zai-rs 0.2.0

//! Shared SSE (Server-Sent Events) line parsing utilities.
//!
//! Extracts the common logic of buffering raw byte chunks, splitting on `\n`,
//! trimming `\r\n`, and yielding `data: ` prefixed payload lines.

/// Incremental SSE event parser.
///
/// Unlike [`extract_sse_data_lines`], this parser follows event boundaries and
/// joins multiple `data:` lines in the same event with `\n`.
#[derive(Debug, Default)]
pub struct SseEventParser {
    buf: Vec<u8>,
    event_data: Vec<Vec<u8>>,
}

impl SseEventParser {
    pub fn new() -> Self {
        Self::default()
    }

    /// Push a transport byte chunk and return completed SSE event payloads.
    pub fn push(&mut self, new_bytes: &[u8]) -> Vec<Vec<u8>> {
        self.buf.extend_from_slice(new_bytes);
        let mut events = Vec::new();

        while let Some(newline_index) = self.buf.iter().position(|&b| b == b'\n') {
            let mut line = self.buf.drain(..=newline_index).collect::<Vec<_>>();
            if line.ends_with(b"\n") {
                line.pop();
            }
            if line.ends_with(b"\r") {
                line.pop();
            }

            if line.is_empty() {
                if !self.event_data.is_empty() {
                    events.push(join_event_data(&self.event_data));
                    self.event_data.clear();
                }
                continue;
            }

            if line.starts_with(b":") {
                continue;
            }

            if let Some(rest) = line.strip_prefix(b"data:") {
                self.event_data.push(trim_one_leading_space(rest).to_vec());
            }
        }

        events
    }
}

fn trim_one_leading_space(bytes: &[u8]) -> &[u8] {
    bytes.strip_prefix(b" ").unwrap_or(bytes)
}

fn join_event_data(lines: &[Vec<u8>]) -> Vec<u8> {
    let mut event = Vec::new();
    for (idx, line) in lines.iter().enumerate() {
        if idx > 0 {
            event.push(b'\n');
        }
        event.extend_from_slice(line);
    }
    event
}

/// Process a new chunk of bytes, extract completed SSE data lines.
///
/// Appends `new_bytes` to `buf`, then extracts all complete lines (delimited
/// by `\n`). For each line, it:
/// - Strips trailing `\r` and `\n`
/// - Skips empty lines
/// - Strips the `"data: "` prefix and yields the remaining bytes
///
/// Returns a vector of data payload slices (borrowed from `buf`).
/// Lines that are not prefixed with `"data: "` are silently skipped.
///
/// If a `data: [DONE]` line is encountered, it is yielded as a
/// `[b"[DONE]"]` entry so the caller can detect stream termination.
pub fn extract_sse_data_lines(buf: &mut Vec<u8>, new_bytes: &[u8]) -> Vec<Vec<u8>> {
    buf.extend_from_slice(new_bytes);
    let mut results = Vec::new();

    let Some(last_newline) = buf.iter().rposition(|&b| b == b'\n') else {
        return results;
    };

    let completed = &buf[..=last_newline];
    for line_with_nl in completed.split_inclusive(|&b| b == b'\n') {
        let mut line = line_with_nl;
        if let Some(line_without_nl) = line.strip_suffix(b"\n") {
            line = line_without_nl;
        }
        if let Some(line_without_cr) = line.strip_suffix(b"\r") {
            line = line_without_cr;
        }
        if line.is_empty() {
            continue;
        }
        const PREFIX: &[u8] = b"data: ";
        if let Some(rest) = line.strip_prefix(PREFIX) {
            results.push(rest.to_vec());
        }
    }

    buf.drain(..=last_newline);

    results
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_single_complete_line() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b"data: hello\n");
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0], b"hello");
    }

    #[test]
    fn test_partial_then_complete() {
        let mut buf = Vec::new();
        let lines1 = extract_sse_data_lines(&mut buf, b"data: hel");
        assert!(lines1.is_empty());

        let lines2 = extract_sse_data_lines(&mut buf, b"lo\n");
        assert_eq!(lines2.len(), 1);
        assert_eq!(lines2[0], b"hello");
    }

    #[test]
    fn test_crlf_line_endings() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b"data: world\r\n");
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0], b"world");
    }

    #[test]
    fn test_multiple_events_in_one_chunk() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b"data: first\n\ndata: second\n");
        assert_eq!(lines.len(), 2);
        assert_eq!(lines[0], b"first");
        assert_eq!(lines[1], b"second");
    }

    #[test]
    fn test_done_marker() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b"data: [DONE]\n");
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0], b"[DONE]");
    }

    #[test]
    fn test_non_data_lines_skipped() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b": comment\nid: 123\ndata: payload\n");
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0], b"payload");
    }

    #[test]
    fn test_empty_lines_ignored() {
        let mut buf = Vec::new();
        let lines = extract_sse_data_lines(&mut buf, b"\n\n\ndata: hello\n\n");
        assert_eq!(lines.len(), 1);
        assert_eq!(lines[0], b"hello");
    }

    #[test]
    fn event_parser_yields_complete_events() {
        let mut parser = SseEventParser::new();
        assert!(parser.push(b"data: hel").is_empty());
        assert_eq!(parser.push(b"lo\r\n\r\n"), vec![b"hello".to_vec()]);
    }

    #[test]
    fn event_parser_joins_multi_data_lines() {
        let mut parser = SseEventParser::new();
        let events = parser.push(b"data: {\"a\":\ndata: 1}\n\n");
        assert_eq!(events, vec![b"{\"a\":\n1}".to_vec()]);
    }

    #[test]
    fn event_parser_ignores_comments_and_non_data_fields() {
        let mut parser = SseEventParser::new();
        let events = parser.push(b": keepalive\nid: 1\ndata: payload\n\n");
        assert_eq!(events, vec![b"payload".to_vec()]);
    }
}