# Description This PR adds a new method to `EngineInterface`: `register_ctrlc_handler` which takes a closure to run when the plugin's driving engine receives a ctrlc-signal. It also adds a mirror of the `signals` attribute from the main shell `EngineState`. This is an example of how a plugin which makes a long poll http request can end the request on ctrlc: https://github.com/cablehead/nu_plugin_http/blob/main/src/commands/request.rs#L68-L77 To facilitate the feature, a new attribute has been added to `EngineState`: `ctrlc_handlers`. This is a Vec of closures that will be run when the engine's process receives a ctrlc signal. When plugins are added to an `engine_state` during a `merge_delta`, the engine passes the ctrlc_handlers to the plugin's `.configure_ctrlc_handler` method, which gives the plugin a chance to register a handler that sends a ctrlc packet through the `PluginInterface`, if an instance of the plugin is currently running. On the plugin side: `EngineInterface` also has a ctrlc_handlers Vec of closures. Plugin calls can use `register_ctrlc_handler` to register a closure that will be called in the plugin process when the PluginInput::Ctrlc command is received. For future reference these are some alternate places that were investigated for tying the ctrlc trigger to transmitting a Ctrlc packet through the `PluginInterface`: - Directly from `src/signals.rs`: the handler there would need a reference to the Vec<Arc<RegisteredPlugins>>, which would require us to wrap the plugins in a Mutex, which we don't want to do. - have `PersistentPlugin.get_plugin` pass down the engine's CtrlcHandlers to .get and then to .spawn (if the plugin isn't already running). Once we have CtrlcHandlers in spawn, we can register a handler to write directly to PluginInterface. We don't want to double down on passing engine_state to spawn this way though, as it's unpredictable because it would depend on whether the plugin has already been spawned or not. - pass `ctrlc_handlers` to PersistentPlugin::new so it can store it on itself so it's available to spawn. - in `PersistentPlugin.spawn`, create a handler that sends to a clone of the GC event loop's tx. this has the same issues with regards to how to get CtrlcHandlers to the spawn method, and is more complicated than a handler that writes directly to PluginInterface # User-Facing Changes No breaking changes --------- Co-authored-by: Ian Manske <ian.manske@pm.me>
355 lines
12 KiB
Rust
355 lines
12 KiB
Rust
//! Implements the stream multiplexing interface for both the plugin side and the engine side.
|
|
|
|
use nu_plugin_protocol::{ByteStreamInfo, ListStreamInfo, PipelineDataHeader, StreamMessage};
|
|
use nu_protocol::{
|
|
engine::Sequence, ByteStream, IntoSpanned, ListStream, PipelineData, Reader, ShellError,
|
|
Signals,
|
|
};
|
|
use std::{
|
|
io::{Read, Write},
|
|
sync::Mutex,
|
|
thread,
|
|
};
|
|
|
|
pub mod stream;
|
|
|
|
use crate::Encoder;
|
|
|
|
use self::stream::{StreamManager, StreamManagerHandle, StreamWriter, WriteStreamMessage};
|
|
|
|
pub mod test_util;
|
|
|
|
#[cfg(test)]
|
|
mod tests;
|
|
|
|
/// The maximum number of list stream values to send without acknowledgement. This should be tuned
|
|
/// with consideration for memory usage.
|
|
const LIST_STREAM_HIGH_PRESSURE: i32 = 100;
|
|
|
|
/// The maximum number of raw stream buffers to send without acknowledgement. This should be tuned
|
|
/// with consideration for memory usage.
|
|
const RAW_STREAM_HIGH_PRESSURE: i32 = 50;
|
|
|
|
/// Read input/output from the stream.
|
|
pub trait PluginRead<T> {
|
|
/// Returns `Ok(None)` on end of stream.
|
|
fn read(&mut self) -> Result<Option<T>, ShellError>;
|
|
}
|
|
|
|
impl<R, E, T> PluginRead<T> for (R, E)
|
|
where
|
|
R: std::io::BufRead,
|
|
E: Encoder<T>,
|
|
{
|
|
fn read(&mut self) -> Result<Option<T>, ShellError> {
|
|
self.1.decode(&mut self.0)
|
|
}
|
|
}
|
|
|
|
impl<R, T> PluginRead<T> for &mut R
|
|
where
|
|
R: PluginRead<T>,
|
|
{
|
|
fn read(&mut self) -> Result<Option<T>, ShellError> {
|
|
(**self).read()
|
|
}
|
|
}
|
|
|
|
/// Write input/output to the stream.
|
|
///
|
|
/// The write should be atomic, without interference from other threads.
|
|
pub trait PluginWrite<T>: Send + Sync {
|
|
fn write(&self, data: &T) -> Result<(), ShellError>;
|
|
|
|
/// Flush any internal buffers, if applicable.
|
|
fn flush(&self) -> Result<(), ShellError>;
|
|
|
|
/// True if this output is stdout, so that plugins can avoid using stdout for their own purpose
|
|
fn is_stdout(&self) -> bool {
|
|
false
|
|
}
|
|
}
|
|
|
|
impl<E, T> PluginWrite<T> for (std::io::Stdout, E)
|
|
where
|
|
E: Encoder<T>,
|
|
{
|
|
fn write(&self, data: &T) -> Result<(), ShellError> {
|
|
let mut lock = self.0.lock();
|
|
self.1.encode(data, &mut lock)
|
|
}
|
|
|
|
fn flush(&self) -> Result<(), ShellError> {
|
|
self.0.lock().flush().map_err(|err| ShellError::IOError {
|
|
msg: err.to_string(),
|
|
})
|
|
}
|
|
|
|
fn is_stdout(&self) -> bool {
|
|
true
|
|
}
|
|
}
|
|
|
|
impl<W, E, T> PluginWrite<T> for (Mutex<W>, E)
|
|
where
|
|
W: std::io::Write + Send,
|
|
E: Encoder<T>,
|
|
{
|
|
fn write(&self, data: &T) -> Result<(), ShellError> {
|
|
let mut lock = self.0.lock().map_err(|_| ShellError::NushellFailed {
|
|
msg: "writer mutex poisoned".into(),
|
|
})?;
|
|
self.1.encode(data, &mut *lock)
|
|
}
|
|
|
|
fn flush(&self) -> Result<(), ShellError> {
|
|
let mut lock = self.0.lock().map_err(|_| ShellError::NushellFailed {
|
|
msg: "writer mutex poisoned".into(),
|
|
})?;
|
|
lock.flush().map_err(|err| ShellError::IOError {
|
|
msg: err.to_string(),
|
|
})
|
|
}
|
|
}
|
|
|
|
impl<W, T> PluginWrite<T> for &W
|
|
where
|
|
W: PluginWrite<T>,
|
|
{
|
|
fn write(&self, data: &T) -> Result<(), ShellError> {
|
|
(**self).write(data)
|
|
}
|
|
|
|
fn flush(&self) -> Result<(), ShellError> {
|
|
(**self).flush()
|
|
}
|
|
|
|
fn is_stdout(&self) -> bool {
|
|
(**self).is_stdout()
|
|
}
|
|
}
|
|
|
|
/// An interface manager handles I/O and state management for communication between a plugin and
|
|
/// the engine. See `PluginInterfaceManager` in `nu-plugin-engine` for communication from the engine
|
|
/// side to a plugin, or `EngineInterfaceManager` in `nu-plugin` for communication from the plugin
|
|
/// side to the engine.
|
|
///
|
|
/// There is typically one [`InterfaceManager`] consuming input from a background thread, and
|
|
/// managing shared state.
|
|
pub trait InterfaceManager {
|
|
/// The corresponding interface type.
|
|
type Interface: Interface + 'static;
|
|
|
|
/// The input message type.
|
|
type Input;
|
|
|
|
/// Make a new interface that communicates with this [`InterfaceManager`].
|
|
fn get_interface(&self) -> Self::Interface;
|
|
|
|
/// Consume an input message.
|
|
///
|
|
/// When implementing, call [`.consume_stream_message()`](Self::consume_stream_message) for any encapsulated
|
|
/// [`StreamMessage`]s received.
|
|
fn consume(&mut self, input: Self::Input) -> Result<(), ShellError>;
|
|
|
|
/// Get the [`StreamManager`] for handling operations related to stream messages.
|
|
fn stream_manager(&self) -> &StreamManager;
|
|
|
|
/// Prepare [`PipelineData`] after reading. This is called by `read_pipeline_data()` as
|
|
/// a hook so that values that need special handling can be taken care of.
|
|
fn prepare_pipeline_data(&self, data: PipelineData) -> Result<PipelineData, ShellError>;
|
|
|
|
/// Consume an input stream message.
|
|
///
|
|
/// This method is provided for implementors to use.
|
|
fn consume_stream_message(&mut self, message: StreamMessage) -> Result<(), ShellError> {
|
|
self.stream_manager().handle_message(message)
|
|
}
|
|
|
|
/// Generate `PipelineData` for reading a stream, given a [`PipelineDataHeader`] that was
|
|
/// received from the other side.
|
|
///
|
|
/// This method is provided for implementors to use.
|
|
fn read_pipeline_data(
|
|
&self,
|
|
header: PipelineDataHeader,
|
|
signals: &Signals,
|
|
) -> Result<PipelineData, ShellError> {
|
|
self.prepare_pipeline_data(match header {
|
|
PipelineDataHeader::Empty => PipelineData::Empty,
|
|
PipelineDataHeader::Value(value) => PipelineData::Value(value, None),
|
|
PipelineDataHeader::ListStream(info) => {
|
|
let handle = self.stream_manager().get_handle();
|
|
let reader = handle.read_stream(info.id, self.get_interface())?;
|
|
ListStream::new(reader, info.span, signals.clone()).into()
|
|
}
|
|
PipelineDataHeader::ByteStream(info) => {
|
|
let handle = self.stream_manager().get_handle();
|
|
let reader = handle.read_stream(info.id, self.get_interface())?;
|
|
ByteStream::from_result_iter(reader, info.span, signals.clone(), info.type_).into()
|
|
}
|
|
})
|
|
}
|
|
}
|
|
|
|
/// An interface provides an API for communicating with a plugin or the engine and facilitates
|
|
/// stream I/O. See `PluginInterface` in `nu-plugin-engine` for the API from the engine side to a
|
|
/// plugin, or `EngineInterface` in `nu-plugin` for the API from the plugin side to the engine.
|
|
///
|
|
/// There can be multiple copies of the interface managed by a single [`InterfaceManager`].
|
|
pub trait Interface: Clone + Send {
|
|
/// The output message type, which must be capable of encapsulating a [`StreamMessage`].
|
|
type Output: From<StreamMessage>;
|
|
|
|
/// Any context required to construct [`PipelineData`]. Can be `()` if not needed.
|
|
type DataContext;
|
|
|
|
/// Write an output message.
|
|
fn write(&self, output: Self::Output) -> Result<(), ShellError>;
|
|
|
|
/// Flush the output buffer, so messages are visible to the other side.
|
|
fn flush(&self) -> Result<(), ShellError>;
|
|
|
|
/// Get the sequence for generating new [`StreamId`](nu_plugin_protocol::StreamId)s.
|
|
fn stream_id_sequence(&self) -> &Sequence;
|
|
|
|
/// Get the [`StreamManagerHandle`] for doing stream operations.
|
|
fn stream_manager_handle(&self) -> &StreamManagerHandle;
|
|
|
|
/// Prepare [`PipelineData`] to be written. This is called by `init_write_pipeline_data()` as
|
|
/// a hook so that values that need special handling can be taken care of.
|
|
fn prepare_pipeline_data(
|
|
&self,
|
|
data: PipelineData,
|
|
context: &Self::DataContext,
|
|
) -> Result<PipelineData, ShellError>;
|
|
|
|
/// Initialize a write for [`PipelineData`]. This returns two parts: the header, which can be
|
|
/// embedded in the particular message that references the stream, and a writer, which will
|
|
/// write out all of the data in the pipeline when `.write()` is called.
|
|
///
|
|
/// Note that not all [`PipelineData`] starts a stream. You should call `write()` anyway, as
|
|
/// it will automatically handle this case.
|
|
///
|
|
/// This method is provided for implementors to use.
|
|
fn init_write_pipeline_data(
|
|
&self,
|
|
data: PipelineData,
|
|
context: &Self::DataContext,
|
|
) -> Result<(PipelineDataHeader, PipelineDataWriter<Self>), ShellError> {
|
|
// Allocate a stream id and a writer
|
|
let new_stream = |high_pressure_mark: i32| {
|
|
// Get a free stream id
|
|
let id = self.stream_id_sequence().next()?;
|
|
// Create the writer
|
|
let writer =
|
|
self.stream_manager_handle()
|
|
.write_stream(id, self.clone(), high_pressure_mark)?;
|
|
Ok::<_, ShellError>((id, writer))
|
|
};
|
|
match self.prepare_pipeline_data(data, context)? {
|
|
PipelineData::Value(value, ..) => {
|
|
Ok((PipelineDataHeader::Value(value), PipelineDataWriter::None))
|
|
}
|
|
PipelineData::Empty => Ok((PipelineDataHeader::Empty, PipelineDataWriter::None)),
|
|
PipelineData::ListStream(stream, ..) => {
|
|
let (id, writer) = new_stream(LIST_STREAM_HIGH_PRESSURE)?;
|
|
Ok((
|
|
PipelineDataHeader::ListStream(ListStreamInfo {
|
|
id,
|
|
span: stream.span(),
|
|
}),
|
|
PipelineDataWriter::ListStream(writer, stream),
|
|
))
|
|
}
|
|
PipelineData::ByteStream(stream, ..) => {
|
|
let span = stream.span();
|
|
let type_ = stream.type_();
|
|
if let Some(reader) = stream.reader() {
|
|
let (id, writer) = new_stream(RAW_STREAM_HIGH_PRESSURE)?;
|
|
let header = PipelineDataHeader::ByteStream(ByteStreamInfo { id, span, type_ });
|
|
Ok((header, PipelineDataWriter::ByteStream(writer, reader)))
|
|
} else {
|
|
Ok((PipelineDataHeader::Empty, PipelineDataWriter::None))
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<T> WriteStreamMessage for T
|
|
where
|
|
T: Interface,
|
|
{
|
|
fn write_stream_message(&mut self, msg: StreamMessage) -> Result<(), ShellError> {
|
|
self.write(msg.into())
|
|
}
|
|
|
|
fn flush(&mut self) -> Result<(), ShellError> {
|
|
<Self as Interface>::flush(self)
|
|
}
|
|
}
|
|
|
|
/// Completes the write operation for a [`PipelineData`]. You must call
|
|
/// [`PipelineDataWriter::write()`] to write all of the data contained within the streams.
|
|
#[derive(Default)]
|
|
#[must_use]
|
|
pub enum PipelineDataWriter<W: WriteStreamMessage> {
|
|
#[default]
|
|
None,
|
|
ListStream(StreamWriter<W>, ListStream),
|
|
ByteStream(StreamWriter<W>, Reader),
|
|
}
|
|
|
|
impl<W> PipelineDataWriter<W>
|
|
where
|
|
W: WriteStreamMessage + Send + 'static,
|
|
{
|
|
/// Write all of the data in each of the streams. This method waits for completion.
|
|
pub fn write(self) -> Result<(), ShellError> {
|
|
match self {
|
|
// If no stream was contained in the PipelineData, do nothing.
|
|
PipelineDataWriter::None => Ok(()),
|
|
// Write a list stream.
|
|
PipelineDataWriter::ListStream(mut writer, stream) => {
|
|
writer.write_all(stream)?;
|
|
Ok(())
|
|
}
|
|
// Write a byte stream.
|
|
PipelineDataWriter::ByteStream(mut writer, mut reader) => {
|
|
let span = reader.span();
|
|
let buf = &mut [0; 8192];
|
|
writer.write_all(std::iter::from_fn(move || match reader.read(buf) {
|
|
Ok(0) => None,
|
|
Ok(len) => Some(Ok(buf[..len].to_vec())),
|
|
Err(err) => Some(Err(ShellError::from(err.into_spanned(span)))),
|
|
}))?;
|
|
Ok(())
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Write all of the data in each of the streams. This method returns immediately; any necessary
|
|
/// write will happen in the background. If a thread was spawned, its handle is returned.
|
|
pub fn write_background(
|
|
self,
|
|
) -> Result<Option<thread::JoinHandle<Result<(), ShellError>>>, ShellError> {
|
|
match self {
|
|
PipelineDataWriter::None => Ok(None),
|
|
_ => Ok(Some(
|
|
thread::Builder::new()
|
|
.name("plugin stream background writer".into())
|
|
.spawn(move || {
|
|
let result = self.write();
|
|
if let Err(ref err) = result {
|
|
// Assume that the background thread error probably won't be handled and log it
|
|
// here just in case.
|
|
log::warn!("Error while writing pipeline in background: {err}");
|
|
}
|
|
result
|
|
})?,
|
|
)),
|
|
}
|
|
}
|
|
}
|