Module freshness 

The freshness sweep (spec/sync/README.md §4.3): the file_stats cache against a filesystem snapshot — as a pure plan (sweep_plan) and as the I/O around it; read-only disk drift; the cache rebuild.

Structs

CacheRow

A file_stats row.

DiskDrift

How the database disagrees with the disk (§4.3 “Disk drift”).

DiskEntry

One walked file with its stat.

SweepPlan

The §4.3 decisions over a snapshot.

SweepResult

The checkpoint result plus the scan counters.

Functions

detect_disk_drift

Read-only: the same cache and snapshot, counted, touching nothing.

file_stats_rows

The cache as the fixtures project it: (path, mtime_ns, size, hash hex) by path.

freshness_sweep

§4.3: plan over the cache and the snapshot, refresh touched-not-edited stats, checkpoint changed ++ deletions at ts (with git_head on the row), refresh the cache for the changed paths and drop the deletions’ rows.

load_cache

The repo’s file_stats rows in row order.

rebuild_file_stats

§4.3 (since 1.1): delete the repo’s rows, then walk the tree and record a row only for a file whose bytes’ hash equals its live doc’s file_hash — the cache may say “known” only about bytes the store already holds. A file with no live doc, or whose bytes differ from what was ingested, gets no row, so the next sweep still sees it as a candidate and drift still reports it. Returns the number of files walked, recorded or not.

record_file_stat

§4.3 step 4 record_file_stat: upsert the fresh stat and hash; a path that vanished meanwhile deletes its row.

snapshot

The walk with each file’s stat (a file that vanished between the walk and the stat is skipped).

sweep_plan

§4.3 steps 1–2 as a pure function: hash_of(path) is called once per candidate, in candidate order.