Given

public:

Let's try this again.

This attribute block can be terminated by a corresponding protection attribute block:

// Begin private
private:
  void foo() {}

// End private
public:
   void bar() {}

But not all attributes have a corresponding attribute to turn them off:

// Begin @nogc
@nogc:
...

// End with scope/file

It would be a nice convenience to have something like this:

@nogc:

:end

Yes, I'm aware I could do this:

@nogc {}

But given the choice of : and {}, I'm tending to prefer the former these days. The lack of a way to turn off some attributes forces me to use {}. I recall discussions about something like @nogc(off) or some such. Was that ever approved? If not, I'd love to have something like :end, though I'm not attached to the name.

On Wednesday, 28 September 2016 at 09:43:38 UTC, Mike Parker wrote:
> Let's try this again.
>
> This attribute block can be terminated by a corresponding protection attribute block:
>
> // Begin private
> private:
>   void foo() {}
>
> // End private
> public:
>    void bar() {}
>
> But not all attributes have a corresponding attribute to turn them off:
>
> // Begin @nogc
> @nogc:
> ...
>
> // End with scope/file
>
> It would be a nice convenience to have something like this:
>
> @nogc:
>
> :end
>
> Yes, I'm aware I could do this:
>
> @nogc {}
>
> But given the choice of : and {}, I'm tending to prefer the former these days. The lack of a way to turn off some attributes forces me to use {}. I recall discussions about something like @nogc(off) or some such. Was that ever approved? If not, I'd love to have something like :end, though I'm not attached to the name.

AFAIR, last year or so, Andrei approved the proposal for attr(bool expr), eg:

@nogc:

// no gc code here

@nogc(false):

// code that's allowed to use the gc here

class Base(bool overridableImpl)
{
   final:


   final(overridableImpl) void impl() { /* ... */
}

But still no one has stepped in to do the implementation.

On Wednesday, 28 September 2016 at 09:48:27 UTC, ZombineDev wrote:
> On Wednesday, 28 September 2016 at 09:43:38 UTC, Mike Parker wrote:
>> [...]
>
> AFAIR, last year or so, Andrei approved the proposal for attr(bool expr), eg:
>
> @nogc:
>
> // no gc code here
>
> @nogc(false):
>
> // code that's allowed to use the gc here
>
> class Base(bool overridableImpl)
> {
>    final:
>
>
>    final(overridableImpl) void impl() { /* ... */
> }
>
> But still no one has stepped in to do the implementation.

I meant:
class Base(bool overridableImpl)
{
   final:
   /* final methods here */

   final(!overridableImpl) void impl() { /* ... */ }

   /* final methods here */
}

On Wednesday, 28 September 2016 at 09:48:27 UTC, ZombineDev wrote:

> AFAIR, last year or so, Andrei approved the proposal for attr(bool expr), eg:
>

Yeah, that's what I couldn't remember. I'll dig up the discussion read up on it again.

On 9/28/16 11:48 AM, ZombineDev wrote:
>
> AFAIR, last year or so, Andrei approved the proposal for attr(bool
> expr), eg:
>
> @nogc:
>
> // no gc code here
>
> @nogc(false):

Yah, we've got to have something. There's been quibbles over syntax but nothing better came up that is as comprehensive. @nogc(true) and generally @nogc(ctBoolean) should work as well. -- Andrei

On Wednesday, 28 September 2016 at 10:14:25 UTC, Mike Parker wrote:
> Yeah, that's what I couldn't remember. I'll dig up the discussion read up on it again.

These two DIPs drafts also try to tackle the attribute problem:

http://wiki.dlang.org/DIP64
http://wiki.dlang.org/DIP79