Edit report at https://bugs.php.net/bug.php?id=50688&edit=1
ID: 50688 Comment by: mbrowne83 at gmail dot com Reported by: jcampbell at remindermedia dot com Summary: Using exceptions inside usort() callback function causes a warning Status: Assigned Type: Bug Package: Arrays related Operating System: Fedora Core 12 PHP Version: 5.*, 6 Assigned To: stas Block user comment: N Private report: N New Comment: This will probably be obvious to most, but I just wanted to mention that you can always prefix the usort function with the @ symbol to prevent the warning...of course that would also suppress any other types of notices or warnings that might occcur anywhere within the sorting function... Previous Comments: ------------------------------------------------------------------------ [2012-02-24 18:04:02] keith at breadvault dot com This same problem arises when using Mockery to mock the object whose method is being used by usort(), even though the method itself neither is mocked nor handles any exceptions. The proxy generated by Mockery must wrap the target class's methods with some exception-handling code. Unfortunately this forced me to code a workaround that would not use usort. My hack extracts from the objects in the array the values being sorted on, sorts that array of values using asort() (to preserve the keys), and finally rebuilds the list of objects using the keys in the order that they appear in the asorted list of values. Yuck. ------------------------------------------------------------------------ [2012-02-21 22:56:31] eric_haney at yahoo dot com It took me a while to figure out that some code called from usort was throwing, catching, and (gracefully) handling an Exception. Then I found this post. Quite frustrating. I turned off warnings with ini_set before calling usort, then turned them on again after. This is an effective workaround for now, but I'd love to clean that nastiness out of my code. It is also my opinion that usort should be allowed to change the elements in the array. EG: an instance variable of an object may be lazy-loaded as a result of a method call from within a usort callback. Should a warning really be issued in that case? ------------------------------------------------------------------------ [2011-10-10 21:44:56] poehler at interworx dot com This bug is still present as of PHP 5.3.8, we ran into it today and spent most of a day trying to figure out what was causing the error message "Array was modified by the user comparison function", when CLEARLY, NOTHING was changing the array at all! The exception was not thrown/caught directly in the usort function but rather in a constructor of a class that was called about 3 or 4 functions deep from the usort, making it very difficult to track down. After finally figuring out the exception was somehow related, we searched google and found this bug report. I'm sure we can agree that the minor act of catching an exception should not result in usort throwing a warning message. This bug is a huge timewaster :( ------------------------------------------------------------------------ [2010-10-07 23:34:54] philipwhiuk at hotmail dot com I notice this is still affecting PHP 5.3.3 (Windows/Apache install). Is this likely to be fixed soon - is it a question of developer time and priority or is it too difficult to fix? It's quite irritating - I realise that the obvious solution is to avoid throwing the exception (ha-ha) but it's a useful function and exceptions are... inevitable. ------------------------------------------------------------------------ [2010-05-31 17:24:05] ajrattink at correct dot net I printed a debug line from my usort callback. It called debug_backtrace() to print the line and sourcefile in the debuglog. And therefor triggered the error. Even more, it did not sort. Maybe the phpmanual should state that usort() callbacks are not allowed to write loglines. I also think that usort() callbacks that DO change the array are perfectly legal, as long as they don't change the sort. Maybe your sorter code needs stackoverflow protection or whatever, but calling certain code 'invalid', because it causes your code to SEGV is a stupid way to solve a bug. ------------------------------------------------------------------------ The remainder of the comments for this report are too long. To view the rest of the comments, please view the bug report online at https://bugs.php.net/bug.php?id=50688 -- Edit this bug report at https://bugs.php.net/bug.php?id=50688&edit=1