Re: Some issues calling atk methods

[Piñeiro <apinheiro igalia com>]

From: Dan Winship <danw gnome org>

> On 08/12/2010 01:56 PM, Piñeiro wrote:
>> After thinking a little about it, I realize that this .get_name that
>> is being called is atk_action_get_name [2], as this class implements
>> AtkAction interface.
> 
>> Although I tried some (random) ways to call atk_object_get_name on
>> this object, I was not able to do that.
>> 
>> Any idea or hint?
> 
> hm... try Atk.Object.prototype.get_name.call(atk_label) ?

Just tested. Doesn't work. It says that call is not defined. Anyway, I
tested this:

  >>> Atk.Object.prototype.get_name (atk_label)

And returns NULL, what it is wrong as the object has a name. Although
it call the method, it fails in this atk_object_get_name check:

 g_return_val_if_fail (ATK_IS_OBJECT (accessible), NULL);

Launching the g_return_val_if_fail warning. So I guess that atk_label
is not valid as a real AtkObject, after all it is a proxy:

 >>> atk_label = label.get_accessible ()
 r (2) = [object instance proxy GIName:Atk.GObjectAccessible jsobj 0xXXXXXX native 0xXXXXXXX]

There is any way to call this method using the real AtkObject
instance?

Anyway, although we get that, that doesn't seems a really clean way to
call this method, and not really general. Although enough if we don't
find too many cases (see below, last paragraph).

> That assumes AtkLabel isn't overriding AtkObjectClass->get_name... if it
> is, I don't think there's any way in gjs to refer to AtkLabel's
> implementation of the method.

Well, sorry, there isn't any AtkLabel class/interface. I guess that
you are talking about AtkAction.

No, AtkAction is not overriding AtkObjectClass->get_name. CallyActor
class is an AtkObjectClass that implements AtkAction (so implementing
AtkAction->get_name). StallyLabel [1] is overriding
AtkObject->get_name. But AtkAction->get_name is not overriding
AtkObject->get_name.

Because if not, this problem would also arise in the "C World" calling
atk_object_get_name, not just using javascript.

And also, take into account that the signature is
different. atk_object_get_name just receives the
instance. atk_action_get_name receives the instance and a gint, as it
returns the name of the nth action of this atkobject.

> Does this work usefully in other language bindings? I think this is just
> considered bad class design, and the answer is "don't do that then"...

You mean trying Atk with python, for example? I haven't tested it.

Also, you mean that AtkAction shouldn't have a method ->get_name
because AtkObject has already a method called in the same way? Well,
after thinking a little about it, and comparing it with other
languages, probably it is true, although just in the automatic
bindings world. I mean that this wasn't a problem at all until now.

Probably this atk_action_get_name can be called
atk_action_get_action_name, as it is requesting the name of a specific
action. But it sound redundant when you are just using C.

Finally, take into account that I was working without problem with Atk
until I found this problem, so not sure how many methods are
affected. Probably I should require to check if there are other
methods in the different interfaces with this problem.

Thanks for your hints.

BR

[1] https://bugzilla.gnome.org/show_bug.cgi?id=626658

===
API (apinheiro igalia com)