Struct DebugStats 

pub struct DebugStats {

    pub parses: u64,
    pub ref_index_locks: u64,
    pub warm_sweeps_completed: u64,
    pub workspace_symbol_index_ready: bool,
    pub workspace_index_walks: u64,
    pub reachability_scan_passes: u64,
    pub analysis_compute_count: u64,
    pub warm_start_replays_completed: u64,
    pub warm_start_untrusted_reanalyzed: u64,
    pub vendor_warm_sweeps_completed: u64,
    pub debug_gate_held: Option<String>,
    pub workspace_file_count: u64,
    pub token_cache_len: u64,
    pub text_cache_len: u64,
    pub parsed_cache_len: u64,
    pub analysis_cache_len: u64,
    pub owned_program_cache_len: u64,
    pub decl_fingerprints_len: u64,
    pub vendor_index_cache_len: u64,
    pub mir_ref_query_cache_hits: u64,
    pub mir_subtype_query_cache_hits: u64,
    pub mir_mention_scans_recorded: u64,
}

Response for the $/php-lsp/debugStats custom request.

Fields

parses: u64

Cumulative count of real ParsedDoc parses (cache misses).

ref_index_locks: u64

Times mir’s reference index was locked. Reads are per-key posting lookups and edits commit one file’s postings, so this grows by a small bounded amount per operation — never per candidate file.

warm_sweeps_completed: u64

Analysis warm sweeps run to completion (not cancelled). Lets benches and tests wait for the post-index sweep before baselining.

workspace_symbol_index_ready: bool

Whether mir’s workspace symbol index singleton is populated (seeded from the disk cache at warm start, or built by the first sweep).

workspace_index_walks: u64

Executions of mir’s tracked O(all-files) symbol-index walk. A cache-warm session must keep this at zero — the regression tests assert it.

reachability_scan_passes: u64

Workspace-wide FQN/text-needle scan passes for reference-scope narrowing (DocumentStore::resolve_reachability_queries). A request with many narrowing-requiring declarations (e.g. code lens on a file with several classes/public-static-methods) must pay one pass per request, not one per declaration — the regression tests assert the delta across such a request.

analysis_compute_count: u64

Real FileAnalyzer::analyze runs (never a cache hit). Concurrent callers racing for the same uncached file’s analysis (e.g. did_open diagnostics vs. a fast-following hover) must serialize onto one run, not one per caller — the regression test asserts the delta.

warm_start_replays_completed: u64

DocumentStore::warm_start_indexes calls that replayed at least one file. A runtime-added workspace folder runs this in the same spawned task as its scan, before the refresh-request notifications go out — tests await a count bump here instead of guessing a fixed delay.

warm_start_untrusted_reanalyzed: u64

Files handed to a background reanalysis because their replayed reference postings were flagged untrusted by mir’s warm_start_files. See DocumentStore::warm_start_untrusted_reanalyzed.

vendor_warm_sweeps_completed: u64

Throttled vendor warm-analysis sweeps run to completion. Always 0 today — the sweep doesn’t exist yet (ROADMAP 0c step 2). See DocumentStore::vendor_warm_sweeps_completed.

debug_gate_held: Option<String>

Section currently parked at the debug gate (see debug_gate.rs), if any. The responsiveness tests poll this to prove a handler’s blocking work is both off the serve future and genuinely in flight.

workspace_file_count: u64

Files mirrored into the salsa workspace. Denominator for the per-file cache sizes below.

token_cache_len: u64

Entries in the semantic-tokens delta cache.

text_cache_len: u64

Entries in the mirrored-source-text cache. Expected to track workspace_file_count.

parsed_cache_len: u64

Entries in the read-through ParsedDoc cache (LRU-capped).

analysis_cache_len: u64

Entries in the per-file FileAnalysis cache (LRU-capped).

owned_program_cache_len: u64

Entries in the owned-Program cache (LRU-capped).

decl_fingerprints_len: u64

Entries in the declaration-fingerprint cache. Expected to track workspace_file_count.

vendor_index_cache_len: u64

Entries in the lazily-loaded vendor FileIndex cache. Not LRU-capped today — the one cache size here worth watching on a long-running session with heavy vendor navigation.

mir_ref_query_cache_hits: u64

Hits served from mir’s indexed_references_to memoization cache (see mir_analyzer::AnalysisSession::ref_query_cache_hits).

mir_subtype_query_cache_hits: u64

Hits served from mir’s indexed_subtype_classes memoization cache (see mir_analyzer::AnalysisSession::subtype_query_cache_hits).

mir_mention_scans_recorded: u64

Per-file text scans recorded in mir’s own ClassMentionIndex — the single shared backend every resolve_reachability_queries needle is answered by. The ratio against reachability_scan_passes is the per-file mention cache’s hit rate — a warm, unedited repeat query should not move this at all.

Trait Implementations

impl Debug for DebugStats

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for DebugStats

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for DebugStats

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.