#18773 [Ctl->Csd]: Quicksort result is different between Windows and Unix.

ID: 18773 Updated by: zeev@php.net Reported By: rgaidot@swt.fr -Status: Critical +Status: Closed Bug Type: Arrays related Operating System: Windows 2000 Pro PHP Version: 4CVS-2002-08-07 New Comment: Ok, technical background first: This bug has to do with several facts: 1. PHP performs numeric comparison between strings that 'look like numbers'. For instance, 000000073 is a string that looks like a number. A numeric comparison will be performed if and only if both of the compared strings look like numbers. 2. In the implementation of this functionality, the C library function strtod() is used. 3. strtod() recognizes exponential notations, so 000000E2 is actually 0.0E+2, or 0. 4. Under Windows, the letter 'D' is also recognized as the exponent sign so 000000D2 is also 0.0E+2. Don't ask me why, but it's clearly documented in MSDN :) The 4th point is the source of the difference between the behavior you see under UNIX and the one you see under Windows. Try changing the '000000D2' element to '000000E2', and you should see identical behavior under UNIX too (haven't tested it, but I'm pretty sure this would happen). Solution? There's an easy one, and a difficult one. Easy one - add a letter (other than A-F) in the beginning of each element you wish to sort on. For intsance, if you start any string with 'S' or 'Z', PHP will not mistake it for a numeric string. Drawback - you may not be in control of the data, so it may not work. Complex one - implement the comparison function yourself, and if both elements are strings, call strcmp() instead of using >=. You would also have to implement array comparison yourself so that it calls your comparison function recursively. If you use the built-in >= operator, it will use the numeric-string aware comparison function. I hope you're not lost, and I hope this helps :) Previous Comments: ------------------------------------------------------------------------ [2002-08-20 13:59:04] zeev@php.net Ok, almost :) <?php print_r(sort(array('0D2', '004'))); ?> ------------------------------------------------------------------------ [2002-08-20 13:55:39] zeev@php.net Reduced to: <?php print_r(array('0D2', '004')); ?> Further details soon... ------------------------------------------------------------------------ [2002-08-08 08:31:26] kalowsky@php.net I'd like to mark this as critical because my local tests show: MacOSX results: Sort 1 Sort 2 1 5 2 3 3 4 4 2 5 1 Win2000 results: Sort 1 Sort 2 1 1 2 3 3 4 4 5 5 2 Neither sort went over correctly. ------------------------------------------------------------------------ [2002-08-07 05:02:12] eru@php.net reclassified, updated version, verified ------------------------------------------------------------------------ [2002-08-07 04:46:18] rgaidot@swt.fr I have tried to use the sort function on a string array. The array strings include both integer and char characters. (e.g. 0000AER, 000000S0, 00000073) I have noted that the result differs between the two systems Windows and Unix. Indeed the result on Unix is correct, but it is not correct on Windows. I have made a C++ test application in order to compare the results between the two systems and I did not note any problem. <PHPScript_Test> $a=array( array('000000S0','1','5'), array('000000O2','2','4'), array('000000D2','3','3'), array('000000MH','4','2'), array('00000073','5','1') ); echo "<h1>Data</h1>"; echo "<pre>"; var_dump($a); echo "</pre>"; sort($a); echo "<h1>Result</h1>"; echo "<pre>"; var_dump($a); echo "</pre>"; </PHPScript_Test> Thanks, Régis Gaidot ------------------------------------------------------------------------ -- Edit this bug report at http://bugs.php.net/?id=18773&edit=1