pub fn hash_and_parse_parallel(
files: &[WalkedFile],
stored_mtimes: &HashMap<String, u64>,
) -> HashParseOutputExpand description
Combined mtime-check + parse pass.
For each file:
- If
mtime_secsmatches the stored value → skip entirely (zero disk I/O). - Otherwise → read file bytes, run tree-sitter, record updated mtime.
This eliminates the full I/O sweep on re-init when files are unchanged: a re-init with no edits costs only the walk + mtime comparison (≈130ms), not a full disk read of all source files (≈2100ms on 58k-file repos).