Struct LogWriter 

pub struct LogWriter<C> { /* private fields */ }

Background-task-fronted log writer.

Construction (via write_agent_completion etc.) spawns a tokio task that owns the per-stream [LogWriterState]. The handle here is a thin sender + JoinHandle pair.

Implementations

impl<C> LogWriter<C>

pub fn write(&self, chunk: C) -> Result<(), Error>

Hand off one chunk to the listener task. Returns Err(Error::Instance(_)) only when the listener has already exited — typically because an earlier DB write failed. The caller should treat that error the same way it would treat an upstream stream error: stop reading, surface upward.

pub fn written_once(&self) -> bool

Sync peek: has the listener completed at least one successful apply_chunk batch? Flips false → true exactly once, immediately after the first batch’s write completes and before the listener parks on the next recv.

pub async fn wait_written_once(&self) -> Result<(), Error>

Async wait that resolves once the listener has completed its first successful apply_chunk batch. Returns immediately if that already happened. Errors only if the listener task exited before its first successful write (DB error on the very first batch).

pub async fn finalize(self) -> Result<(), Error>

Consume the writer. Drops the sender (signaling EOF to the listener) and awaits the task. Returns only once:

• the channel is empty: recv() returns None only after the listener has drained every queued chunk, AND
• no work is in flight: the task’s future has fully completed, so no row-bucket joins or blob writes remain pending.

Surfaces the first DB error the task encountered, if any.