array_unique vs array_flip

Question

If I had an array of signed integers e.g:

Array
(
    [0] => -3
    [1] => 1
    [2] => 2
    [3] => 3
    [4] => 3
)

To get unique values I would instinctively use array_unique but after consideration I could perform array_flip twice which would have the same effect, and I think it would be quicker?

array_unique O(n log n) because of the sort operation it uses

array_flip O(n)

Am I correct in my assumptions?

UPDATE / EXAMPLE:

$intArray1 = array(-4,1,2,3);
print_r($intArray1);
$intArray1 = array_flip($intArray1);
print_r($intArray1);
$intArray1 = array_flip($intArray1);
print_r($intArray1);

Array
(
    [0] => -3
    [1] => 1
    [2] => 2
    [3] => 3
    [4] => 3
)
Array
(
    [-3] => 0
    [1] => 1
    [2] => 2
    [3] => 4
)
Array
(
    [0] => -3
    [1] => 1
    [2] => 2
    [4] => 3
)

Alasdair · Accepted Answer · 2011-11-30 06:05:58Z

37

I benchmarked it for you: CodePad

Your intuition on this was correct!

$test=array();
for($run=0; $run<1000; $run++)
$test[]=rand(0,100);

$time=microtime(true);

for($run=0; $run<100; $run++)
$out=array_unique($test);

$time=microtime(true)-$time;
echo 'Array Unique: '.$time."\n";

$time=microtime(true);

for($run=0; $run<100; $run++)
$out=array_keys(array_flip($test));

$time=microtime(true)-$time;
echo 'Keys Flip: '.$time."\n";

$time=microtime(true);

for($run=0; $run<100; $run++)
$out=array_flip(array_flip($test));

$time=microtime(true)-$time;
echo 'Flip Flip: '.$time."\n";

Output:

Array Unique: 1.1829199790955
Keys Flip: 0.0084578990936279
Flip Flip: 0.0083951950073242

Note that array_keys(array_flip($array)) will give a new key values in order, which in many cases may be what you want (identical except much faster to array_values(array_unique($array))), whereas array_flip(array_flip($array)) is identical (except much faster) to array_unique($array) where the keys remain the same.

edited Nov 30, 2011 at 6:05

answered Nov 30, 2011 at 5:52

Alasdair

14.3k18 gold badges95 silver badges151 bronze badges

Sign up to request clarification or add additional context in comments.

9 Comments

Leith Over a year ago

There's also array_keys(array_count_values($array)) - see CodePad

Alasdair Over a year ago

@Leith, array_keys(array_count_values( can never be faster than array_keys(array_flip( but the result will always be the same.

Leith Over a year ago

@Alasdair, why do you say that? The modified CodePad benchmark I linked in my previous comment indicates a significant improvement.

Alasdair Over a year ago

@Leith, yes you are right. Your benchmark does indeed say its faster. That's quite interesting.

CMCDragonkai Over a year ago

Is array_flip O(n) or something better?

|

Schnepper · Accepted Answer · 2015-09-21 20:33:05Z

5

Caution: this technique is NOT a drop-in replacement for array_unique(). It only works for arrays with values that are are valid keys. (eg: string, integer, things can can be cast to int). And certainly does not work for arrays of objects.

$input = [true, false, 1, 0, 1.2, "1", "two", "0"];
var_export(array_unique($input));
array (
  0 => true,
  1 => false,
  3 => 0,
  4 => 1.2,
  6 => 'two',
)

vs:

var_export(array_keys(array_flip($input)));

PHP Warning:  array_flip(): Can only flip STRING and INTEGER values! 
in php shell code on line 1

array (
  0 => 1,
  1 => 0,
  2 => 'two',
)

answered Sep 21, 2015 at 20:33

Schnepper

511 silver badge2 bronze badges

1 Comment

Shapeshifter Over a year ago

this isn't an answer really, but saved me some time so thanks!

Dogbert · Accepted Answer · 2011-11-30 05:51:48Z

Nothing is better than running your own benchmark.

➜  8321620  cat first.php
<?php

$arr = array(-3, 1, 2, 3, 3);

for($i = 0; $i <= 1000000; $i++) {
    array_unique($arr);
}
➜  8321620  time php first.php
php first.php  3.24s user 0.01s system 99% cpu 3.251 total
➜  8321620  cat second.php
<?php

$arr = array(-3, 1, 2, 3, 3);

for($i = 0; $i <= 1000000; $i++) {
    array_flip(array_flip($arr));
}
➜  8321620  time php second.php
php second.php  1.50s user 0.01s system 99% cpu 1.514 total

Update: Array with 1000 elements.

➜  8321620  cat first.php
<?php

$arr = array();
for($i = 0; $i <= 1000; $i++) {
    $arr[] = rand(0, 1000);
}

for($i = 0; $i <= 10000; $i++) {
    array_unique($arr);
}
➜  8321620  time php first.php
php first.php  27.50s user 0.03s system 99% cpu 27.534 total
➜  8321620  cat second.php
<?php

$arr = array();
for($i = 0; $i <= 1000; $i++) {
    $arr[] = rand(0, 1000);
}

for($i = 0; $i <= 10000; $i++) {
    array_flip(array_flip($arr));
}
➜  8321620  time php second.php 
php second.php  1.59s user 0.01s system 99% cpu 1.604 total

So yes, your assumption was correct.

Galen · Accepted Answer · 2011-11-30 05:43:54Z

-3

you would have to use

array_keys( array_flip( $array ) );

which would take more time

I would go for array_unique. It has the added benefit of explaining whats happening.

answered Nov 30, 2011 at 5:43

Galen

30.2k9 gold badges75 silver badges90 bronze badges

7 Comments

Dogbert Over a year ago

Isn't array_keys also O(n)?

Galen Over a year ago

My point is you have to use 2 functions, not 1. It will take more time and obfuscate the code as well.

Lizard Over a year ago

I have given an example of my array_flip code, array_keys isn't required

Galen Over a year ago

Well, now you're using array_flip twice

Dogbert Over a year ago

@Galen, Twice O(n) is much faster than once O(n log n).

|

Collectives™ on Stack Overflow

array_unique vs array_flip

4 Answers 4

9 Comments

1 Comment

Comments

7 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

9 Comments

1 Comment

Comments

7 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related