[OpenLayers-Dev] constructor signatures

[christopher.schmidt at nokia.com]

On Jun 18, 2010, at 11:02 AM, ext Tim Schaub wrote:
> To clarify here, specifically for the WMTS layer, these are the strictly 
> required properties:
> 
>  * url
>  * layer identifier
>  * style identifier
>  * matrixSet identifier
> 
> And who knows, we might get smarter in the future and find that we can 
> derive sensible defaults for those*.  Anyway, at least four required 
> arguments to start with.  And by required I mean that the layer will not 
> render anything when added to a map if those arguments are not provided. 
>  I'm putting together a patch for the layer that will make the 
> constructor throw an error when the layer is not properly configured - 
> to reduce the chance for user error.

A good plan. And stuff like this -- specifically, the potential that
the defaults could *change* in the future -- is the kind of thing that
makes the current 'required are seperate from not required' silly. 
(In a WMS layer, 'params' is required -- but in reality, there's no reason
you *have* to have parameters for a given layer -- same kind of problem.)

> Having a single argument for the WMTS layer constructor also makes 
> things nicer for creating layers from parsed capabilities.  This is a 
> detail specific to the WMTS layer, but I had it in mind when making the 
> decision.  In general, a single configuration argument is more 
> convenient to pass around, extend, serialize (assuming people want to 
> persist app configuration), etc.

I'm reasonably convinced that this is the right thing for the future.

>> Backing out on something that is one of our few points of consistency --
>> Layers require a string as their first argument -- would make me 
>> sad, but I'm not going to argue that we need to revert a change
>> because of it. I just think it's a worthwhile convention to hold onto
>> until we really go ahead and change many of our conventions for the
>> Glorious Future.
>> 
>> Going forward into 3.x, I am okay with abandoning this convention
>> in favor of making everything optional. I don't like it, personally, 
>> but if this is common Javascript convention, I'm willing to go with 
>> it. 
>> 
> 
> I completely understand this.  I suspected this was about a change that 
> would make you sad.
> 
> So, right now we have an opportunity.  We can create new constructors 
> that are forward compatible, or we can create new constructors that 
> follow old conventions (and need to be changed in the future).
> 
> I hear that you (Chris), Eric, and Bart favor going with old conventions 
> rather than forward compatibility.

That about sums it up. :) 

Regards,
-- 
Christopher Schmidt
Nokia