mirror of https://github.com/helix-editor/helix
don't overload LS with completion resolve requests
While moving completion resolve to the event system in #9668 we introduced what is essentially a "DOS attack" on slow LSPs. Completion resolve requests were made in the render loop and debounced with a timeout. Once the timeout expired the resolve request was made. The problem is the next frame would immediately request a new completion resolve request (and mark the old one as obsolete but because LSP has no notion of cancelation the server would still process it). So we were in essence sending one completion request to the server every 150ms and only stopped if the server managed to respond before we rendered a new frame. This caused overload on slower machines/with slower LS. In this PR I revamped the resolve handler so that a request is only ever resolved once. Both by checking if a request is already in-flight and by marking failed resolve requests as resolved.pull/10555/head
Pascal Kuthe
2 months ago
committed by
Blaž Hrastnik
parent
b834806dbc
commit
38ee845b05
@ -0,0 +1,153 @@
|
||||
use std::sync::Arc;
|
||||
|
||||
use helix_lsp::lsp;
|
||||
use tokio::sync::mpsc::Sender;
|
||||
use tokio::time::{Duration, Instant};
|
||||
|
||||
use helix_event::{send_blocking, AsyncHook, CancelRx};
|
||||
use helix_view::Editor;
|
||||
|
||||
use crate::handlers::completion::CompletionItem;
|
||||
use crate::job;
|
||||
|
||||
/// A hook for resolving incomplete completion items.
|
||||
///
|
||||
/// From the [LSP spec](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#textDocument_completion):
|
||||
///
|
||||
/// > If computing full completion items is expensive, servers can additionally provide a
|
||||
/// > handler for the completion item resolve request. ...
|
||||
/// > A typical use case is for example: the `textDocument/completion` request doesn't fill
|
||||
/// > in the `documentation` property for returned completion items since it is expensive
|
||||
/// > to compute. When the item is selected in the user interface then a
|
||||
/// > 'completionItem/resolve' request is sent with the selected completion item as a parameter.
|
||||
/// > The returned completion item should have the documentation property filled in.
|
||||
pub struct ResolveHandler {
|
||||
last_request: Option<Arc<CompletionItem>>,
|
||||
resolver: Sender<ResolveRequest>,
|
||||
}
|
||||
|
||||
impl ResolveHandler {
|
||||
pub fn new() -> ResolveHandler {
|
||||
ResolveHandler {
|
||||
last_request: None,
|
||||
resolver: ResolveTimeout {
|
||||
next_request: None,
|
||||
in_flight: None,
|
||||
}
|
||||
.spawn(),
|
||||
}
|
||||
}
|
||||
|
||||
pub fn ensure_item_resolved(&mut self, editor: &mut Editor, item: &mut CompletionItem) {
|
||||
if item.resolved {
|
||||
return;
|
||||
}
|
||||
let needs_resolve = item.item.documentation.is_none()
|
||||
|| item.item.detail.is_none()
|
||||
|| item.item.additional_text_edits.is_none();
|
||||
if !needs_resolve {
|
||||
item.resolved = true;
|
||||
return;
|
||||
}
|
||||
if self.last_request.as_deref().is_some_and(|it| it == item) {
|
||||
return;
|
||||
}
|
||||
let Some(ls) = editor.language_servers.get_by_id(item.provider).cloned() else {
|
||||
item.resolved = true;
|
||||
return;
|
||||
};
|
||||
if matches!(
|
||||
ls.capabilities().completion_provider,
|
||||
Some(lsp::CompletionOptions {
|
||||
resolve_provider: Some(true),
|
||||
..
|
||||
})
|
||||
) {
|
||||
let item = Arc::new(item.clone());
|
||||
self.last_request = Some(item.clone());
|
||||
send_blocking(&self.resolver, ResolveRequest { item, ls })
|
||||
} else {
|
||||
item.resolved = true;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
struct ResolveRequest {
|
||||
item: Arc<CompletionItem>,
|
||||
ls: Arc<helix_lsp::Client>,
|
||||
}
|
||||
|
||||
#[derive(Default)]
|
||||
struct ResolveTimeout {
|
||||
next_request: Option<ResolveRequest>,
|
||||
in_flight: Option<(helix_event::CancelTx, Arc<CompletionItem>)>,
|
||||
}
|
||||
|
||||
impl AsyncHook for ResolveTimeout {
|
||||
type Event = ResolveRequest;
|
||||
|
||||
fn handle_event(
|
||||
&mut self,
|
||||
request: Self::Event,
|
||||
timeout: Option<tokio::time::Instant>,
|
||||
) -> Option<tokio::time::Instant> {
|
||||
if self
|
||||
.next_request
|
||||
.as_ref()
|
||||
.is_some_and(|old_request| old_request.item == request.item)
|
||||
{
|
||||
timeout
|
||||
} else if self
|
||||
.in_flight
|
||||
.as_ref()
|
||||
.is_some_and(|(_, old_request)| old_request.item == request.item.item)
|
||||
{
|
||||
self.next_request = None;
|
||||
None
|
||||
} else {
|
||||
self.next_request = Some(request);
|
||||
Some(Instant::now() + Duration::from_millis(150))
|
||||
}
|
||||
}
|
||||
|
||||
fn finish_debounce(&mut self) {
|
||||
let Some(request) = self.next_request.take() else { return };
|
||||
let (tx, rx) = helix_event::cancelation();
|
||||
self.in_flight = Some((tx, request.item.clone()));
|
||||
tokio::spawn(request.execute(rx));
|
||||
}
|
||||
}
|
||||
|
||||
impl ResolveRequest {
|
||||
async fn execute(self, cancel: CancelRx) {
|
||||
let future = self.ls.resolve_completion_item(&self.item.item);
|
||||
let Some(resolved_item) = helix_event::cancelable_future(future, cancel).await else {
|
||||
return;
|
||||
};
|
||||
job::dispatch(move |_, compositor| {
|
||||
if let Some(completion) = &mut compositor
|
||||
.find::<crate::ui::EditorView>()
|
||||
.unwrap()
|
||||
.completion
|
||||
{
|
||||
let resolved_item = match resolved_item {
|
||||
Ok(item) => CompletionItem {
|
||||
item,
|
||||
resolved: true,
|
||||
..*self.item
|
||||
},
|
||||
Err(err) => {
|
||||
log::error!("completion resolve request failed: {err}");
|
||||
// set item to resolved so we don't request it again
|
||||
// we could also remove it but that oculd be odd ui
|
||||
let mut item = (*self.item).clone();
|
||||
item.resolved = true;
|
||||
item
|
||||
}
|
||||
};
|
||||
completion.replace_item(&self.item, resolved_item);
|
||||
};
|
||||
})
|
||||
.await
|
||||
}
|
||||
}
|
||||
Loading…
Reference in New Issue