Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:26525
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain php.net from 81.169.182.136 cause and error)
Date: Sun, 12 Nov 2006 15:04:28 +0100
Reply-To: Marcus Boerger <helly@php.net>
Message-ID: <1425479062.20061112150428@marcus-boerger.de>
To: Jessie Hernandez <jrhernandez05@gmail.com>
Cc: internals@lists.php.net
In-Reply-To: <3D.90.24000.85F66554@pb1.pair.com>
References: <4554AE0D.4080600@caedmon.net> <99.20.32592.5ED45554@pb1.pair.com> <19674119.20061111112203@marcus-boerger.de> <3D.90.24000.85F66554@pb1.pair.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] Re: Namespaces in PHP 6 - ++$take
From: helly@php.net (Marcus Boerger)

Hello Jessie,

  we already have consts in our classes, if you look close then you will see
that the table does not conatain 'namespace:const' but only 'namespace:class'
and still there are conflicts already.

best regards
marcus

Sunday, November 12, 2006, 1:48:22 AM, you wrote:

> Hi Marcus,

> As I said, I wanted to only enable classes inside namespaces, and NOT 
> allow functions or constants. class::const would stay the same, so there 
> are no conflicts.


> Marcus Boerger wrote:
>> Hello Jessie,
>> 
>>   sorry to say this but look at the following, keeping in mind that the
>> whitespace is optional.
>> 
>> $x = $foo ? class:const           : $var                    // 1, 0
>> $x = $foo ? class:const           : const;                  // 1, 0
>> $x = $foo ? class:const           : class:const;            // 1, 1
>> $x = $foo ? class:const           : namespace:class:const;  // 1, 2
>> $x = $foo ? namespace:class:const : $var;                   // 2, 0
>> $x = $foo ? namespace:class:const : const;                  // 2, 0
>> $x = $foo ? namespace:class:const : class:const;            // 2, 1
>> $x = $foo ? namespace:class:const : namespace:class:const;  // 2, 2
>> 
>> Feel free to continue that list, it has conflicts whatever you do.
>> 
>> Saturday, November 11, 2006, 5:13:22 AM, you wrote:
>> 
>>> Hello,
>> 
>>> I haven't had time to work on my patch, but thinking about this some 
>>> more, I'm convinced namespaces should only contain classes. The only 
>>> problem that was present when permitting functions/constants to be 
>>> inside namespaces was the ambiguity in ternary expressions. By just 
>>> supporting classes inside namespaces, this issue would go away. Besides, 
>>> I'll dare say that most, if not all, the developers who want namespaces 
>>> will only group classes with it anyways,
>> 
>>> Also, by only supporting classes, we can use ":" instead of the long 
>>> ":::" separator and everyone would be happy.
>> 
>>> Last I checked, the only things that were pending on my patch were some 
>>> scoping issues which I had to fix. These are listed below. Once these 
>>> are done, the patch can be formally proposed. If anyone wants to take a 
>>> look at the patch so far and/or work on the remaining issues below, let 
>>> me know and I'll either post the patch or email it.
>> 
>> 
>>> Pending:
>> 
>>> 1) The extends clause should resolve imported class names.
>> 
>>> import class ns:DateBase;
>> 
>>> namespace ns{ class Date extends DateBase{} }
>> 
>>> 2) To access a global symbol, use global:<class_name> syntax.
>> 
>> I would prefer ':::'<class_name> or '\'<class_name> but that doesn't really
>> matter as we have 'global' as a keyword already.
>> 
>>> 3) Type hints should also consider imported classes.
>> 
>>> 4) When an import is done (with alias or not), and a global class with 
>>> that same name exists, what is the desired behavior? Error? Global takes 
>>> precedence?
>> 
>> Maybe this new discussion gave one hint. Aliases could be solved with a
>> flag. Just copy the classwith a new name into the classlist again and flagit
>> as copy. Maybe the original class gets a list of the copies of the copies a
>> pointer to the original but that would be an implementation detail. As soon
>> as that is done importdoes nothingelse then copying classes on a single
>> class table. That said namespaces would, if after all, simply contain other
>> copies to the original classes. In the extremecase we can start with
>> namespaces only being a 'stupid' list. Reflection could then travers all
>> classes to see in which namespaces it was registered.
>> 
>> This btw would also fix 3) as the namespace seperator would be a normal
>> sign in class lookup, it would simply be disallowed in definitions of
>> class/interface/namespaces.
>> 
>> Best regards,
>>  Marcus




Best regards,
 Marcus