tree.v

1//! Walking the working tree: ignore rules, hashing, and the file cache.23module snapshot45import os6import src.storage78// One tracked file in one tree.9pub struct FileEntry {10pub:11	path       string12	hash       string13	size       i6414	executable bool15}1617// walk hashes every tracked file under `root` and returns them sorted by path.18// Symlinks are skipped, as Koh 2 skipped them, so migration stays lossless.19pub fn walk(store &storage.Store, root string, ig Ignore) ![]FileEntry {20	cache := store.load_cache()!21	mut entries := []FileEntry{}22	walk_dir(store, root, '', ig, cache, mut entries)!23	entries.sort(a.path < b.path)24	return entries25}2627fn walk_dir(store &storage.Store, root string, rel string, ig Ignore, cache map[string]storage.CacheRow, mut out []FileEntry) ! {28	dir := if rel == '' { root } else { os.join_path(root, rel) }29	for name in os.ls(dir)! {30		child := if rel == '' { name } else { '${rel}/${name}' }31		st := lstat(os.join_path(root, child)) or { continue }32		// The kind is known by now, so a pattern ending in `/` can be applied33		// to directories only.34		if ig.matches(child, st.kind == .dir) {35			continue36		}37		match st.kind {38			.dir {39				walk_dir(store, root, child, ig, cache, mut out)!40			}41			.file {42				if st.size > storage.max_file_size {43					return error('${child} is ${st.size} bytes; koh stores code and text, not files this large')44				}45				hash := hash_file(store, root, child, st, cache)!46				out << FileEntry{47					path:       child48					hash:       hash49					size:       st.size50					executable: is_executable(st.mode)51				}52			}53			else {}54		}55	}56}5758fn hash_file(store &storage.Store, root string, rel string, st Stat, cache map[string]storage.CacheRow) !string {59	if row := cache[rel] {60		if row.size == st.size && row.mtime_ns == st.mtime_ns && row.ctime_ns == st.ctime_ns {61			return row.hash62		}63	}64	data := os.read_bytes(os.join_path(root, rel))!65	hash := storage.hash_hex(data)66	store.save_cache(storage.CacheRow{67		path:     rel68		size:     st.size69		mtime_ns: st.mtime_ns70		ctime_ns: st.ctime_ns71		hash:     hash72	})!73	return hash74}7576// store_objects writes the bytes of every entry the database does not have77// yet, so a save only pays for content it has never seen.78pub fn store_objects(store &storage.Store, root string, entries []FileEntry) ! {79	for entry in entries {80		if store.has_object(entry.hash)! {81			continue82		}83		data := os.read_bytes(os.join_path(root, entry.path))!84		stored := store.put_object(data)!85		if stored != entry.hash {86			return error('${entry.path} changed while saving')87		}88	}89}9091// same_tree compares two sorted entry lists by path, hash and executable bit.92pub fn same_tree(a []FileEntry, b []FileEntry) bool {93	if a.len != b.len {94		return false95	}96	for i in 0 .. a.len {97		if a[i].path != b[i].path || a[i].hash != b[i].hash || a[i].executable != b[i].executable {98			return false99		}100	}101	return true102}