Files
sourcekit-lsp/Sources/InProcessClient/InProcessSourceKitLSPClient.swift
Alex Hoppen 7f4f92e5bd Rename build system to build server in most cases
The term *build system* predated our wide-spread adoption of BSP for communicating between SourceKit-LSP to the build system and was never really the correct term anyway – ie. a `JSONCompilationDatabaseBuildSystem` never really sounded right. We now have a correct term for the communication layer between SourceKit-LSP: A build server. Rename most occurrences of *build system* to *build server* to reflect this. There are unfortunately a couple lingering instances of *build system* that we can’t change, most notably: `fallbackBuildSystem` in the config file, the `workspace/waitForBuildSystemUpdates` BSP extension request and the `synchronize-for-build-system-updates` experimental feature.
2025-08-02 08:45:01 +02:00

149 lines
5.5 KiB
Swift

//===----------------------------------------------------------------------===//
//
// This source file is part of the Swift.org open source project
//
// Copyright (c) 2014 - 2024 Apple Inc. and the Swift project authors
// Licensed under Apache License v2.0 with Runtime Library Exception
//
// See https://swift.org/LICENSE.txt for license information
// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
//
//===----------------------------------------------------------------------===//
import BuildServerIntegration
public import Foundation
public import LanguageServerProtocol
import LanguageServerProtocolExtensions
import SKLogging
public import SKOptions
package import SourceKitLSP
import SwiftExtensions
import TSCExtensions
package import ToolchainRegistry
import struct TSCBasic.AbsolutePath
/// Launches a `SourceKitLSPServer` in-process and allows sending messages to it.
public final class InProcessSourceKitLSPClient: Sendable {
private let server: SourceKitLSPServer
private let nextRequestID = AtomicUInt32(initialValue: 0)
public convenience init(
toolchainPath: URL?,
options: SourceKitLSPOptions = SourceKitLSPOptions(),
capabilities: ClientCapabilities = ClientCapabilities(),
workspaceFolders: [WorkspaceFolder],
messageHandler: any MessageHandler
) async throws {
try await self.init(
toolchainRegistry: ToolchainRegistry(installPath: toolchainPath),
options: options,
capabilities: capabilities,
workspaceFolders: workspaceFolders,
messageHandler: messageHandler
)
}
/// Create a new `SourceKitLSPServer`. An `InitializeRequest` is automatically sent to the server.
///
/// `messageHandler` handles notifications and requests sent from the SourceKit-LSP server to the client.
package init(
toolchainRegistry: ToolchainRegistry,
options: SourceKitLSPOptions = SourceKitLSPOptions(),
hooks: Hooks = Hooks(),
capabilities: ClientCapabilities = ClientCapabilities(),
workspaceFolders: [WorkspaceFolder],
messageHandler: any MessageHandler
) async throws {
let serverToClientConnection = LocalConnection(receiverName: "client")
self.server = SourceKitLSPServer(
client: serverToClientConnection,
toolchainRegistry: toolchainRegistry,
options: options,
hooks: hooks,
onExit: {
serverToClientConnection.close()
}
)
serverToClientConnection.start(handler: messageHandler)
_ = try await self.send(
InitializeRequest(
processId: nil,
rootPath: nil,
rootURI: nil,
initializationOptions: nil,
capabilities: capabilities,
trace: .off,
workspaceFolders: workspaceFolders
)
)
}
/// Send the request to `server` and return the request result.
///
/// - Important: Because this is an async function, Swift concurrency makes no guarantees about the execution ordering
/// of this request with regard to other requests to the server. If execution of requests in a particular order is
/// necessary and the response of the request is not awaited, use the version of the function that takes a
/// completion handler
public func send<R: RequestType>(_ request: R) async throws -> R.Response {
let requestId = ThreadSafeBox<RequestID?>(initialValue: nil)
return try await withTaskCancellationHandler {
return try await withCheckedThrowingContinuation { continuation in
if Task.isCancelled {
// Check if the task has been cancelled before we send the request to LSP to avoid any kind of work if
// possible.
return continuation.resume(throwing: CancellationError())
}
requestId.value = self.send(request) {
continuation.resume(with: $0)
}
if Task.isCancelled, let requestId = requestId.takeValue() {
// The task might have been cancelled after the above cancellation check but before `requestId` was assigned
// a value. To cover that case, check for cancellation here again. Note that we won't cancel twice from here
// and the `onCancel` handler because we take the request ID out of the `ThreadSafeBox` before sending the
// `CancelRequestNotification`.
self.send(CancelRequestNotification(id: requestId))
}
}
} onCancel: {
if let requestId = requestId.takeValue() {
self.send(CancelRequestNotification(id: requestId))
}
}
}
/// Send the request to `server` and return the request result via a completion handler.
@discardableResult
public func send<R: RequestType>(
_ request: R,
reply: @Sendable @escaping (LSPResult<R.Response>) -> Void
) -> RequestID {
let requestID = RequestID.string("sk-\(Int(nextRequestID.fetchAndIncrement()))")
server.handle(request, id: requestID, reply: reply)
return requestID
}
/// Send the request to `server` and return the request result via a completion handler.
///
/// The request ID must not start with `sk-` to avoid conflicting with the request IDs that are created by
/// `send(:reply:)`.
public func send<R: RequestType>(
_ request: R,
id: RequestID,
reply: @Sendable @escaping (LSPResult<R.Response>) -> Void
) {
if case .string(let string) = id {
if string.starts(with: "sk-") {
logger.fault("Manually specified request ID must not have reserved prefix 'sk-'")
}
}
server.handle(request, id: id, reply: reply)
}
/// Send the notification to `server`.
public func send(_ notification: some NotificationType) {
server.handle(notification)
}
}