const - Best practices

[Jonathan M Davis]

On Saturday, December 31, 2011 08:02:52 Caligo wrote:
> On Sun, Jan 2, 2011 at 4:43 PM, Jonathan M Davis <jmdavisProg at gmx.com>wrote:
> > - Jonathan M Davis
> 
> Almost a year has passed, and what is the state of const now?  Is it still
> broken?  I've been reading some threads on SO, and I'm a bit confused as to
> what's correct in my code and what's not.

Progress has definitely been made. inout finally works. Dynamic arrays are being 
changed to instantiate as tail-const with IFTI, so templated functions in 
Phobos will do much better with const (I think that the change is in github 
but not released yet; if not, it should be on github soon). A number of bugs 
have been fixed as well. But not everything has been sorted out yet. I believe 
that the two biggest issues are

1. Const-correctness. We have pull requests to fix this which are under 
discussion, but due to some of the non-trivial implications of the changes, 
they haven't been merged in yet.

2. Postblit constructors don't work with const or immutable (which makes a 
fair bit of sense when you consider how they work by memcpying and then 
changing the state of the copy). Walter and Andrei supposedly have a solution, 
but they haven't revealed what it is yet, let alone implemented it.

So, the situation is definitely better than it was, and const is quite usable 
in plenty of circumstances, but there are still definitely outstanding issues 
which need to be resolved. Fortunately however, solutions for them _are_ in 
the works.

- Jonathan M Davis