Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:73488
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.217.180 as permitted sender)
Content-Type: text/plain; charset=koi8-r; format=flowed; delsp=yes
To: "Crypto Compress" <cryptocompress@googlemail.com>
Cc: "PHP Developers Mailing List" <internals@lists.php.net>
References: <op.xdkwnbfdc9evq2@nikita-pc> <5339497A.2040506@googlemail.com>
Date: Mon, 31 Mar 2014 15:02:25 +0400
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Message-ID: <op.xdk2abhec9evq2@nikita-pc>
In-Reply-To: <5339497A.2040506@googlemail.com>
User-Agent: Opera Mail/12.16 (Win32)
Subject: Re: [PHP-DEV] usort callback should always return integer (not float)
From: inefedor@gmail.com ("Nikita Nefedov")

On Mon, 31 Mar 2014 14:54:50 +0400, Crypto Compress  
<cryptocompress@googlemail.com> wrote:

> Am 31.03.2014 11:00, schrieb Nikita Nefedov:
>> Hi,
>>
>> I've just stumbled upon this limitation in sorting (compare) callbacks.  
>> Is there any
>> reason why we can't handle floats in this case?
>>
>> I think it would be nice to be able to return float from the callback
>> because it's often happens that all it does is just `return
>> $a->someMethod() - $b->someMethod();` (where someMethod() returns  
>> float) or `return $a["something"] -
>> $b["something"];` and to make it work I need to write additional if  
>> statement. I think the reasoning could be complex behavior of doubles?  
>> Especially comparing doubles. But will we see any double-related  
>> pitfalls when all our code does with it is `return retval < 0 ? -1 :  
>> retval > 0 ? 1 : 0;` (so essentially - compares them to zero)?
>>
>>
>> I made a PR [1] on github for 5.6 version. I think this change is  
>> pretty small and could go into 5.6 not breaking our release process  
>> rules? Though it WILL introduce some BC breaks for userland code that  
>> uses floats. But this code was already broken when it was relying on  
>> old behavior.
>>
>> [1] https://github.com/php/php-src/pull/638
>>
>
> Hi,
>
> try this:
>
> <?php
> $retval = 1.0 - 0.7 - 0.3;
> var_dump($retval < 0 ? 'lt' : $retval > 0 ? 'gt' : 'equal'); // result  
> is *not* equal
>
> cryptocompress

Yes, I understand the pitfalls of floating point number arithmetics. But  
isn't it userland's problem?