Consecutive krsort()

[k0z]

If i have an array of data (the values don't matter, only the keys), such as:

$myArray[21]  = 0;
$myArray[22] = 0;
$myArray[23] = 0;
$myArray[25]  = 0;
$myArray[24] = 0;
$myArray[26] = 0;

 

Running a krsort() on this array turns it into:

$myArray[26]
$myArray[25]
..
..
$myArray[21]

 

The problem is, these keys represent days of the month. If I had an array of:

$myArray[29]
$myArray[28]
$myArray[30]
$myArray[2]
$myArray[1]

 

I'd actually want this to be turned into: (EDIT: SORRY, the below values should be reversed because this is KRSORT and not KSORT. THIS GOES FOR ALL OF THE FOLLOWING EXAMPLES)

$myArray[28]
$myArray[29]
$myArray[30]
$myArray[1]
$myArray[2]

 

To represent an actual month.  In addition, the data might not roll over the next month, so the data could be exactly what the first example was.

 

Obviously, different months have different lengths, so the data could potentially be:

$myArray[28]
$myArray[27]
$myArray[1]
$myArray[2]

 

And this would need to be turned into:

$myArray[27]
$myArray[28]
$myArray[1]
$myArray[2]

 

What would the easiest way to do this be?

 

Thanks.

[Pikachu2000]

uksort($array, "strnatcmp") should do the trick.

[k0z]

uksort($array, "strnatcmp") should do the trick.

 

Tried your code:

 

$myarray = array(1=>'test',30=>'test',2=>'test',3=>'test');
uksort($myarray, "strnatcmp");
var_dump($myarray);

 

This should have returned the keys in order: 30,1,2,3

But it returned them in 1,2,3,30

 

?

[Pikachu2000]

Sorry, overlooked the ordering, make it uksort($array, "strnatcmpdesc");

[k0z]

Sorry, overlooked the ordering, make it uksort($array, "strnatcmpdesc");

 

Now receiving the error:

 

Warning: uksort() [function.uksort]: Invalid comparison function 

[Pikachu2000]

That's odd. What version of PHP are you using? I'm on 5.2.12 and just created a new script with this:

 

<?php
$myarray = array(1=>'one',30=>'thirty',2=>'two',3=>'three');
uksort($myarray, "strnatcmpdesc");
var_dump($myarray);
?>

 

 

 

 

Which output this:

 

array(4) { [1]=> string(3) "one" [30]=> string(6) "thirty" [2]=> string(3) "two" [3]=> string(5) "three" }

[Pikachu2000]

Nevermind, you're right. I forgot I had set display_errors = off earlier. It is throwing a warning level error for me. It works, but it still throws a warning. I'm too tired to look into it right now, but perhaps tomorrow, to satisfy my curiosity.

[Ruzzas]

throw a @ infront of it.

[k0z]

With or without the warning, the returned data is still incorrect. It returns:

array(4) { [1]=>  string(3) "one" [30]=>  string(6) "thirty" [2]=>  string(3) "two" [3]=>  string(5) "three" }

 

when it should be 30, 1, 2, 3

[PFMaBiSmAd]

Your numerical indexes don't contain enough information for this to work. With the current data, there is simply no way for a computer to know that the 1,2 are from a greater month than the 28,29,and 30 and to order them correctly.

 

Dates consist of year, month, and day parts and each of those parts are significant if you are making greater-than/less-than comparisons and sorting date values. If you make your problem general purpose and carry all three parts, your problem will be solved (look back at the Y2K problem.)

[k0z]

Your numerical indexes don't contain enough information for this to work. With the current data, there is simply no way for a computer to know that the 1,2 are from a greater month than the 28,29,and 30 and to order them correctly.

 

Dates consist of year, month, and day parts and each of those parts are significant if you are making greater-than/less-than comparisons and sorting date values. If you make your problem general purpose and carry all three parts, your problem will be solved (look back at the Y2K problem.)

That's not *quite* correct, actually. If you look at the data:

1,2,28,29,30

 

The amount of numbers between the end of a month to 1 (with respect to months) can be a maximum of 3 (because february starts @ the 28th). Furthermore, the jump from e.g. 2 to 20, for e.g., is much greater. If there is a deviation of a number LARGER than 3, then it is obvious it should go through a simple krsort();

 

I'm aware that a much simpler solution would be to maintain the month - but currently, it's not an option.

[Pikachu2000]

You know what? I have to admit I completely misread the question, and misinterpreted what you were trying to do. I believe it's time for sleep now.

[PFMaBiSmAd]

Your scheme falls apart if I want to operate on, for example, the 3rd through the end of the current month and the 1st and 2nd of next month. The values would be indistinguishable.

 

Even with the gap notion, one simple sort won't work. You would need to split the data at the gap, sort each group, then put the data back together.

 

... but currently, it's not an option.

^^^ It's never too late to fix a design problem. The earlier you fix a problem, the more time you will end up saving over the time you will waste working around the design flaws.