[PATCH] D79740: Align mapped_iterator::reference type with mapped_iterator::operator*() return value.

[David Blaikie via Phabricator via llvm-commits]

dblaikie added a comment.

In D79740#2030762 <https://reviews.llvm.org/D79740#2030762>, @dblaikie wrote:

> In D79740#2030720 <https://reviews.llvm.org/D79740#2030720>, @craig.topper wrote:
>
> > @dblaikie are you saying that a mapped_iterator that returns by value should never be used by a reverse_iterator?
>
>
> That's my hypothesis/understanding, yes. I'm open to other ideas/counterpoints.


An alternative: A mapped_iterator that invokes the functor on increment, and stores the result in the iterator itself. That way its op* can return a reference to that object, etc. Then you satisfy the other iterator requirements.

So, I guess, we could make that happen implicitly - if the functor returns by value we switch to an iterator implementation that stashes that value in a member, otherwise if it returns by reference we can assume it has storage elsewhere and use the existing implementation.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D79740/new/

https://reviews.llvm.org/D79740