Posted by amygdala on 07/31/07 02:24

"Rik" <luiheidsgoeroe@hotmail.com> schreef in bericht
news:op.twa29wl2qnv3q9@metallium...
> On Tue, 31 Jul 2007 03:23:32 +0200, amygdala <noreply@noreply.com> wrote:
>
>> Hi all,
>>
>> I'm starting this new project in which I'ld like to implement sort of a
>> design pattern I have seen being used in the CodeIgniter framework.
>>
>> Basically, the site will examine the URI and based on the segments of the
>> URI it will fire up some controller class, for instance, say I have an
>> inbox
>> in which end-users can view messages they got from other users, they'ld
>> start at:
>>
>> http://www.somesite.com/inbox/view/
>>
>> This will then call upon the function view in the class inbox:
>>
>> class Inbox extends Controller
>> {
>> // ...
>> public function view()
>> {
>> // ...
>> }
>> // ...
>> }
>>
>> Any trailing segments of the URI will be passed as supplement arguments
>> to
>> the function.
>>
>> Now, if $_POST variables are sent to some page and are needed by the
>> called
>> upon function in the appropriate class, would it be bad practice to
>> simply
>> extract those $_POST variables from the superglobal $_POST inside this
>> function? Preferably I would like to send the $_POST variables as
>> arguments
>> to the function, but I'm not sure how to implement this correctly, nor
>> am I
>> even completely sure as to why this would be better practice, other than
>> that the function will only accept predefined arguments.
>>
>> What are your views on implementing such a 'pattern' (if you will). Would
>> you even suggest such an implementation, or am I bound to get into
>> trouble
>> with proposed pattern?
>
>
> There's nothing wrong with using $_POST directly. If however, you're
> writing an app that might get it's instructions another way (for instance
> the commandline, SOAP, instruction files), it might be handy to let a
> controller decide what kind of request it gets, and create a request
> object accordingly. When this is used in the code further one, it saves
> you having to write the entire interfaces for these 'channels' again, and
> assures that all requests are handled in the same manner. Offcourse, the
> response for all these channels is usually different, then again 'what
> kind of response do I want' i just part of the request offcourse. (Even
> when just talking HTTP GETS/POST there might be a difference between calls
> for entire HTML pages or for instance JSON objects).

Yes of course, this all makes sense too. And you're probably right, given
the type of requests you mentioned here, they would probably call for
different response types too. And indeed, this wouldn't necessarily mean
that the request interfaces (the arguments) need to be different.

But I have to agree with Jerry's comment (see Jerry's reply) here that if
for instance I decided to use $_SESSION of even $_GET vars for calling my
class functions it would break the interfaces. I did not think of that yet.

> --
> Rik Wasmus

[Back to original message]