Am 08.08.2013 15:12 schrieb "Matthieu Napoli" <matth...@mnapoli.fr>:
>
> Le 08/08/2013 13:11, Martin Keckeis a écrit :
>
>> Am 08.08.2013 12:34 schrieb "Jordi Boggiano" <j.boggi...@seld.be>:
>>>
>>>
>>> On 08.08.2013 10:34, Leigh wrote:
>>>>
>>>> I'm not sure what problem this is really trying to solve, the
>>
>> boilerplate
>>>>
>>>> code you mention is very explicit and it is very clear to the reader
>>
>> what
>>>>
>>>> is being done. Each property documented with its type, purpose and
>>>> visibility in a common place for easy reference (at the top of the
>>
>> class).
>>>>
>>>> Each property that takes a value from the constructor assigned in the
>>>> constructor. Very clear.
>>>
>>>
>>> I for one am pretty tired of writing this boilerplate in every second
>>> class I write. Using dependency injection you end up having to write a
>>> LOT of those usually, and constructors typically only contain
assignments.
>>>
>>> Adding a property means: declaring the property, adding the ctor arg,
>>> adding the assignment in the ctor. You have to write the property name 4
>>> times. With this RFC in it'd come down to writing it once, and would
>>> avoid having undeclared properties because someone forgot.
>>>
>>> I am very supportive the idea, although I see that it could be confusing
>>> to some. Maybe the syntax needs to change, but the overall change is
>>> much welcome.
>>
>>
>> The syntax itself i feel is self explaining. You read it and you know
whats
>> going on.
>>
>> But somehow i feel this is "wrong". Declaring visibility directly in the
>> arguments? This looks very scary to me. (Maybe only because i've never
seen
>> it before...)
>>
>> Other reasons against:
>> Phpdoc generator break
>> Code completion break
>> If constructor is not at the beginning never see a var declaration
>> Auto generation of set/get can be achieved with IDE..so its not much work
>
>
> I second the feeling about the syntax, there are too many disadvantages.
However the idea is excellent.
>
> Maybe an alternative approach could do, here is a random suggestion (a
bit more verbose):
>
> class MyClass {
> public $foo;
> protected $bar;
>
> public function __construct($this->foo, $this->bar, $baz) {
> // $this->foo and $this->bar are now set
> $baz->doAnything(); // $baz is a standard parameter
> }
> }
>
> This would even be compatible with an interface:
>
> interface MyInterface {
> function __construct($foo, $bar, $baz);
>
> }
Another thing which came into my mind:
I normall create a set/get method also to a construct parameter to
overwrite and unittest it...
So then i cant use this, because i construct i also use the set mehod to be
sure the var is assigned always the same (e.g. additional input check or
exceptions)
