Ban user-redefined init?

[IgorDeepakM]

Ban user-redefined init is mentioned in https://dpldocs.info/opend/roadmap.html

I'm trying to find information about this but I haven't really found anything about it. This is probably an old discussion I haven't been able to find and haven't found anything about that you can redefine .init for the built in types?

What is a user-redefined .init?

[adamdruppe]

The language doesn't prohibit you from having a random member called init on your structs/classes/etc, but it also doesn't work right and just breaks things if you do it.

[dkorpel]

This is probably an old discussion I haven't been able to find and haven't found anything about

See: dlang/dmd#12512

[IgorDeepakM]

In one comment.

I think we should:

Disallow overriding .init
Allow user-defined default constructors

Does this open up fixing one of the biggest warts in D, not allowing a default constructor for structs in D?

struct S
{
    int x;
    int y;

    this()
    {
        x = 1;
        y = 2;
    }
}

Or are there other obstacles preventing this?

[dkorpel]

I think the two issues are orthogonal

[adamdruppe]

that's what is meant by "allow user-defined default constructors" but yeah it is separate from the .init issue

[quickfur]

User-defined default ctors are a completely orthogonal issue. If we want to discuss that, I suggest opening a new discussion so that we don't get confused.

[quickfur]

We should definitely do this. This should have been done years ago. It does not make any sense at all for user code to redefine .init; it breaks so many things and so many expectations for the sake of a practically non-existent use case.

Kill it with fire I say!

[adamdruppe]

yeah, i used .init in some extern(obj-c) stuff because that's the name of it in that language... but yeah even there it caused trouble in some of my code. so i think it is best to ban it - even in laces where it works, it is subtly broken so better to make it openly broken