rust-cli / rexpect

.github/workflows/ci.yml
https://docs.rs/rexpect
MIT License
330 stars 58 forks source link

Resuming read after timeout #125

Open amosjyng opened 2 months ago

amosjyng commented 2 months ago

I have a

pub session: Option<Arc<Mutex<PtySession>>>,

I want to get the existing output of a process after 100ms. This seemed to work at first:

        let output = {
            let session_mutex = self.session.as_mut().ok_or(anyhow!("No session"))?;
            let mut session = session_mutex.lock()?;
            match session.exp_eof() {
                Ok(output) => output,
                Err(e) => {
                    match e {
                        rexpect::error::Error::Timeout { got, .. } => got
                            .replace("`\\n`\n", "\n")
                            .replace("`\\r`", "\r")
                            .replace("`^`", "\u{1b}"),
                        _ => return Err(e.into()),
                    }
                }
            }
        };

but then I realized that session.exp_eof does not drain the buffer if a timeout is reached, meaning that if I try to read the next chunk, the entire output up to this point is going to be returned yet again. This seems to work:

    fn drain_read_buffer(&mut self) -> ZammResult<String> {
        let mut output = String::new();
        if let Some(session) = self.session.as_mut() {
            let reader = &mut session.lock()?.reader;
            while let Some(chunk) = reader.try_read() {
                output.push(chunk);
            }
        }
        Ok(output)
    }

    ...

        let mut output = String::new();
        loop {
            sleep(Duration::from_millis(100));
            let new_output = self.drain_read_buffer()?;
            if new_output.is_empty() {
                break;
            }
            output.push_str(&new_output);
        }

But I'm wondering, is there a better way to implement "read whatever is available after 100ms" than this?