symbol_cache.py
python
sha256:9ca3eeef4f199d5e8d0b21e50c384a2468e2c2477bd9b157f29fd0f7f06fddcd
feat: add muse reflog expire subcommand and reflog.expire-d…
Sonnet 4.6
patch
71 days ago
| 1 | """Persistent symbol-tree cache — eliminates re-parsing on every CLI invocation. |
| 2 | |
| 3 | Architecture |
| 4 | ------------ |
| 5 | Every call to ``symbols_for_snapshot`` currently re-parses all semantic files |
| 6 | from their raw bytes: ~1,300 ms for a 400-file Python codebase. The parse |
| 7 | result is fully determined by the file content — the same bytes always produce |
| 8 | the same ``SymbolTree``. |
| 9 | |
| 10 | ``SymbolCache`` exploits this by persisting the parse results keyed by the |
| 11 | SHA-256 of the file bytes (``object_id`` from the content-addressed object |
| 12 | store, or a freshly computed SHA-256 for working-tree reads). On a warm cache |
| 13 | hit the parse step is skipped entirely; the full snapshot loads in ~22 ms |
| 14 | instead of ~1,300 ms — a 60× speedup. |
| 15 | |
| 16 | Cache key |
| 17 | --------- |
| 18 | The key is the 64-character lowercase SHA-256 hex digest of the **raw file |
| 19 | bytes**, which is: |
| 20 | |
| 21 | * For committed files — the ``object_id`` from the snapshot manifest |
| 22 | (content-addressed, so it is already the SHA-256 of the bytes). |
| 23 | * For working-tree files — ``hashlib.sha256(raw_bytes).hexdigest()``, |
| 24 | computed from the bytes already read from disk. |
| 25 | |
| 26 | Because the key is content-addressed, every hit is guaranteed correct. |
| 27 | A file edit produces a new SHA-256 → cache miss → fresh parse → new entry. |
| 28 | Old entries for changed files become dead weight but never produce wrong |
| 29 | results. |
| 30 | |
| 31 | Storage |
| 32 | ------- |
| 33 | ``.muse/cache/symbols.json`` — a plain JSON document:: |
| 34 | |
| 35 | { |
| 36 | "version": 2, |
| 37 | "entries": { |
| 38 | "<sha256-hex>": { |
| 39 | "<file::address>": { |
| 40 | "kind": "function", |
| 41 | "name": "run", |
| 42 | ... |
| 43 | } |
| 44 | } |
| 45 | } |
| 46 | } |
| 47 | |
| 48 | Writes are atomic (Pattern A): ``mkstemp`` gives each writer a unique temp |
| 49 | file so two concurrent saves cannot interleave bytes; ``os.replace`` is the |
| 50 | atomic rename. |
| 51 | |
| 52 | Pruning |
| 53 | ------- |
| 54 | Entries accumulate as new file versions are committed. The cache is naturally |
| 55 | bounded by the object store: once ``muse gc`` collects an object, its symbol |
| 56 | cache entry is dead weight. A future ``muse gc`` integration can prune the |
| 57 | cache via :meth:`SymbolCache.prune`. Until then, the cache grows at the rate |
| 58 | of unique file-content changes — typically kilobytes per commit. |
| 59 | """ |
| 60 | |
| 61 | import hashlib |
| 62 | import pathlib |
| 63 | from typing import TypeGuard, get_args |
| 64 | |
| 65 | from muse.core.cache_base import MsgpackCache, _RawCacheMap |
| 66 | from muse.core.types import MsgpackValue |
| 67 | from muse.plugins.code.ast_parser import SymbolKind, SymbolRecord, SymbolTree |
| 68 | |
| 69 | type _SymbolTreeMap = dict[str, "SymbolTree"] |
| 70 | |
| 71 | _CACHE_VERSION = 2 |
| 72 | _CACHE_FILENAME = "symbols.json" |
| 73 | |
| 74 | # All string fields in SymbolRecord, used for structural validation on load. |
| 75 | _STR_FIELDS: frozenset[str] = frozenset({ |
| 76 | "kind", "name", "qualified_name", "content_id", |
| 77 | "body_hash", "signature_id", "metadata_id", "canonical_key", |
| 78 | }) |
| 79 | _INT_FIELDS: frozenset[str] = frozenset({"lineno", "end_lineno"}) |
| 80 | _VALID_KINDS: frozenset[str] = frozenset(get_args(SymbolKind)) |
| 81 | |
| 82 | def _object_id_of(raw: bytes) -> str: |
| 83 | """Return the content-addressed ``sha256:`` ID of *raw* bytes.""" |
| 84 | from muse.core.types import blob_id |
| 85 | return blob_id(raw) |
| 86 | |
| 87 | def _is_symbol_record(obj: MsgpackValue) -> TypeGuard[SymbolRecord]: |
| 88 | """Return ``True`` if *obj* is structurally a valid ``SymbolRecord``. |
| 89 | |
| 90 | Used during cache deserialization to validate entries from JSON without |
| 91 | constructing a new dict — the validated dict IS the SymbolRecord at runtime. |
| 92 | """ |
| 93 | if not isinstance(obj, dict): |
| 94 | return False |
| 95 | for field in _STR_FIELDS: |
| 96 | if not isinstance(obj.get(field), str): |
| 97 | return False |
| 98 | for field in _INT_FIELDS: |
| 99 | if not isinstance(obj.get(field), int): |
| 100 | return False |
| 101 | return obj["kind"] in _VALID_KINDS |
| 102 | |
| 103 | class SymbolCache(MsgpackCache): |
| 104 | """Persistent JSON cache mapping object_id → SymbolTree. |
| 105 | |
| 106 | Inherits load/save/get/put/prune/size/empty from :class:`MsgpackCache` |
| 107 | (Pattern A — mkstemp + replace). |
| 108 | |
| 109 | Typical lifecycle inside ``symbols_for_snapshot``:: |
| 110 | |
| 111 | cache = SymbolCache.load(muse_dir) |
| 112 | for file_path, object_id in manifest.items(): |
| 113 | tree = cache.get(object_id) |
| 114 | if tree is None: |
| 115 | raw = read_object(root, object_id) |
| 116 | tree = parse_symbols(raw, file_path) |
| 117 | cache.put(object_id, tree) |
| 118 | cache.save() |
| 119 | |
| 120 | Attributes |
| 121 | ---------- |
| 122 | _cache_dir : pathlib.Path | None |
| 123 | Absolute path to ``.muse/cache/``. ``None`` for in-memory-only |
| 124 | instances (``empty()``). ``save()`` is a no-op when ``None``. |
| 125 | _dirty : bool |
| 126 | Set to ``True`` by ``put()`` and ``prune()`` when entries change. |
| 127 | Reset to ``False`` by a successful ``save()``. |
| 128 | """ |
| 129 | |
| 130 | _CACHE_FILENAME = "symbols.json" |
| 131 | _CACHE_VERSION = 2 |
| 132 | _TEMP_PREFIX = ".symbols_" |
| 133 | |
| 134 | @classmethod |
| 135 | def _deserialize_entries(cls, raw: _RawCacheMap) -> _RawCacheMap: |
| 136 | """Validate and convert raw JSON entries to a typed SymbolTree map. |
| 137 | |
| 138 | Each entry is a dict of address → SymbolRecord. Invalid entries (wrong |
| 139 | structure) are skipped so a partially corrupt file does not poison the |
| 140 | whole cache. |
| 141 | """ |
| 142 | entries: _SymbolTreeMap = {} |
| 143 | for obj_id, tree_raw in raw.items(): |
| 144 | if not isinstance(obj_id, str) or not isinstance(tree_raw, dict): |
| 145 | continue |
| 146 | tree: SymbolTree = {} |
| 147 | valid = True |
| 148 | for addr, rec_raw in tree_raw.items(): |
| 149 | if not _is_symbol_record(rec_raw): |
| 150 | valid = False |
| 151 | break |
| 152 | tree[addr] = rec_raw |
| 153 | if valid: |
| 154 | entries[obj_id] = tree |
| 155 | return entries |
| 156 | |
| 157 | def _serialize_entries(self) -> _RawCacheMap: |
| 158 | """Return ``_entries`` as-is — SymbolTree dicts are already JSON-serialisable.""" |
| 159 | return self._entries # type: ignore[return-value] |
| 160 | |
| 161 | # Type-narrowing overrides (correctness identical to base; annotations differ) |
| 162 | |
| 163 | def get(self, object_id: str) -> SymbolTree | None: |
| 164 | """Return the cached ``SymbolTree`` for *object_id*, or ``None`` on miss.""" |
| 165 | return self._entries.get(object_id) # type: ignore[return-value] |
| 166 | |
| 167 | def put(self, object_id: str, tree: SymbolTree) -> None: |
| 168 | """Store *tree* under *object_id* and mark the cache dirty.""" |
| 169 | super().put(object_id, tree) |
| 170 | |
| 171 | def prune(self, live_ids: set[str]) -> None: |
| 172 | """Remove entries whose object IDs are not in *live_ids*. |
| 173 | |
| 174 | Call this from ``muse gc`` after identifying all reachable object IDs |
| 175 | to keep the cache from growing unboundedly. |
| 176 | """ |
| 177 | super().prune(live_ids) |
| 178 | |
| 179 | def load_symbol_cache(root: pathlib.Path) -> SymbolCache: |
| 180 | """Convenience loader: return a ``SymbolCache`` for a repository root. |
| 181 | |
| 182 | Returns ``SymbolCache.empty()`` when *root* has no ``.muse`` directory |
| 183 | so callers never need to guard against a missing repo. |
| 184 | """ |
| 185 | return SymbolCache.from_root(root) # type: ignore[return-value] |
File History
1 commit
sha256:9ca3eeef4f199d5e8d0b21e50c384a2468e2c2477bd9b157f29fd0f7f06fddcd
feat: add muse reflog expire subcommand and reflog.expire-d…
Sonnet 4.6
patch
71 days ago