Re: [Vala] Vala 0.22.0 - requiring instance of Class to access Class class-members.

[Tal Hadad <tal_hd hotmail com>]

> Hello,
>
> This is the second time I've tried to post to the list, but I haven't 
> seen my first attempt show up in the vala-list archive yet, so I thought 
> I would resend, just in case.
>
> I'm still learning my way around vala so I may be doing something wrong, 
> but according to the docs at https://wiki.gnome.org/Vala/Manual/Classes:
>
>   * Class members are shared between all instances of a class. They can
>     be accessed without an instance of the class, and class methods will
>     execute in the scope of the class.

I'll mark some points:

Class members are sharing ACCESS between all instances of a class.
They
 can     be accessed without an instance of the class(BUT HAVE TO 
SPECIFY ANOTHER INSTANCE TO GET TO), and class methods will     execute 
in the scope of the class(WITH A GIVEN INSTANCE).

There's an only one exception to my marking - static members.
Static
 members belong to the class, not to the instance, and therefore its 
value is "shared" between all instances and the class.

More about static vs non-static members(in OOP):
http://www.daniweb.com/software-development/csharp/threads/96978/static-vs-non-static-methods

> This does not appear to be the case.  In the following example I cannot 
> access the prop method of the Parent or Child classes without 
> instantiating an instance of those classes.
>
> public class Parent : Object {
>      protected class int _prop;
>      public class int prop() { return _prop; }
>      class construct {
>          _prop = 0;
>      }
> }
>
> public class Child : Parent {
>      class construct {
>          _prop = 1;
>      }
> }
>
> Parent p = new Parent();
> print ("Parent.prop: %d\n", p.prop());         // Works as expected
> print ("Parent.prop: %d\n", Parent.prop());    // error: Access to 
> instance member `Parent.prop' denied
>
> Child c = new Child();
> print ("Child.prop: %d\n", c.prop());        // Works as expected
> print ("Child.prop: %d\n", Child.prop());    // error: The name `prop' 
> does not exist in the context of `Child'
>
> Is this the desired behavior or are class methods, suppose to only be 
> accessible from class instances?
>
> Kind Regards,
>
> Tom

This is desirable behavior.
Notice that it works fine when calling method "prop" by instance.
When you call this method by a class name, it fails, since ONLY static method are allowed to be called by a 
class name.
Child.prop() would fail even if prop() is static, since the method belongs to Parent, and not to Child.

> I believe it's a bug in the compiler. It's non-sense to require an instance
> when those are class-level instances.

Sorry, but I don't understand you. Why it's a bug in compiler?

> However as a workaround you can use static fields. You can still initialize
> them with class construct { } or static construct { }.

So why don't just do this?

> Also if you were thinking of overriding a protected class field, I hardly
> believe that's implemented (there's a bug somewhere about that)

No I checked this in the past, Vala doesn't support this.
It could theoretically can be possible by register this with GObject class struct, but it's a source for ABI 
breaks.
Better to implicate this with a virtual static get method.

I do accept that this could be a nice feature.

Tal