To maintain index association, replace usort with uasort in the code:

	firstname	lastname	age
[2]	James	Brown	31
[4]	Michael	Davis	43
[0]	Mary	Johnson	25
[5]	Sarah	Miller	24
[6]	Patrick	Miller	27
[1]	Amanda	Miller	18
[3]	Patricia	Williams	7

This is useful when the 'index' is a database id that needs to be associated with each record.

Sorting on multiple fields

If you want to sort by lastname AND firstname, then you might think that just applying two sorts in sequence would give you the desired result:

<?PHP function compare_firstname($a, $b) { return strnatcmp($a['firstname'], $b['firstname']); } // sort alphabetically by firstname, then by lastname usort($data, 'compare_firstname'); usort($data, 'compare_lastname'); ?>

You might think so, but you'd be wrong. Using the usort function this isn't the case. As the PHP manual says, since version 4.1.0 the sort algorithm is no longer stable (see Quick Sort Algorithm for more detail):

If two members compare as equal, their order in the sorted array is undefined

This means that, after ordering by first name, the result of then sorting by last name will not necessarily retain the correct order of first names.

So what's the solution? We need a new function:

<?PHP function compare_fullname($a, $b) { $retval = strnatcmp($a['lastname'], $b['lastname']); if(!$retval) return strnatcmp($a['firstname'], $b['firstname']); return $retval; } // sort alphabetically by firstname and lastname usort($data, 'compare_fullname'); ?>

Finally, the result we wanted:

	firstname	lastname	age
[0]	James	Brown	31
[1]	Michael	Davis	43
[2]	Mary	Johnson	25
[3]	Amanda	Miller	18
[4]	Patrick	Miller	27
[5]	Sarah	Miller	24
[6]	Patricia	Williams	7

This function works because it first compares the last names, and only if they are identical will it compare the first names to work out the correct order.

Using an 'anonymous' function

This is where it starts to get interesting. You can create an anonymous function based on information gathered at run time. This allows for greater flexibility, without too much extra coding.

<?PHP function makeSortFunction($field) { $code = "return strnatcmp(\$a['$field'], \$b['$field']);"; return create_function('$a,$b', $code); } $compare = makeSortFunction('age'); usort($data, $compare); ?>

The call can then be shortened to:

<?PHP usort($data, makeSortFunction('age')); ?>

Finally, you can wrap it all into a single function as shown below. Now the code's starting to look more 'readable' which is a good sign.

<?PHP function orderBy($data, $field) { $code = "return strnatcmp(\$a['$field'], \$b['$field']);"; usort($data, create_function('$a,$b', $code)); return $data; } $data = orderBy($data, 'age'); ?>

	firstname	lastname	age
[0]	Patricia	Williams	7
[1]	Amanda	Miller	18
[2]	Sarah	Miller	24
[3]	Mary	Johnson	25
[4]	Patrick	Miller	27
[5]	James	Brown	31
[6]	Michael	Davis	43

Using 'pass by reference'

We can also pass the $data variable by reference and no longer worry about returning a value from the function. This is how built-in functions such as usort work (see the link below for more details from the PHP website).

<?PHP function orderBy(&$data, $field) { $code = "return strnatcmp(\$a['$field'], \$b['$field']);"; usort($data, create_function('$a,$b', $code)); } orderBy($data, 'age'); ?>

Simply adding the & to the function definition means that instead of passing a copy of $data to the function, a reference or pointer is passed. That means that any manipulation of the array within the function will also apply outside the function so no return command is required.

The function is otherwise equivalent to the 'pass by value' version.

Sorting based on a list of values

There are situations where it's not possible to use a numeric or alphabet-based sort routine. Suppose that you have the following data to present on a website:

	name	position
[0]	Mary Johnson	Secretary
[1]	Amanda Miller	Member
[2]	James Brown	Member
[3]	Patricia Williams	Member
[4]	Michael Davis	President
[5]	Sarah Miller	Vice-President
[6]	Patrick Miller	Member

And that you want to sort these names by their position, based on the following list:

<?PHP $sortorder = array( 'President', 'Vice-President', 'Secretary', 'Member' ); ?>

Again, it's not as difficult as you might think. All we need is the following custom function:

<?PHP function mySort($a, $b) { global $sortorder; if($a['position'] == $b['position']) { return strcmp($a['name'], $b['name']); } $cmpa = array_search($a['position'], $sortorder); $cmpb = array_search($b['position'], $sortorder); return ($cmpa > $cmpb) ? 1 : -1; } ?>

Note: If you don't have a second field that you want to sort by, just replace the highlighted line with return 0;

All we're doing here is looking at the values in the position field. If they're identical then we sort by the name field instead. Otherwise we do a comparison of where in the sortorder list the relevant values appear and sort based on that. The output is as follows:

	name	position
[0]	Michael Davis	President
[1]	Sarah Miller	Vice-President
[2]	Mary Johnson	Secretary
[3]	Amanda Miller	Member
[4]	James Brown	Member
[5]	Patricia Williams	Member
[6]	Patrick Miller	Member

As detailed above for simpler cases we can now enhance this function to make it more generic and re-usable, but I'll leave that as an exercise for you. You would need pass the sortorder array to the function along with details of which fields to sort by.

If you're working on the web with this type of data, you might also want to read the article Formatting Data as HTML which presents instructions for converting PHP data to HTML tables, lists and forms.

Multiple row sorting using a closure

Thomas Heuer from Germany has been kind enough to provide the following code and examples, which looks very nice, but will only work with PHP 5.3 and higher:

<?PHP function sortArray($data, $field) { if(!is_array($field)) $field = array($field); usort($data, function($a, $b) use($field) { $retval = 0; foreach($field as $fieldname) { if($retval == 0) $retval = strnatcmp($a[$fieldname],$b[$fieldname]); } return $retval; }); return $data; } ?>

Example calls:

<?PHP $data = sortArray($data, 'age'); $data = sortArray($data, array('lastname', 'firstname')); ?>

Note: I haven't been able to test this code yet as we're still running PHP 5.2, but it looks like it will work. You should also be able to modify the function to use 'pass by reference' as per previous examples.

References

• PHP: Array Functions
• PHP: Passing by Reference

Translations

• Russian - by Anton Afanasyev

< PHP

User Comments and Notes

Send Feedback

Load Feedback Form

© Copyright 2013 Chirp Internet - Page Last Modified: 26 May 2012