Struct LspSessionPool 

pub struct LspSessionPool { /* private fields */ }

P1-1: each live language-server session is wrapped in its own Mutex<LspSession>. The pool itself is a lock-free DashMap so that requests routed to different (command, args) sessions (e.g. pyright vs rust-analyzer in a polyglot monorepo) no longer serialize on a single pool-level mutex. Requests hitting the same session still take a session-local mutex — the LSP JSON-RPC wire is inherently serial per stdin/stdout pair.

Implementations

impl LspSessionPool

pub fn new(project: ProjectRoot) -> Self

pub fn reset(&self, project: ProjectRoot) -> Self

Replace the project root and close all existing sessions.

pub fn session_count(&self) -> usize

pub fn readiness_snapshot(&self) -> Vec<ReadinessSnapshot>

Snapshot the per-session readiness state for all currently pooled LSP servers. Cheap and lock-free: it allocates a Vec and clones a handful of atomic counters per session. Intended for the MCP get_lsp_readiness handler and for bench-harness polling loops that need to wait for indexing to complete instead of sleeping a fixed duration.

pub fn find_referencing_symbols( &self, request: &LspRequest, ) -> Result<Vec<LspReference>>

pub fn get_diagnostics( &self, request: &LspDiagnosticRequest, ) -> Result<Vec<LspDiagnostic>>

pub fn search_workspace_symbols( &self, request: &LspWorkspaceSymbolRequest, ) -> Result<Vec<LspWorkspaceSymbol>>

pub fn get_type_hierarchy( &self, request: &LspTypeHierarchyRequest, ) -> Result<HashMap<String, Value>>

pub fn get_rename_plan( &self, request: &LspRenamePlanRequest, ) -> Result<LspRenamePlan>