Taiki Endo

Thanks for the proposal. I don't have a strong opinion on whether futures-rs should include this, since async-stream and others already have the ability to do this.

https://github.com/tokio-rs/async-stream/issues/102#issuecomment-1946871459 > I don't remember exactly what happened, but I suspect I could not publish without removing it due to circular dependencies. > > https://github.com/tokio-rs/tokio/blob/b32826bc937a34e4d871c89bb2c3711ed3e20cdc/tokio-test/Cargo.toml#L22 (My understanding has not changed...

So I think the issue with async-stream can be solved if https://github.com/rust-lang/cargo/issues/4242 is solved or a workaround that works for dev-dependencies that have a version is added (a [known workaround](https://github.com/rust-lang/cargo/issues/4242#issuecomment-1539720335)...

Sorry for the slow response, I thought this was a patch that was 1k lines of complex implementation without any tests or documentation. I would like to merge this once...

> is this pr in a kind of internal testing now before merging CI is failing on the master branch (mostly due to https://github.com/rust-lang/rust/issues/124800) and we need to fix that...

Thanks! I have tested this with some URDFs, and I found the following two issues: - Serializing https://github.com/openrr/openrr/blob/38b67711531ff339b4a037badd4e9f4150c6436b/openrr-planner/sample.urdf produces the wrong URDF. - Deserializing https://github.com/ToyotaResearchInstitute/hsr_description/blob/9281be02b759e96107183b8c78d944ed8144d61b/robots/hsrb4s.urdf fails with "UrdfError(Other("Expecting `Event::Start`"))" error....

> I had a look at the second Urdf you found and found the root cause, it is a very sneaky one... [here](https://github.com/ToyotaResearchInstitute/hsr_description/blob/9281be02b759e96107183b8c78d944ed8144d61b/robots/hsrb4s.urdf#L729). If you look _very_ closely there is...

Thanks for the PR. IIRC, I implemented `Valuable` for the reference to match the `Valuable` implementation of `str`.

Another case of this problem is when I want to pass `Value::Error` to visitors. This can be avoided by making the error static (like the following example), but the problem...

> so a conversion to string should be avoided. For both cases, we could possibly add a `Display` variant: Hmm. The creation of `String` is a red herring. The struct...