Could crates.io require updating with the master branch on GitHub?

[DouglasGray]

Hello, I'm new to Rust so I'm probably just a bit confused but I was looking through the docs for what appears to be the latest version of tokio-tungstenite on crates.io (v0.6.0), specifically the source for WebSocketStream in lib.rs (https://docs.rs/tokio-tungstenite/0.6.0/src/tokio_tungstenite/lib.rs.html#162-164), and it appears that the code for WebSocketStream is a fair bit different to that on the GitHub master branch. For example, comparing the struct and the poll function in 'impl Stream for WebSocketStream':

(crates.io source)

pub struct WebSocketStream<S> {
    inner: WebSocket<S>,
}

(GitHub source)

pub struct WebSocketStream<S> {
    inner: WebSocket<S>,
    stream_ended: bool,
}

(crates.io source)

fn poll(&mut self) -> Poll<Option<Message>, WsError> {
        self.inner.read_message().map(|m| Some(m)).to_async()
}

(GitHub source)

fn poll(&mut self) -> Poll<Option<Message>, WsError> {
        if self.stream_ended {
            self.stream_ended = false;
            return Ok(Async::Ready(None))
        }

        self.inner.read_message().map(|m| {
            if m.is_close() {
                self.stream_ended = true;
            }
            Some(m)
        }).to_async()
}

At least for me I can confirm that when I add 'tokio-tungstenite = "*"' to my Cargo.toml it uses the code that I can see on crates.io. Is this the case for anyone else?

(Edit: changed format for code blocks)

[DouglasGray]

Of course I understand that the master branch and what's officially out there can be out of sync, but it seems that the change to signal the end of the stream is a fairly important one? Unless I'm missing something in the source that I see on crates.io (which is very likely)

[daniel-abramov]

Hi, yes, the master branch is the current (active) state, the latest versions of tokio-tungstenite and tungstenite has not been published to crates.io yet.

At least for me I can confirm that when I add 'tokio-tungstenite = "*"' to my Cargo.toml it uses the code that I can see on crates.io. Is this the case for anyone else?

Yes. The version from master branch has not yet been published. The changes you're referring to have been made to implement some improvements which were suggested to us, namely we changed the semantics for the Stream implementation so that it corresponds to the "expected" behavior of Tokio, i.e. a normal close will not result in an error on Stream anymore, but return a None. And the Message enum has been extended by an additional Close message, so you still have the possibility to analyze the close code in a newest version (master branch) and your Stream will return None when the connection is closed in a normal way (and not an Err like it did before).

Some people already mentioned in multiple comments that we have to publish tungstenite-rs and tokio-tungstenite to crates.io, I pinged @agalakhov to do that. Perhaps he has forgotten to do that.

[DouglasGray]

Ok gotcha, thank you for explaining!