core/muxing: Replace Into<io::Error> bound on StreamMuxer with std::error::Error
#2710
Conversation
thomaseizinger
changed the base branch from
master
to
refactor/introduce-substream-box
core/src/muxing/boxed.rs
Outdated
| @@ -38,6 +39,7 @@ impl<T> StreamMuxer for Wrap<T> | |||
| where | |||
| T: StreamMuxer, | |||
| T::Substream: AsyncRead + AsyncWrite + Send + Unpin + 'static, | |||
| T::Error: Into<Box<dyn Error + Send + Sync>>, | |||
There was a problem hiding this comment.
| T::Error: Into<Box<dyn Error + Send + Sync>>, | |
| T::Error: Error + Send + Sync>, |
Nit: Wouldn't this simpler trait bound also work, given that Into<Box<dyn ..>> is auto-implemented for E: Error + Send + Sync?
There was a problem hiding this comment.
Probably yes. I've copied the bounds from std::io::Error::new because that is where I am passing the type to. Omitting the Into here is probably okay :)
There was a problem hiding this comment.
I decided to put the Error bound directly on the associated type so I don't have to repeat it everywhere. The Send + Sync bounds stay on higher layers for consistency with other bounds.
There was a problem hiding this comment.
I decided to put the
Errorbound directly on the associated type so I don't have to repeat it everywhere.
It is only in 4 locations, right? a479eb8 I would prefer bounds to be as close to where they are used as possible. If I am not mistaken StreamMuxer doesn't need type Error to implement Error, right? I liked how previously this pull request would remove the bound on StreamMuxer.
There was a problem hiding this comment.
It doesn't. It also doesn't necessarily need Substream to implement AsyncRead + AsyncWrite but it is a bit pointless if it doesn't.
If we remove the Error bounds, we should also remove AsyncRead + AsyncWrite for consistency but that will get a bit redundant I think.
The current style seems to be that functional bounds are put on associated types and auto-traits are left off.
| @@ -134,7 +134,7 @@ where | |||
| pub fn poll( | |||
| &mut self, | |||
| cx: &mut Context<'_>, | |||
| ) -> Poll<Result<SubstreamEvent<TMuxer, TUserData>, IoError>> { | |||
| ) -> Poll<Result<SubstreamEvent<TMuxer, TUserData>, TMuxer::Error>> { | |||
There was a problem hiding this comment.
Not in scope for this PR, but while reviewing I was wondering: Do we even need the TMuxer generic on Muxing? I think the actual type is always StreamMuxerBox. Muxing is private the swarm crate and only used by Connection and there we already use the concrete StreamMuxerBox.
There was a problem hiding this comment.
I am going to inline the entire Muxing type into Connection once poll_outbound is removed from StreamMuxer. This is already done in #2648 if you want to see what it looks like :)
I am also tempted to inline HandlerWrapper and centralize all the connection handling logic directly in Connection. This would remove quite a bit of indirection because substreams would directly be passed from the StreamMuxer to the ConnectionHandler.
thomaseizinger
changed the title
Into<io::Error> bound from StreamMuxer::ErrorInto<io::Error> bound from StreamMuxer::Error with std::Error
thomaseizinger
changed the title
Into<io::Error> bound from StreamMuxer::Error with std::ErrorInto<io::Error> bound on StreamMuxer with std::error::Error
thomaseizinger
force-pushed
the
refactor/introduce-substream-box
branch
from
cdfa81e
to
f014c29
Compare
thomaseizinger
force-pushed
the
refactor/remove-io-error-bound
branch
from
37a23c4
to
27638ac
Compare
|
Rebase only to stay up to date with #2707. |
This allows us to preserve the type information of a muxer's concrete error as long as possible. For `StreamMuxerBox`, we leverage `io::Error`'s capability of wrapping any error that implements `Into<Box<dyn Error>>`.
thomaseizinger
force-pushed
the
refactor/remove-io-error-bound
branch
from
c2106db
to
a28a7c5
Compare
|
Rebase only includes diff of master since this has been opened: https://github.com/libp2p/rust-libp2p/compare/c2106db78e6af1fc78f33c1f780498c4c5f59a70..a28a7c51eded33886ccd279e64c423afafb5e907 |
Description
This allows us to preserve the type information of a muxer's concrete
error as long as possible. For
StreamMuxerBox, we leverageio::Error'scapability of wrapping any error that implements
Into<Box<dyn Error>>.Links to any relevant issues
This is a stacked PR on top of #2707 and thus a draft but is otherwise ready to review.Open Questions
Change checklist
I have added tests that prove my fix is effective or that my feature works