Re: Initing sublacss with superclass method

On 11 Jul 2025, at 14:17, Riccardo Mottola <[email protected]> =
wrote:
>=20
> Hi,
>=20
> In Grr there is this code:
>=20
>     wsAndTagClosing =3D [NSMutableCharacterSet
> characterSetWithCharactersInString: @"/>"];
>=20
> Clang warns me with:
>=20
> NSString+TolerantHTML.m:109:21: warning: incompatible pointer types
> assigning to 'NSMutableCharacterSet *' from 'NSCharacterSet *'
> [-Wincompatible-pointer-types]
>   109 |     wsAndTagClosing =3D [NSMutableCharacterSet
> characterSetWithCharactersInString: @"/>"];
>=20
> GCC gives a more generic issue about "distinct objective C type".
>=20
> Now, characterSetWithCharactersInString: is a class method of
> NSCharacterSet:
>=20
> =
https://www.gnustep.org/resources/documentation/Developer/Base/Reference/N=
SCharacterSet.html#method$NSCharacterSet+characterSetWithCharactersInStrin=
g$
>=20
> and NSMutableCharacterSet is a subclass of NSCharacterSet
>=20
> Shouldn't thus the method called on the subclass still return a valid
> subclass and not the superclass?
>=20
> Could it be a header issue? we define it as:
> + (NSCharacterSet*) characterSetWithCharactersInString: =
(NSString*)aString;
>=20
> maybe it should return instancetype?

Factory methods on the superclass, if called on the subclass, are not =
required to return an instance of the subclass, they can return a =
totally unrelated subclass.  For example, a bunch of the NSString and =
NSDictionary factory methods will return a specialised version for small =
of arguments.  If you=E2=80=99re doing this kind of trick, you need an =
explicit check in the method implementation to see if `self` is the =
expected class. =20

If that check exists, or if this method isn=E2=80=99t returning a =
subclass instance, it=E2=80=99s safe to turn it into instancetype and =
this warning will go away.  If not, then this is a real warning and the =
mutable subclass is not being returned.

You can also try `NSLog(@=E2=80=9CWhat even is this? %@=E2=80=9C, =
[wsAndTagClosing class]);` to see.

David