# Description This helps to ensure data produced on a stream is immediately available to the consumer of the stream. The BufWriter introduced for performance reasons in 0.93 exposed the behavior that data messages wouldn't make it to the other side until they filled the buffer in @cablehead's [`nu_plugin_from_sse`](https://github.com/cablehead/nu_plugin_from_sse). I had originally not flushed on every `Data` message because I figured that it isn't really critical that the other side sees those messages immediately, since they're not used for control and they are flushed when waiting for acknowledgement or when the buffer is too full anyway. Increasing the amount of data that can be sent with a single underlying write increases performance, but this interferes with some plugins that want to use streams in a more real-time way. In the future I would like to make this configurable, maybe even per-command, so that a command can decide what the priority is. But for now I think this is reasonable. In the worst case, this decreases performance by about 40%, when sending very small values (just numbers). But for larger values, this PR actually increases performance by about 20%, because I've increased the buffer size about 2x to 16,384 bytes. The previous value of 8,192 bytes was too small to fit a full buffer coming from an external command, so doubling it makes sense, and now a write of a buffer from an external command can be done in exactly one write call, which I think makes sense. I'm doing this at the same time because flushing each data message would make it very likely that each individual data message from an external stream would require exactly two writes rather than approximately one (amortized). Again, hopefully the tradeoff isn't too bad, and if it is I'll just make it configurable. # User-Facing Changes - Performance of plugin streams will be a bit different - Plugins that expect to send streams in real-time will work again # Tests + Formatting - 🟢 `toolkit fmt` - 🟢 `toolkit clippy` - 🟢 `toolkit test` - 🟢 `toolkit test stdlib`
632 lines
22 KiB
Rust
632 lines
22 KiB
Rust
use nu_plugin_protocol::{StreamData, StreamId, StreamMessage};
|
|
use nu_protocol::{ShellError, Span, Value};
|
|
use std::{
|
|
collections::{btree_map, BTreeMap},
|
|
iter::FusedIterator,
|
|
marker::PhantomData,
|
|
sync::{mpsc, Arc, Condvar, Mutex, MutexGuard, Weak},
|
|
};
|
|
|
|
#[cfg(test)]
|
|
mod tests;
|
|
|
|
/// Receives messages from a stream read from input by a [`StreamManager`].
|
|
///
|
|
/// The receiver reads for messages of type `Result<Option<StreamData>, ShellError>` from the
|
|
/// channel, which is managed by a [`StreamManager`]. Signalling for end-of-stream is explicit
|
|
/// through `Ok(Some)`.
|
|
///
|
|
/// Failing to receive is an error. When end-of-stream is received, the `receiver` is set to `None`
|
|
/// and all further calls to `next()` return `None`.
|
|
///
|
|
/// The type `T` must implement [`FromShellError`], so that errors in the stream can be represented,
|
|
/// and `TryFrom<StreamData>` to convert it to the correct type.
|
|
///
|
|
/// For each message read, it sends [`StreamMessage::Ack`] to the writer. When dropped,
|
|
/// it sends [`StreamMessage::Drop`].
|
|
#[derive(Debug)]
|
|
pub struct StreamReader<T, W>
|
|
where
|
|
W: WriteStreamMessage,
|
|
{
|
|
id: StreamId,
|
|
receiver: Option<mpsc::Receiver<Result<Option<StreamData>, ShellError>>>,
|
|
writer: W,
|
|
/// Iterator requires the item type to be fixed, so we have to keep it as part of the type,
|
|
/// even though we're actually receiving dynamic data.
|
|
marker: PhantomData<fn() -> T>,
|
|
}
|
|
|
|
impl<T, W> StreamReader<T, W>
|
|
where
|
|
T: TryFrom<StreamData, Error = ShellError>,
|
|
W: WriteStreamMessage,
|
|
{
|
|
/// Create a new StreamReader from parts
|
|
fn new(
|
|
id: StreamId,
|
|
receiver: mpsc::Receiver<Result<Option<StreamData>, ShellError>>,
|
|
writer: W,
|
|
) -> StreamReader<T, W> {
|
|
StreamReader {
|
|
id,
|
|
receiver: Some(receiver),
|
|
writer,
|
|
marker: PhantomData,
|
|
}
|
|
}
|
|
|
|
/// Receive a message from the channel, or return an error if:
|
|
///
|
|
/// * the channel couldn't be received from
|
|
/// * an error was sent on the channel
|
|
/// * the message received couldn't be converted to `T`
|
|
pub fn recv(&mut self) -> Result<Option<T>, ShellError> {
|
|
let connection_lost = || ShellError::GenericError {
|
|
error: "Stream ended unexpectedly".into(),
|
|
msg: "connection lost before explicit end of stream".into(),
|
|
span: None,
|
|
help: None,
|
|
inner: vec![],
|
|
};
|
|
|
|
if let Some(ref rx) = self.receiver {
|
|
// Try to receive a message first
|
|
let msg = match rx.try_recv() {
|
|
Ok(msg) => msg?,
|
|
Err(mpsc::TryRecvError::Empty) => {
|
|
// The receiver doesn't have any messages waiting for us. It's possible that the
|
|
// other side hasn't seen our acknowledgements. Let's flush the writer and then
|
|
// wait
|
|
self.writer.flush()?;
|
|
rx.recv().map_err(|_| connection_lost())??
|
|
}
|
|
Err(mpsc::TryRecvError::Disconnected) => return Err(connection_lost()),
|
|
};
|
|
|
|
if let Some(data) = msg {
|
|
// Acknowledge the message
|
|
self.writer
|
|
.write_stream_message(StreamMessage::Ack(self.id))?;
|
|
// Try to convert it into the correct type
|
|
Ok(Some(data.try_into()?))
|
|
} else {
|
|
// Remove the receiver, so that future recv() calls always return Ok(None)
|
|
self.receiver = None;
|
|
Ok(None)
|
|
}
|
|
} else {
|
|
// Closed already
|
|
Ok(None)
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<T, W> Iterator for StreamReader<T, W>
|
|
where
|
|
T: FromShellError + TryFrom<StreamData, Error = ShellError>,
|
|
W: WriteStreamMessage,
|
|
{
|
|
type Item = T;
|
|
|
|
fn next(&mut self) -> Option<T> {
|
|
// Converting the error to the value here makes the implementation a lot easier
|
|
match self.recv() {
|
|
Ok(option) => option,
|
|
Err(err) => {
|
|
// Drop the receiver so we don't keep returning errors
|
|
self.receiver = None;
|
|
Some(T::from_shell_error(err))
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Guaranteed not to return anything after the end
|
|
impl<T, W> FusedIterator for StreamReader<T, W>
|
|
where
|
|
T: FromShellError + TryFrom<StreamData, Error = ShellError>,
|
|
W: WriteStreamMessage,
|
|
{
|
|
}
|
|
|
|
impl<T, W> Drop for StreamReader<T, W>
|
|
where
|
|
W: WriteStreamMessage,
|
|
{
|
|
fn drop(&mut self) {
|
|
if let Err(err) = self
|
|
.writer
|
|
.write_stream_message(StreamMessage::Drop(self.id))
|
|
.and_then(|_| self.writer.flush())
|
|
{
|
|
log::warn!("Failed to send message to drop stream: {err}");
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Values that can contain a `ShellError` to signal an error has occurred.
|
|
pub trait FromShellError {
|
|
fn from_shell_error(err: ShellError) -> Self;
|
|
}
|
|
|
|
// For List streams.
|
|
impl FromShellError for Value {
|
|
fn from_shell_error(err: ShellError) -> Self {
|
|
Value::error(err, Span::unknown())
|
|
}
|
|
}
|
|
|
|
// For Raw streams, mostly.
|
|
impl<T> FromShellError for Result<T, ShellError> {
|
|
fn from_shell_error(err: ShellError) -> Self {
|
|
Err(err)
|
|
}
|
|
}
|
|
|
|
/// Writes messages to a stream, with flow control.
|
|
///
|
|
/// The `signal` contained
|
|
#[derive(Debug)]
|
|
pub struct StreamWriter<W: WriteStreamMessage> {
|
|
id: StreamId,
|
|
signal: Arc<StreamWriterSignal>,
|
|
writer: W,
|
|
ended: bool,
|
|
}
|
|
|
|
impl<W> StreamWriter<W>
|
|
where
|
|
W: WriteStreamMessage,
|
|
{
|
|
fn new(id: StreamId, signal: Arc<StreamWriterSignal>, writer: W) -> StreamWriter<W> {
|
|
StreamWriter {
|
|
id,
|
|
signal,
|
|
writer,
|
|
ended: false,
|
|
}
|
|
}
|
|
|
|
/// Check if the stream was dropped from the other end. Recommended to do this before calling
|
|
/// [`.write()`], especially in a loop.
|
|
pub fn is_dropped(&self) -> Result<bool, ShellError> {
|
|
self.signal.is_dropped()
|
|
}
|
|
|
|
/// Write a single piece of data to the stream.
|
|
///
|
|
/// Error if something failed with the write, or if [`.end()`] was already called
|
|
/// previously.
|
|
pub fn write(&mut self, data: impl Into<StreamData>) -> Result<(), ShellError> {
|
|
if !self.ended {
|
|
self.writer
|
|
.write_stream_message(StreamMessage::Data(self.id, data.into()))?;
|
|
// Flush after each data message to ensure they do predictably appear on the other side
|
|
// when they're generated
|
|
//
|
|
// TODO: make the buffering configurable, as this is a factor for performance
|
|
self.writer.flush()?;
|
|
// This implements flow control, so we don't write too many messages:
|
|
if !self.signal.notify_sent()? {
|
|
self.signal.wait_for_drain()
|
|
} else {
|
|
Ok(())
|
|
}
|
|
} else {
|
|
Err(ShellError::GenericError {
|
|
error: "Wrote to a stream after it ended".into(),
|
|
msg: format!(
|
|
"tried to write to stream {} after it was already ended",
|
|
self.id
|
|
),
|
|
span: None,
|
|
help: Some("this may be a bug in the nu-plugin crate".into()),
|
|
inner: vec![],
|
|
})
|
|
}
|
|
}
|
|
|
|
/// Write a full iterator to the stream. Note that this doesn't end the stream, so you should
|
|
/// still call [`.end()`].
|
|
///
|
|
/// If the stream is dropped from the other end, the iterator will not be fully consumed, and
|
|
/// writing will terminate.
|
|
///
|
|
/// Returns `Ok(true)` if the iterator was fully consumed, or `Ok(false)` if a drop interrupted
|
|
/// the stream from the other side.
|
|
pub fn write_all<T>(&mut self, data: impl IntoIterator<Item = T>) -> Result<bool, ShellError>
|
|
where
|
|
T: Into<StreamData>,
|
|
{
|
|
// Check before starting
|
|
if self.is_dropped()? {
|
|
return Ok(false);
|
|
}
|
|
|
|
for item in data {
|
|
// Check again after each item is consumed from the iterator, just in case the iterator
|
|
// takes a while to produce a value
|
|
if self.is_dropped()? {
|
|
return Ok(false);
|
|
}
|
|
self.write(item)?;
|
|
}
|
|
Ok(true)
|
|
}
|
|
|
|
/// End the stream. Recommend doing this instead of relying on `Drop` so that you can catch the
|
|
/// error.
|
|
pub fn end(&mut self) -> Result<(), ShellError> {
|
|
if !self.ended {
|
|
// Set the flag first so we don't double-report in the Drop
|
|
self.ended = true;
|
|
self.writer
|
|
.write_stream_message(StreamMessage::End(self.id))?;
|
|
self.writer.flush()
|
|
} else {
|
|
Ok(())
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<W> Drop for StreamWriter<W>
|
|
where
|
|
W: WriteStreamMessage,
|
|
{
|
|
fn drop(&mut self) {
|
|
// Make sure we ended the stream
|
|
if let Err(err) = self.end() {
|
|
log::warn!("Error while ending stream in Drop for StreamWriter: {err}");
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Stores stream state for a writer, and can be blocked on to wait for messages to be acknowledged.
|
|
/// A key part of managing stream lifecycle and flow control.
|
|
#[derive(Debug)]
|
|
pub struct StreamWriterSignal {
|
|
mutex: Mutex<StreamWriterSignalState>,
|
|
change_cond: Condvar,
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
pub struct StreamWriterSignalState {
|
|
/// Stream has been dropped and consumer is no longer interested in any messages.
|
|
dropped: bool,
|
|
/// Number of messages that have been sent without acknowledgement.
|
|
unacknowledged: i32,
|
|
/// Max number of messages to send before waiting for acknowledgement.
|
|
high_pressure_mark: i32,
|
|
}
|
|
|
|
impl StreamWriterSignal {
|
|
/// Create a new signal.
|
|
///
|
|
/// If `notify_sent()` is called more than `high_pressure_mark` times, it will wait until
|
|
/// `notify_acknowledge()` is called by another thread enough times to bring the number of
|
|
/// unacknowledged sent messages below that threshold.
|
|
fn new(high_pressure_mark: i32) -> StreamWriterSignal {
|
|
assert!(high_pressure_mark > 0);
|
|
|
|
StreamWriterSignal {
|
|
mutex: Mutex::new(StreamWriterSignalState {
|
|
dropped: false,
|
|
unacknowledged: 0,
|
|
high_pressure_mark,
|
|
}),
|
|
change_cond: Condvar::new(),
|
|
}
|
|
}
|
|
|
|
fn lock(&self) -> Result<MutexGuard<StreamWriterSignalState>, ShellError> {
|
|
self.mutex.lock().map_err(|_| ShellError::NushellFailed {
|
|
msg: "StreamWriterSignal mutex poisoned due to panic".into(),
|
|
})
|
|
}
|
|
|
|
/// True if the stream was dropped and the consumer is no longer interested in it. Indicates
|
|
/// that no more messages should be sent, other than `End`.
|
|
pub fn is_dropped(&self) -> Result<bool, ShellError> {
|
|
Ok(self.lock()?.dropped)
|
|
}
|
|
|
|
/// Notify the writers that the stream has been dropped, so they can stop writing.
|
|
pub fn set_dropped(&self) -> Result<(), ShellError> {
|
|
let mut state = self.lock()?;
|
|
state.dropped = true;
|
|
// Unblock the writers so they can terminate
|
|
self.change_cond.notify_all();
|
|
Ok(())
|
|
}
|
|
|
|
/// Track that a message has been sent. Returns `Ok(true)` if more messages can be sent,
|
|
/// or `Ok(false)` if the high pressure mark has been reached and [`.wait_for_drain()`] should
|
|
/// be called to block.
|
|
pub fn notify_sent(&self) -> Result<bool, ShellError> {
|
|
let mut state = self.lock()?;
|
|
state.unacknowledged =
|
|
state
|
|
.unacknowledged
|
|
.checked_add(1)
|
|
.ok_or_else(|| ShellError::NushellFailed {
|
|
msg: "Overflow in counter: too many unacknowledged messages".into(),
|
|
})?;
|
|
|
|
Ok(state.unacknowledged < state.high_pressure_mark)
|
|
}
|
|
|
|
/// Wait for acknowledgements before sending more data. Also returns if the stream is dropped.
|
|
pub fn wait_for_drain(&self) -> Result<(), ShellError> {
|
|
let mut state = self.lock()?;
|
|
while !state.dropped && state.unacknowledged >= state.high_pressure_mark {
|
|
state = self
|
|
.change_cond
|
|
.wait(state)
|
|
.map_err(|_| ShellError::NushellFailed {
|
|
msg: "StreamWriterSignal mutex poisoned due to panic".into(),
|
|
})?;
|
|
}
|
|
Ok(())
|
|
}
|
|
|
|
/// Notify the writers that a message has been acknowledged, so they can continue to write
|
|
/// if they were waiting.
|
|
pub fn notify_acknowledged(&self) -> Result<(), ShellError> {
|
|
let mut state = self.lock()?;
|
|
state.unacknowledged =
|
|
state
|
|
.unacknowledged
|
|
.checked_sub(1)
|
|
.ok_or_else(|| ShellError::NushellFailed {
|
|
msg: "Underflow in counter: too many message acknowledgements".into(),
|
|
})?;
|
|
// Unblock the writer
|
|
self.change_cond.notify_one();
|
|
Ok(())
|
|
}
|
|
}
|
|
|
|
/// A sink for a [`StreamMessage`]
|
|
pub trait WriteStreamMessage {
|
|
fn write_stream_message(&mut self, msg: StreamMessage) -> Result<(), ShellError>;
|
|
fn flush(&mut self) -> Result<(), ShellError>;
|
|
}
|
|
|
|
#[derive(Debug, Default)]
|
|
struct StreamManagerState {
|
|
reading_streams: BTreeMap<StreamId, mpsc::Sender<Result<Option<StreamData>, ShellError>>>,
|
|
writing_streams: BTreeMap<StreamId, Weak<StreamWriterSignal>>,
|
|
}
|
|
|
|
impl StreamManagerState {
|
|
/// Lock the state, or return a [`ShellError`] if the mutex is poisoned.
|
|
fn lock(
|
|
state: &Mutex<StreamManagerState>,
|
|
) -> Result<MutexGuard<StreamManagerState>, ShellError> {
|
|
state.lock().map_err(|_| ShellError::NushellFailed {
|
|
msg: "StreamManagerState mutex poisoned due to a panic".into(),
|
|
})
|
|
}
|
|
}
|
|
|
|
#[derive(Debug)]
|
|
pub struct StreamManager {
|
|
state: Arc<Mutex<StreamManagerState>>,
|
|
}
|
|
|
|
impl StreamManager {
|
|
/// Create a new StreamManager.
|
|
pub fn new() -> StreamManager {
|
|
StreamManager {
|
|
state: Default::default(),
|
|
}
|
|
}
|
|
|
|
fn lock(&self) -> Result<MutexGuard<StreamManagerState>, ShellError> {
|
|
StreamManagerState::lock(&self.state)
|
|
}
|
|
|
|
/// Create a new handle to the StreamManager for registering streams.
|
|
pub fn get_handle(&self) -> StreamManagerHandle {
|
|
StreamManagerHandle {
|
|
state: Arc::downgrade(&self.state),
|
|
}
|
|
}
|
|
|
|
/// Process a stream message, and update internal state accordingly.
|
|
pub fn handle_message(&self, message: StreamMessage) -> Result<(), ShellError> {
|
|
let mut state = self.lock()?;
|
|
match message {
|
|
StreamMessage::Data(id, data) => {
|
|
if let Some(sender) = state.reading_streams.get(&id) {
|
|
// We should ignore the error on send. This just means the reader has dropped,
|
|
// but it will have sent a Drop message to the other side, and we will receive
|
|
// an End message at which point we can remove the channel.
|
|
let _ = sender.send(Ok(Some(data)));
|
|
Ok(())
|
|
} else {
|
|
Err(ShellError::PluginFailedToDecode {
|
|
msg: format!("received Data for unknown stream {id}"),
|
|
})
|
|
}
|
|
}
|
|
StreamMessage::End(id) => {
|
|
if let Some(sender) = state.reading_streams.remove(&id) {
|
|
// We should ignore the error on the send, because the reader might have dropped
|
|
// already
|
|
let _ = sender.send(Ok(None));
|
|
Ok(())
|
|
} else {
|
|
Err(ShellError::PluginFailedToDecode {
|
|
msg: format!("received End for unknown stream {id}"),
|
|
})
|
|
}
|
|
}
|
|
StreamMessage::Drop(id) => {
|
|
if let Some(signal) = state.writing_streams.remove(&id) {
|
|
if let Some(signal) = signal.upgrade() {
|
|
// This will wake blocked writers so they can stop writing, so it's ok
|
|
signal.set_dropped()?;
|
|
}
|
|
}
|
|
// It's possible that the stream has already finished writing and we don't have it
|
|
// anymore, so we fall through to Ok
|
|
Ok(())
|
|
}
|
|
StreamMessage::Ack(id) => {
|
|
if let Some(signal) = state.writing_streams.get(&id) {
|
|
if let Some(signal) = signal.upgrade() {
|
|
// This will wake up a blocked writer
|
|
signal.notify_acknowledged()?;
|
|
} else {
|
|
// We know it doesn't exist, so might as well remove it
|
|
state.writing_streams.remove(&id);
|
|
}
|
|
}
|
|
// It's possible that the stream has already finished writing and we don't have it
|
|
// anymore, so we fall through to Ok
|
|
Ok(())
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Broadcast an error to all stream readers. This is useful for error propagation.
|
|
pub fn broadcast_read_error(&self, error: ShellError) -> Result<(), ShellError> {
|
|
let state = self.lock()?;
|
|
for channel in state.reading_streams.values() {
|
|
// Ignore send errors.
|
|
let _ = channel.send(Err(error.clone()));
|
|
}
|
|
Ok(())
|
|
}
|
|
|
|
// If the `StreamManager` is dropped, we should let all of the stream writers know that they
|
|
// won't be able to write anymore. We don't need to do anything about the readers though
|
|
// because they'll know when the `Sender` is dropped automatically
|
|
fn drop_all_writers(&self) -> Result<(), ShellError> {
|
|
let mut state = self.lock()?;
|
|
let writers = std::mem::take(&mut state.writing_streams);
|
|
for (_, signal) in writers {
|
|
if let Some(signal) = signal.upgrade() {
|
|
// more important that we send to all than handling an error
|
|
let _ = signal.set_dropped();
|
|
}
|
|
}
|
|
Ok(())
|
|
}
|
|
}
|
|
|
|
impl Default for StreamManager {
|
|
fn default() -> Self {
|
|
Self::new()
|
|
}
|
|
}
|
|
|
|
impl Drop for StreamManager {
|
|
fn drop(&mut self) {
|
|
if let Err(err) = self.drop_all_writers() {
|
|
log::warn!("error during Drop for StreamManager: {}", err)
|
|
}
|
|
}
|
|
}
|
|
|
|
/// A [`StreamManagerHandle`] supports operations for interacting with the [`StreamManager`].
|
|
///
|
|
/// Streams can be registered for reading, returning a [`StreamReader`], or for writing, returning
|
|
/// a [`StreamWriter`].
|
|
#[derive(Debug, Clone)]
|
|
pub struct StreamManagerHandle {
|
|
state: Weak<Mutex<StreamManagerState>>,
|
|
}
|
|
|
|
impl StreamManagerHandle {
|
|
/// Because the handle only has a weak reference to the [`StreamManager`] state, we have to
|
|
/// first try to upgrade to a strong reference and then lock. This function wraps those two
|
|
/// operations together, handling errors appropriately.
|
|
fn with_lock<T, F>(&self, f: F) -> Result<T, ShellError>
|
|
where
|
|
F: FnOnce(MutexGuard<StreamManagerState>) -> Result<T, ShellError>,
|
|
{
|
|
let upgraded = self
|
|
.state
|
|
.upgrade()
|
|
.ok_or_else(|| ShellError::NushellFailed {
|
|
msg: "StreamManager is no longer alive".into(),
|
|
})?;
|
|
let guard = upgraded.lock().map_err(|_| ShellError::NushellFailed {
|
|
msg: "StreamManagerState mutex poisoned due to a panic".into(),
|
|
})?;
|
|
f(guard)
|
|
}
|
|
|
|
/// Register a new stream for reading, and return a [`StreamReader`] that can be used to iterate
|
|
/// on the values received. A [`StreamMessage`] writer is required for writing control messages
|
|
/// back to the producer.
|
|
pub fn read_stream<T, W>(
|
|
&self,
|
|
id: StreamId,
|
|
writer: W,
|
|
) -> Result<StreamReader<T, W>, ShellError>
|
|
where
|
|
T: TryFrom<StreamData, Error = ShellError>,
|
|
W: WriteStreamMessage,
|
|
{
|
|
let (tx, rx) = mpsc::channel();
|
|
self.with_lock(|mut state| {
|
|
// Must be exclusive
|
|
if let btree_map::Entry::Vacant(e) = state.reading_streams.entry(id) {
|
|
e.insert(tx);
|
|
Ok(())
|
|
} else {
|
|
Err(ShellError::GenericError {
|
|
error: format!("Failed to acquire reader for stream {id}"),
|
|
msg: "tried to get a reader for a stream that's already being read".into(),
|
|
span: None,
|
|
help: Some("this may be a bug in the nu-plugin crate".into()),
|
|
inner: vec![],
|
|
})
|
|
}
|
|
})?;
|
|
Ok(StreamReader::new(id, rx, writer))
|
|
}
|
|
|
|
/// Register a new stream for writing, and return a [`StreamWriter`] that can be used to send
|
|
/// data to the stream.
|
|
///
|
|
/// The `high_pressure_mark` value controls how many messages can be written without receiving
|
|
/// an acknowledgement before any further attempts to write will wait for the consumer to
|
|
/// acknowledge them. This prevents overwhelming the reader.
|
|
pub fn write_stream<W>(
|
|
&self,
|
|
id: StreamId,
|
|
writer: W,
|
|
high_pressure_mark: i32,
|
|
) -> Result<StreamWriter<W>, ShellError>
|
|
where
|
|
W: WriteStreamMessage,
|
|
{
|
|
let signal = Arc::new(StreamWriterSignal::new(high_pressure_mark));
|
|
self.with_lock(|mut state| {
|
|
// Remove dead writing streams
|
|
state
|
|
.writing_streams
|
|
.retain(|_, signal| signal.strong_count() > 0);
|
|
// Must be exclusive
|
|
if let btree_map::Entry::Vacant(e) = state.writing_streams.entry(id) {
|
|
e.insert(Arc::downgrade(&signal));
|
|
Ok(())
|
|
} else {
|
|
Err(ShellError::GenericError {
|
|
error: format!("Failed to acquire writer for stream {id}"),
|
|
msg: "tried to get a writer for a stream that's already being written".into(),
|
|
span: None,
|
|
help: Some("this may be a bug in the nu-plugin crate".into()),
|
|
inner: vec![],
|
|
})
|
|
}
|
|
})?;
|
|
Ok(StreamWriter::new(id, signal, writer))
|
|
}
|
|
}
|