Re: Pre-PPC: class :abstract

On Fri, 19 Dec 2025 at 11:59, Ovid <[email protected]> wrote:

> On Wed, Dec 17, 2025 at 3:47 PM Branislav Zahradník <
> [email protected]> wrote:
>
>>
>>
>> On Wed, 17 Dec 2025 at 15:38, Paul "LeoNerd" Evans <
>> [email protected]> wrote:
>>
>>
> I always though that abstract has same usefulness as Java's final - hardly
>> to be distinguished from zero.
>>
>
> Final classes can be interesting. In some languages, you no longer have to
> use a vtable to lookup methods, meaning you get a nice performance boost.
> It's also useful when you have extremely sensitive classes (e.g.,
> something security related) which you don't want people to change without
> careful consideration. It also makes it impossible to violate Liskov :)
>

imho this should not be property of the class but of an instance (or data
contact).

I have no problem with restrictions like (hypothetical):
- new final Class ...
- foo (final Class b)

to require exact instance - as well as a hint to compiler that it can do
compile time vtable mapping.



>
> I wouldn't reach for final classes as a first solution, but they're not
> without merit. It's just that the use cases are somewhat restricted (though
> our inability to enforce Liskov might be an argument).
>
> For abstract classes, they have uses, too, such as being the base class
> for a factory and ensuring that every factory class has the required
> interface. They do have one notable fragility, though. If you add a new
> abstract method, all subclasses are immediately broken until they can be
> updated. Maybe not so bad for closed source work, but bad for open source.
>

I get this reasons and all of them are still valid. Though I don't think
there is any reason to use word abstract apart of fact saying
"I'm not implementing method" ... that is still something compiler can
easily guess - it knows that method is declared but not implemented in
class.

Also interfaces and abstract is imho better implemented via role mechanism
- for me (you know how my brain works)
this sounds like "abstract class = role with subset of features".


>
> Best,
> Ovid
> --
> https://curtispoe.org/
>