Raise request concurrency

[radeksimko]

Request concurrency is required in certain scenarios:

• request cancellation Fix request cancellation #314
• publishing diagnostics directly from the handler (we're currently working around the limitation by publishing in a separate routine

  terraform-ls/internal/langserver/diagnostics/diagnostics.go

  Lines 44 to 79 in 87c4e22

  	go n.notify()
  	return n
  	}
  	// PublishHCLDiags accepts a map of hcl diagnostics per file and queues them for publishing.
  	// A dir path is passed which is joined with the filename keys of the map, to form a file URI.
  	// A source string is passed and set for each diagnostic, this is typically displayed in the client UI.
  	func (n *Notifier) PublishHCLDiags(ctx context.Context, dirPath string, diags map[string]hcl.Diagnostics, source string) {
  	select {
  	case <-n.sessCtx.Done():
  	n.closeDiagsOnce.Do(func() {
  	close(n.diags)
  	})
  	return
  	default:
  	}
  	for filename, ds := range diags {
  	n.diags <- diagContext{
  	ctx: ctx, source: source,
  	diags: ilsp.HCLDiagsToLSP(ds, source),
  	uri: lsp.DocumentURI(uri.FromPath(filepath.Join(dirPath, filename))),
  	}
  	}
  	}
  	func (n *Notifier) notify() {
  	for d := range n.diags {
  	if err := jrpc2.PushNotify(d.ctx, "textDocument/publishDiagnostics", lsp.PublishDiagnosticsParams{
  	URI: d.uri,
  	Diagnostics: n.mergeDiags(d.uri, d.source, d.diags),
  	}); err != nil {
  	n.logger.Printf("Error pushing diagnostics: %s", err)
  	}
  	}
  	}

  )
• reporting progress Report progress for indexing operations #5 Report progress when obtaining schema #150 directly from handlers

We can attempt to reverse #120 - i.e. increase concurrency.

Before doing that though it would be very useful to add some telemetry to be able to tell whether requests are still getting received and responded to in the right order. We can leverage the native telemetry/event method of LSP, and/or just log this on the VS Code extension side only.

Related: creachadair/jrpc2#43

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.