pub struct Counts {Show 54 fields
pub dir_opens: u64,
pub dir_enumeration_calls: u64,
pub dir_entries: u64,
pub stats: u64,
pub file_opens: u64,
pub file_reads: u64,
pub bytes_read: u64,
pub allocs: u64,
pub reallocs: u64,
pub frees: u64,
pub bytes_allocated: u64,
pub upserts: u64,
pub parent_resolutions: u64,
pub parent_memo_hits: u64,
pub rollup_merges: u64,
pub control_reads: u64,
pub control_refused: u64,
pub control_sources_shared: u64,
pub entries_allocated: u64,
pub baseline_batches: u64,
pub baseline_accepted_ops: u64,
pub opened_batches: u64,
pub opened_accepted_ops: u64,
pub public_batches: u64,
pub public_accepted_ops: u64,
pub ancestry_overlay_inserts: u64,
pub ancestry_path_comparisons: u64,
pub ancestry_parent_proofs: u64,
pub scanner_prepare_us: u64,
pub scanner_control_projection_us: u64,
pub scanner_reduce_us: u64,
pub content_sidecar_read_us: u64,
pub content_sidecar_parse_us: u64,
pub content_sidecar_candidates_us: u64,
pub content_sidecar_apply_us: u64,
pub detached_builds: u64,
pub detached_entries: u64,
pub detached_walk_us: u64,
pub detached_finish_us: u64,
pub effect_paths: u64,
pub effect_path_bytes: u64,
pub impact_candidates: u64,
pub impact_ancestor_visits: u64,
pub impact_retained_dirty_paths: u64,
pub impact_all_dirty: u64,
pub journal_retained_commits: u64,
pub journal_cloned_commits: u64,
pub journal_oversized_commits: u64,
pub journal_dropped_commits: u64,
pub adaptive_calibration_chunks: u64,
pub adaptive_calibration_entries: u64,
pub adaptive_calibration_work_us: u64,
pub adaptive_scale_ups: u64,
pub adaptive_policy_undecided: u64,
}Expand description
Per-layer tallies for a walk, index build, and content pass.
Fields§
§dir_opens: u64Logical directory-open operations.
dir_enumeration_calls: u64Enumeration syscalls in complete successful macOS bulk listings, including the empty terminator.
One successful directory is typically two calls (one data, one EOF), the same
multiplicity the Linux portable path pays in getdents64. The portable read_dir
path leaves this at 0 because the standard library hides that split. Abandoned
bulk attempts before portable fallback are not counted. A later Linux or openat
campaign can compare this to dir_opens without another sample.
dir_entries: u64Directory entries yielded by enumeration.
stats: u64Logical metadata-stat operations.
file_opens: u64File-open operations performed by content analysis.
file_reads: u64File read calls performed by content analysis.
bytes_read: u64Bytes read from files by content analysis.
allocs: u64Allocation requests observed by the installed counting allocator.
reallocs: u64Reallocation requests observed by the installed counting allocator.
frees: u64Deallocations observed by the installed counting allocator.
bytes_allocated: u64Bytes requested by allocations and allocation growth.
upserts: u64Index upserts applied.
parent_resolutions: u64Parent-path resolutions performed for index updates.
parent_memo_hits: u64Parent resolutions served by the adjacent-parent memo.
rollup_merges: u64Roll-up merge operations.
control_reads: u64Control files read from the filesystem.
control_refused: u64Control sources a control table refused for one of its limits.
Counted where the table decides, so a re-read of a file it has already refused is not: that batch cannot change the table and is never projected. A warm revalidate of a refused tree therefore reports its reads against no refusals, which is the projection being skipped rather than a refusal being lifted.
Control sources that joined a retained identical content instead of parsing again.
entries_allocated: u64Index entries allocated.
baseline_batches: u64Successful detached baseline batches arbitrated by the index.
baseline_accepted_ops: u64Operations accepted from successful detached baseline batches.
opened_batches: u64Successful opened-root batches arbitrated by the index.
opened_accepted_ops: u64Operations accepted from successful opened-root batches.
public_batches: u64Successful arbitrary public batches arbitrated by the index.
public_accepted_ops: u64Operations accepted from successful arbitrary public batches.
ancestry_overlay_inserts: u64Path-keyed structural-overlay inserts during ancestry preflight.
ancestry_path_comparisons: u64Same-parent path comparisons during ancestry preflight.
ancestry_parent_proofs: u64Parent chains proved during ancestry preflight, including memo hits.
scanner_prepare_us: u64Wall microseconds spent preparing trusted scanner batches.
scanner_control_projection_us: u64Wall microseconds spent projecting controls for trusted scanner batches.
scanner_reduce_us: u64Wall microseconds spent reducing prepared scanner batches into the index.
content_sidecar_read_us: u64Wall microseconds spent reading the content sidecar image.
content_sidecar_parse_us: u64Wall microseconds spent on sidecar integrity and record decode.
content_sidecar_candidates_us: u64Wall microseconds spent building the live candidate map for a sidecar restore.
content_sidecar_apply_us: u64Wall microseconds spent applying restored records and rebuilding content roll-ups.
detached_builds: u64Detached cold scans that selected the directory-group builder.
detached_entries: u64Entries consumed by that builder.
detached_walk_us: u64Wall microseconds spent walking while the builder consumed directory groups.
detached_finish_us: u64Wall microseconds spent in the final directories-only roll-up pass.
effect_paths: u64Owned paths retained in exact effective changes.
effect_path_bytes: u64Native encoded bytes retained by exact effective-change paths.
impact_candidates: u64Change and state paths considered for bounded impact derivation.
impact_ancestor_visits: u64Path ancestors visited while deriving bounded impact.
impact_retained_dirty_paths: u64Dirty paths retained in completed bounded impacts.
impact_all_dirty: u64Impact derivations that crossed the dirty-path bound.
journal_retained_commits: u64Exact commits retained in the bounded journal.
journal_cloned_commits: u64Exact commits cloned for attempted journal retention.
journal_oversized_commits: u64Exact commits rejected because one commit exceeded journal capacity.
journal_dropped_commits: u64Older exact commits dropped to admit a new retained commit.
adaptive_calibration_chunks: u64Chunk releases folded into an automatic scan’s worker-scaling calibration.
adaptive_calibration_entries: u64Entries folded into that calibration before it reached a decision.
adaptive_calibration_work_us: u64Worker time, in microseconds, folded into that calibration.
Microseconds rather than nanoseconds because the decision threshold is expressed per entry in microseconds, and a nanosecond total over a multi-million-entry walk is large enough to make the row hard to read.
adaptive_scale_ups: u64Reserve-pool expansions an automatic scan actually performed.
adaptive_policy_undecided: u64Walks that ended while their scaling calibration was still undecided.
A walk shorter than the calibration window never observes enough entries to
decide, so its worker policy is unobservable rather than merely unchanged. That
is the distinction crate::scan fails closed on, and it must not read as a
decision to hold.