Using ByteVec consuming trait methods on BString fails with "cannot move out of dereference of BString"

[Byron]

My actual goal is to convert a BString into a String, in a way that errors if it's not UTF-8. Basically String::from_utf8(my_bstring.into_inner()).

However, when trying it, I see 'cannot move out of dereference of BString', which makes sense as all functionality is implemented via Deref.

The problem is that my API uses BStr and sometimes BString so it would be desirable to be able dissolve the BString once again.

Here is the playground link.

Thanks for your advice.

[BurntSushi]

In terms of how to make progress here, I think you want Vec::from(s).into_string(): https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=7492b80f15720c0b4585e9cc74d2359a

But this is probably a specific example of a general problem. Namely, that BString cannot call methods on ByteVec. I wonder how we might accomplish that. I can think of two ways:

1. Implement ByteVec for BString explicitly.
2. Add inherent methods on BString for all methods on ByteVec that move self.

I know (2) would work, but would be clunky/unfortunate. I'm less sure about (1), since I wonder how trait method resolution would work for methods that take &self or &mut self. But perhaps everything "just works" in that case.

[Byron]

Thanks for the analysis and for showing a way forward.

I know (2) would work, but would be clunky/unfortunate. I'm less sure about (1), since I wonder how trait method resolution would work for methods that take &self or &mut self.

1. Seems like the most natural way even though it will require quite some repetition/call-forwarding. If there is ambiguity when &self and &mut self are involved, I wonder if it would even compile.

On the flip side, fixing this probably makes the API much more ergonomic, for instance, on BString something like as_bstr() won't work, but as_ref() does, which can be confusing to some me.

[BurntSushi]

It turns out that this is a duplicate of #84. Similar ideas on how to move forward are discussed there.

To be honest, I'm still not quite sure what to do. If impl ByteVec for BString makes everything work, then that is definitely the easiest path. But someone has to try it out and see what, if any, failure modes it has.

Anyway, closing in favor of #84 since it was created first and also has a fair bit of context/info there.