symlog.py
python
sha256:986277e88b9de700cfb77ed3385e7b344df08fbb801c9615906da7df9c45c91d
feat: muse#53 item 12 — muse symlog prune-orphans
Sonnet 5
patch
1 day ago
| 1 | """Symlog — live per-symbol journal written at commit time. |
| 2 | |
| 3 | Every commit that changes a symbol's content ID appends an entry to that |
| 4 | symbol's journal. Journals are stored in a directory tree that mirrors the |
| 5 | source tree:: |
| 6 | |
| 7 | .muse/symlogs/ |
| 8 | src/ |
| 9 | billing.py/ |
| 10 | compute_total ← one file per symbol |
| 11 | validate_invoice |
| 12 | tests/ |
| 13 | test_billing.py/ |
| 14 | test_compute_total |
| 15 | |
| 16 | This layout gives O(1) per-symbol reads (no commit-history scan), O(1) per-file |
| 17 | symbol enumeration (one ``os.listdir``), and independent per-symbol GC. |
| 18 | |
| 19 | Entry format |
| 20 | ------------ |
| 21 | Each journal file is an append-only plain-text sequence of lines:: |
| 22 | |
| 23 | <old_content_id> <new_content_id> <commit_id> <author> <ts_unix> <tz_offset> TAB <operation> |
| 24 | |
| 25 | Fields: |
| 26 | |
| 27 | old_content_id sha256:<64-hex> — symbol content ID before the change, or |
| 28 | NULL_CONTENT_ID for created symbols. |
| 29 | new_content_id sha256:<64-hex> — symbol content ID after the change, or |
| 30 | NULL_CONTENT_ID for deleted/renamed-away symbols. |
| 31 | commit_id sha256:<64-hex> — commit that caused this entry. |
| 32 | author sanitized author or agent_id string. |
| 33 | ts_unix Unix seconds as a decimal integer (UTC). |
| 34 | tz_offset UTC offset in +HHMM / -HHMM form (always +0000 today). |
| 35 | operation Free-form description of the lifecycle event. |
| 36 | |
| 37 | Lifecycle sentinels |
| 38 | ------------------- |
| 39 | symbol-created: <name> |
| 40 | old_content_id == NULL_CONTENT_ID. |
| 41 | |
| 42 | symbol-modified: <commit message first line> |
| 43 | Both content IDs are populated. |
| 44 | |
| 45 | symbol-deleted: <name> |
| 46 | new_content_id == NULL_CONTENT_ID. |
| 47 | |
| 48 | symbol-renamed-to: <new_addr> |
| 49 | Terminal entry written to the OLD symbol path. new_content_id == NULL_CONTENT_ID. |
| 50 | |
| 51 | symbol-born-from: <old_addr> |
| 52 | Born-from entry written to the NEW symbol path. old_content_id == NULL_CONTENT_ID. |
| 53 | The ``born_from`` field of the parsed entry is set to <old_addr>. |
| 54 | |
| 55 | Path encoding |
| 56 | ------------- |
| 57 | The symbol name (part after ``::``) is percent-encoded for any character |
| 58 | outside ``[a-zA-Z0-9._-]``. The file path portion is used as-is (it must |
| 59 | already be a valid relative path). ``..`` components in either portion raise |
| 60 | ``ValueError`` before any filesystem access. |
| 61 | |
| 62 | Security model |
| 63 | -------------- |
| 64 | ``author`` and ``operation`` are sanitized before being written to disk: |
| 65 | newlines are stripped to prevent line injection; tabs are stripped from |
| 66 | ``author`` to prevent corruption of the tab-delimited parse boundary. |
| 67 | |
| 68 | ``list_symlog_symbols`` and ``list_symlog_symbols_for_file`` skip symlinks — |
| 69 | a crafted symlink inside ``.muse/symlogs/`` cannot be followed to enumerate |
| 70 | arbitrary filesystem paths. |
| 71 | |
| 72 | ``read_symlog`` enforces a ``_MAX_SYMLOG_BYTES`` file-size cap before reading |
| 73 | to prevent OOM from excessively large or maliciously crafted journal files. |
| 74 | """ |
| 75 | |
| 76 | from __future__ import annotations |
| 77 | |
| 78 | import datetime |
| 79 | import hashlib |
| 80 | import json |
| 81 | import logging |
| 82 | import os |
| 83 | import pathlib |
| 84 | import re |
| 85 | import tempfile |
| 86 | import urllib.parse |
| 87 | from dataclasses import dataclass, field as _field |
| 88 | |
| 89 | from muse.core.paths import symlogs_dir as _symlogs_dir |
| 90 | from muse.core.types import DEFAULT_HASH_ALGO |
| 91 | |
| 92 | logger = logging.getLogger(__name__) |
| 93 | |
| 94 | # --------------------------------------------------------------------------- |
| 95 | # Constants |
| 96 | # --------------------------------------------------------------------------- |
| 97 | |
| 98 | #: Null content ID — used as old_content_id for created symbols and |
| 99 | #: new_content_id for deleted / renamed-away symbols. |
| 100 | NULL_CONTENT_ID: str = f"{DEFAULT_HASH_ALGO}:" + "0" * 64 |
| 101 | |
| 102 | #: Maximum symlog file size that will be read. Files larger than this cap |
| 103 | #: are warned about rather than loaded entirely into memory. |
| 104 | _MAX_SYMLOG_BYTES: int = 10 * 1024 * 1024 # 10 MiB |
| 105 | |
| 106 | #: Characters in the symbol name that do NOT need percent-encoding. |
| 107 | _SAFE_CHARS: str = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789._-" |
| 108 | |
| 109 | #: Leaf filenames at or under this many UTF-8 bytes are always safe on every |
| 110 | #: filesystem this tool supports (POSIX's 255-byte NAME_MAX, with headroom |
| 111 | #: for the ".lock" suffix atomic writes append). Percent-encoded names longer |
| 112 | #: than this fall back to a hashed leaf name instead of raising OSError. |
| 113 | _MAX_LEAF_BYTES: int = 200 |
| 114 | |
| 115 | #: Marker prefix for hashed fallback leaf names. Percent-encoding (see |
| 116 | #: ``_encode_symbol_name``) only ever emits characters from ``_SAFE_CHARS`` |
| 117 | #: or a literal ``%``, so ``!`` can never appear in a normally-encoded leaf — |
| 118 | #: making it an unambiguous signal that the leaf name must be resolved via |
| 119 | #: the sidecar index rather than percent-decoded directly. |
| 120 | _HASHED_LEAF_PREFIX: str = "!" |
| 121 | |
| 122 | #: Sidecar index file, one per symlog directory, mapping hashed leaf names |
| 123 | #: back to their original (pre-hash) symbol name. Only written/consulted for |
| 124 | #: leaves that were too long to encode directly; normal leaves never touch it. |
| 125 | _LEAF_INDEX_FILENAME: str = ".leaf_index.json" |
| 126 | |
| 127 | #: Regex that identifies a ``symbol-born-from:`` operation and captures the prior address. |
| 128 | _BORN_FROM_RE = re.compile(r"^symbol-born-from:\s*(.+)$") |
| 129 | |
| 130 | |
| 131 | # --------------------------------------------------------------------------- |
| 132 | # Helpers |
| 133 | # --------------------------------------------------------------------------- |
| 134 | |
| 135 | def is_null_content_id(content_id: str) -> bool: |
| 136 | """Return ``True`` iff *content_id* is the canonical null content ID.""" |
| 137 | return content_id == NULL_CONTENT_ID |
| 138 | |
| 139 | |
| 140 | def _encode_symbol_name(name: str) -> str: |
| 141 | """Percent-encode characters outside ``[a-zA-Z0-9._-]`` in *name*.""" |
| 142 | return urllib.parse.quote(name, safe=_SAFE_CHARS) |
| 143 | |
| 144 | |
| 145 | def _decode_symbol_name(encoded: str) -> str: |
| 146 | """Reverse percent-encoding applied by ``_encode_symbol_name``.""" |
| 147 | return urllib.parse.unquote(encoded) |
| 148 | |
| 149 | |
| 150 | def _hashed_leaf_name(symbol_name: str) -> str: |
| 151 | """Return a short, collision-resistant leaf name for a too-long symbol name.""" |
| 152 | digest = hashlib.sha256(symbol_name.encode("utf-8")).hexdigest() |
| 153 | return f"{_HASHED_LEAF_PREFIX}{digest[:48]}" |
| 154 | |
| 155 | |
| 156 | def _leaf_index_path(leaf: pathlib.Path) -> pathlib.Path: |
| 157 | """Return the sidecar index path for the symlog directory containing *leaf*.""" |
| 158 | return leaf.parent / _LEAF_INDEX_FILENAME |
| 159 | |
| 160 | |
| 161 | def _record_hashed_leaf(leaf: pathlib.Path, symbol_name: str) -> None: |
| 162 | """Record *symbol_name* under *leaf*'s hashed name in the sidecar index. |
| 163 | |
| 164 | Idempotent and atomic (write-to-temp + ``os.replace``) — safe to call on |
| 165 | every append, not just the first. |
| 166 | """ |
| 167 | index_path = _leaf_index_path(leaf) |
| 168 | try: |
| 169 | data = ( |
| 170 | json.loads(index_path.read_text(encoding="utf-8")) |
| 171 | if index_path.exists() |
| 172 | else {} |
| 173 | ) |
| 174 | except (OSError, ValueError): |
| 175 | data = {} |
| 176 | if data.get(leaf.name) == symbol_name: |
| 177 | return |
| 178 | data[leaf.name] = symbol_name |
| 179 | tmp_path = index_path.with_name(index_path.name + ".tmp") |
| 180 | tmp_path.write_text(json.dumps(data, sort_keys=True), encoding="utf-8") |
| 181 | os.replace(tmp_path, index_path) |
| 182 | |
| 183 | |
| 184 | def _resolve_hashed_leaf(leaf: pathlib.Path) -> str | None: |
| 185 | """Look up the original symbol name for a hashed *leaf* via the sidecar index.""" |
| 186 | index_path = _leaf_index_path(leaf) |
| 187 | if not index_path.exists(): |
| 188 | return None |
| 189 | try: |
| 190 | data = json.loads(index_path.read_text(encoding="utf-8")) |
| 191 | except (OSError, ValueError): |
| 192 | return None |
| 193 | value = data.get(leaf.name) |
| 194 | return value if isinstance(value, str) else None |
| 195 | |
| 196 | |
| 197 | def _reject_traversal(part: str, label: str) -> None: |
| 198 | """Raise ``ValueError`` when *part* contains a ``..`` path component.""" |
| 199 | segments = pathlib.PurePosixPath(part).parts |
| 200 | if ".." in segments: |
| 201 | raise ValueError( |
| 202 | f"Path traversal detected in {label!r}: '..' components are not allowed." |
| 203 | ) |
| 204 | |
| 205 | |
| 206 | def _sanitize_author(s: str) -> str: |
| 207 | """Strip characters that would corrupt the symlog line structure.""" |
| 208 | return s.replace("\n", "").replace("\r", "").replace("\t", "") |
| 209 | |
| 210 | |
| 211 | def _sanitize_operation(s: str) -> str: |
| 212 | """Strip newlines to prevent line injection in the operation field.""" |
| 213 | return s.replace("\n", "").replace("\r", "") |
| 214 | |
| 215 | |
| 216 | # --------------------------------------------------------------------------- |
| 217 | # Types |
| 218 | # --------------------------------------------------------------------------- |
| 219 | |
| 220 | @dataclass(frozen=True) |
| 221 | class SymlogEntry: |
| 222 | """One line of a symbol journal. |
| 223 | |
| 224 | The ``born_from`` field is non-None only when the operation string starts |
| 225 | with ``symbol-born-from:``; it is always derived from ``operation`` at |
| 226 | construction time and is never passed by callers. This ensures the two |
| 227 | fields are always consistent. |
| 228 | """ |
| 229 | |
| 230 | old_content_id: str # sha256:… or NULL_CONTENT_ID |
| 231 | new_content_id: str # sha256:… or NULL_CONTENT_ID |
| 232 | commit_id: str # sha256:… of the commit that caused this entry |
| 233 | author: str # sanitized author / agent_id |
| 234 | timestamp: datetime.datetime |
| 235 | operation: str # "symbol-created: …", "symbol-modified: …", etc. |
| 236 | born_from: str | None = _field(default=None, init=False) |
| 237 | |
| 238 | def __post_init__(self) -> None: |
| 239 | m = _BORN_FROM_RE.match(self.operation) |
| 240 | object.__setattr__(self, "born_from", m.group(1).strip() if m else None) |
| 241 | |
| 242 | |
| 243 | def _make_entry( |
| 244 | old_content_id: str, |
| 245 | new_content_id: str, |
| 246 | commit_id: str, |
| 247 | author: str, |
| 248 | timestamp: datetime.datetime, |
| 249 | operation: str, |
| 250 | ) -> SymlogEntry: |
| 251 | """Construct a ``SymlogEntry`` from parsed line components.""" |
| 252 | return SymlogEntry( |
| 253 | old_content_id=old_content_id, |
| 254 | new_content_id=new_content_id, |
| 255 | commit_id=commit_id, |
| 256 | author=author, |
| 257 | timestamp=timestamp, |
| 258 | operation=operation, |
| 259 | ) |
| 260 | |
| 261 | |
| 262 | # --------------------------------------------------------------------------- |
| 263 | # Path helpers |
| 264 | # --------------------------------------------------------------------------- |
| 265 | |
| 266 | def symlog_path(repo_root: pathlib.Path, symbol_addr: str) -> pathlib.Path: |
| 267 | """Return the journal file path for *symbol_addr*. |
| 268 | |
| 269 | *symbol_addr* must contain exactly one ``::`` separator, e.g. |
| 270 | ``src/billing.py::compute_total``. The file path portion maps to a |
| 271 | directory under ``.muse/symlogs/``; the symbol name portion becomes the |
| 272 | leaf filename (percent-encoded for non-ASCII-safe characters). |
| 273 | |
| 274 | Raises: |
| 275 | ValueError: ``::`` separator is missing, symbol name is empty, |
| 276 | or a ``..`` path component is detected. |
| 277 | """ |
| 278 | if "::" not in symbol_addr: |
| 279 | raise ValueError( |
| 280 | f"Missing '::' separator in symbol address {symbol_addr!r}. " |
| 281 | "Expected form: 'path/to/file.py::symbol_name'." |
| 282 | ) |
| 283 | file_path, _, symbol_name = symbol_addr.partition("::") |
| 284 | if not symbol_name: |
| 285 | raise ValueError( |
| 286 | f"Empty symbol name in symbol address {symbol_addr!r}." |
| 287 | ) |
| 288 | # Reject path traversal in both parts. |
| 289 | _reject_traversal(file_path, "file_path") |
| 290 | _reject_traversal(symbol_name, "symbol_name") |
| 291 | encoded_name = _encode_symbol_name(symbol_name) |
| 292 | if len(encoded_name.encode("utf-8")) > _MAX_LEAF_BYTES: |
| 293 | encoded_name = _hashed_leaf_name(symbol_name) |
| 294 | return _symlogs_dir(repo_root) / file_path / encoded_name |
| 295 | |
| 296 | |
| 297 | def _addr_from_path(symlogs_root: pathlib.Path, leaf: pathlib.Path) -> str: |
| 298 | """Reconstruct a symbol address from a leaf path under *symlogs_root*. |
| 299 | |
| 300 | Raises: |
| 301 | ValueError: *leaf* is a hashed fallback name with no sidecar index |
| 302 | entry (index missing or corrupt). |
| 303 | """ |
| 304 | rel = leaf.relative_to(symlogs_root) |
| 305 | parts = rel.parts |
| 306 | # All parts up to the last are the file path; the last is the encoded symbol name. |
| 307 | file_path = str(pathlib.PurePosixPath(*parts[:-1])) |
| 308 | leaf_name = parts[-1] |
| 309 | if leaf_name.startswith(_HASHED_LEAF_PREFIX): |
| 310 | symbol_name = _resolve_hashed_leaf(leaf) |
| 311 | if symbol_name is None: |
| 312 | raise ValueError( |
| 313 | f"No sidecar index entry for hashed symlog leaf {leaf!r}." |
| 314 | ) |
| 315 | else: |
| 316 | symbol_name = _decode_symbol_name(leaf_name) |
| 317 | return f"{file_path}::{symbol_name}" |
| 318 | |
| 319 | |
| 320 | # --------------------------------------------------------------------------- |
| 321 | # Write |
| 322 | # --------------------------------------------------------------------------- |
| 323 | |
| 324 | def append_symlog( |
| 325 | repo_root: pathlib.Path, |
| 326 | symbol_addr: str, |
| 327 | old_content_id: str, |
| 328 | new_content_id: str, |
| 329 | commit_id: str, |
| 330 | author: str, |
| 331 | operation: str, |
| 332 | ) -> None: |
| 333 | """Append one entry to the symbol journal for *symbol_addr*. |
| 334 | |
| 335 | Parent directories are created automatically on first write. Author and |
| 336 | operation are sanitized to prevent line injection. |
| 337 | |
| 338 | Args: |
| 339 | repo_root: Root of the Muse repository. |
| 340 | symbol_addr: ``"path/to/file.py::symbol_name"``. |
| 341 | old_content_id: Content ID before the change (NULL_CONTENT_ID for created). |
| 342 | new_content_id: Content ID after the change (NULL_CONTENT_ID for deleted). |
| 343 | commit_id: SHA-256 commit ID that caused this entry. |
| 344 | author: Agent ID or author string; sanitized before write. |
| 345 | operation: Lifecycle description; sanitized before write. |
| 346 | """ |
| 347 | safe_author = _sanitize_author(author) |
| 348 | safe_operation = _sanitize_operation(operation) |
| 349 | now = datetime.datetime.now(tz=datetime.timezone.utc) |
| 350 | ts = int(now.timestamp()) |
| 351 | tz_offset = "+0000" |
| 352 | |
| 353 | line = ( |
| 354 | f"{old_content_id} {new_content_id} {commit_id} " |
| 355 | f"{safe_author} {ts} {tz_offset}\t{safe_operation}\n" |
| 356 | ) |
| 357 | |
| 358 | p = symlog_path(repo_root, symbol_addr) |
| 359 | p.parent.mkdir(parents=True, exist_ok=True) |
| 360 | if p.name.startswith(_HASHED_LEAF_PREFIX): |
| 361 | _, _, symbol_name = symbol_addr.partition("::") |
| 362 | _record_hashed_leaf(p, symbol_name) |
| 363 | with p.open("a", encoding="utf-8") as fh: |
| 364 | fh.write(line) |
| 365 | |
| 366 | |
| 367 | # --------------------------------------------------------------------------- |
| 368 | # Parse |
| 369 | # --------------------------------------------------------------------------- |
| 370 | |
| 371 | def _parse_line(line: str) -> SymlogEntry | None: |
| 372 | """Parse one symlog line; return None on malformed input. |
| 373 | |
| 374 | Format:: |
| 375 | |
| 376 | <old_cid> <new_cid> <commit_id> <author…> <ts_unix> <tz_offset> TAB <operation> |
| 377 | """ |
| 378 | line = line.rstrip("\n") |
| 379 | parts = line.split("\t", 1) |
| 380 | if len(parts) != 2: |
| 381 | return None |
| 382 | meta, operation = parts |
| 383 | tokens = meta.split() |
| 384 | # Minimum: old_cid new_cid commit_id author ts tz (6 tokens) |
| 385 | if len(tokens) < 6: |
| 386 | return None |
| 387 | old_content_id = tokens[0] |
| 388 | new_content_id = tokens[1] |
| 389 | commit_id = tokens[2] |
| 390 | ts_str = tokens[-2] |
| 391 | # author may contain spaces — everything between tokens[3] and the last two |
| 392 | author = " ".join(tokens[3:-2]) |
| 393 | try: |
| 394 | ts = datetime.datetime.fromtimestamp(int(ts_str), tz=datetime.timezone.utc) |
| 395 | except (ValueError, OSError): |
| 396 | ts = datetime.datetime.now(tz=datetime.timezone.utc) |
| 397 | return _make_entry( |
| 398 | old_content_id=old_content_id, |
| 399 | new_content_id=new_content_id, |
| 400 | commit_id=commit_id, |
| 401 | author=author, |
| 402 | timestamp=ts, |
| 403 | operation=operation, |
| 404 | ) |
| 405 | |
| 406 | |
| 407 | # --------------------------------------------------------------------------- |
| 408 | # Read |
| 409 | # --------------------------------------------------------------------------- |
| 410 | |
| 411 | def read_symlog( |
| 412 | repo_root: pathlib.Path, |
| 413 | symbol_addr: str, |
| 414 | limit: int = 100, |
| 415 | *, |
| 416 | follow: bool = False, |
| 417 | ) -> list[SymlogEntry]: |
| 418 | """Return symlog entries for *symbol_addr*, newest-first, up to *limit*. |
| 419 | |
| 420 | When *follow* is ``True`` and the oldest entry in this symbol's log has a |
| 421 | ``born_from`` pointer (i.e. the operation starts with ``symbol-born-from:``), |
| 422 | the function recursively reads the prior symbol's log and appends those |
| 423 | entries so the full rename chain is visible in one list. |
| 424 | |
| 425 | Files larger than ``_MAX_SYMLOG_BYTES`` trigger a warning and are still |
| 426 | read (the OS read may truncate at a line boundary, but at least a partial |
| 427 | result is returned rather than an empty list for very large files). |
| 428 | """ |
| 429 | p = symlog_path(repo_root, symbol_addr) |
| 430 | if not p.exists(): |
| 431 | return [] |
| 432 | |
| 433 | try: |
| 434 | size = p.stat().st_size |
| 435 | if size > _MAX_SYMLOG_BYTES: |
| 436 | logger.warning( |
| 437 | "⚠️ Symlog %s is %.1f MiB — exceeds cap of %d MiB; " |
| 438 | "results may be incomplete. Consider pruning old entries.", |
| 439 | p, |
| 440 | size / (1024 * 1024), |
| 441 | _MAX_SYMLOG_BYTES // (1024 * 1024), |
| 442 | ) |
| 443 | lines = p.read_text(encoding="utf-8", errors="replace").splitlines() |
| 444 | except OSError as exc: |
| 445 | logger.warning("⚠️ Could not read symlog %s: %s", p, exc) |
| 446 | return [] |
| 447 | |
| 448 | entries: list[SymlogEntry] = [] |
| 449 | for line in reversed(lines): |
| 450 | if not line.strip(): |
| 451 | continue |
| 452 | entry = _parse_line(line) |
| 453 | if entry is not None: |
| 454 | entries.append(entry) |
| 455 | if len(entries) >= limit: |
| 456 | break |
| 457 | |
| 458 | if follow and entries: |
| 459 | # Check the last (oldest) entry for a born-from pointer. |
| 460 | oldest = entries[-1] |
| 461 | if oldest.born_from: |
| 462 | prior_entries = read_symlog( |
| 463 | repo_root, |
| 464 | oldest.born_from, |
| 465 | limit=max(limit, 10_000), |
| 466 | follow=True, |
| 467 | ) |
| 468 | entries.extend(prior_entries) |
| 469 | |
| 470 | return entries |
| 471 | |
| 472 | |
| 473 | # --------------------------------------------------------------------------- |
| 474 | # Enumerate |
| 475 | # --------------------------------------------------------------------------- |
| 476 | |
| 477 | def list_symlog_symbols(repo_root: pathlib.Path) -> list[str]: |
| 478 | """Return all symbol addresses that have a journal, sorted. |
| 479 | |
| 480 | Symlinks are excluded — only regular files are yielded. |
| 481 | """ |
| 482 | root = _symlogs_dir(repo_root) |
| 483 | if not root.exists(): |
| 484 | return [] |
| 485 | results: list[str] = [] |
| 486 | for leaf in root.rglob("*"): |
| 487 | if leaf.name == _LEAF_INDEX_FILENAME: |
| 488 | continue |
| 489 | if leaf.is_file() and not leaf.is_symlink(): |
| 490 | try: |
| 491 | results.append(_addr_from_path(root, leaf)) |
| 492 | except ValueError: |
| 493 | continue |
| 494 | return sorted(results) |
| 495 | |
| 496 | |
| 497 | def list_symlog_symbols_for_file( |
| 498 | repo_root: pathlib.Path, |
| 499 | file_path: str, |
| 500 | ) -> list[str]: |
| 501 | """Return symbol addresses for all symbols in *file_path* that have a journal. |
| 502 | |
| 503 | Uses a single ``os.listdir`` — O(1) regardless of total journal count. |
| 504 | Symlinks are excluded. |
| 505 | |
| 506 | Args: |
| 507 | repo_root: Root of the Muse repository. |
| 508 | file_path: Repo-relative path to the source file, e.g. ``"src/billing.py"``. |
| 509 | """ |
| 510 | symlogs_root = _symlogs_dir(repo_root) |
| 511 | file_dir = symlogs_root / file_path |
| 512 | if not file_dir.exists(): |
| 513 | return [] |
| 514 | results: list[str] = [] |
| 515 | for p in file_dir.iterdir(): |
| 516 | if p.name == _LEAF_INDEX_FILENAME: |
| 517 | continue |
| 518 | if p.is_file() and not p.is_symlink(): |
| 519 | try: |
| 520 | results.append(_addr_from_path(symlogs_root, p)) |
| 521 | except ValueError: |
| 522 | continue |
| 523 | return sorted(results) |
| 524 | |
| 525 | |
| 526 | # --------------------------------------------------------------------------- |
| 527 | # Expire |
| 528 | # --------------------------------------------------------------------------- |
| 529 | |
| 530 | def expire_symlog( |
| 531 | repo_root: pathlib.Path, |
| 532 | symbol_addr: str, |
| 533 | expire_days: int, |
| 534 | *, |
| 535 | dry_run: bool = False, |
| 536 | ) -> tuple[int, int]: |
| 537 | """Remove entries older than *expire_days* from one symbol journal. |
| 538 | |
| 539 | The write is atomic: content is first written to a sibling ``.lock`` |
| 540 | temp file then renamed into place with ``os.replace``. An empty journal |
| 541 | after expiry is deleted rather than left as a zero-byte stub. |
| 542 | |
| 543 | Args: |
| 544 | repo_root: Root of the Muse repository. |
| 545 | symbol_addr: ``"path/to/file.py::symbol_name"``. |
| 546 | expire_days: Entries whose timestamp is older than this many days |
| 547 | are removed. |
| 548 | dry_run: When ``True``, compute counts but write nothing. |
| 549 | |
| 550 | Returns: |
| 551 | ``(expired_count, kept_count)``. |
| 552 | """ |
| 553 | p = symlog_path(repo_root, symbol_addr) |
| 554 | if not p.exists(): |
| 555 | return 0, 0 |
| 556 | |
| 557 | try: |
| 558 | raw = p.read_text(encoding="utf-8", errors="replace") |
| 559 | except OSError as exc: |
| 560 | logger.warning("⚠️ symlog expire: could not read %s: %s", p, exc) |
| 561 | return 0, 0 |
| 562 | |
| 563 | cutoff_ts = int(datetime.datetime.now(tz=datetime.timezone.utc).timestamp()) - expire_days * 86400 |
| 564 | |
| 565 | kept_lines: list[str] = [] |
| 566 | expired = 0 |
| 567 | kept = 0 |
| 568 | for line in raw.splitlines(keepends=True): |
| 569 | stripped = line.strip() |
| 570 | if not stripped: |
| 571 | continue |
| 572 | entry = _parse_line(stripped) |
| 573 | if entry is None: |
| 574 | kept_lines.append(line) |
| 575 | kept += 1 |
| 576 | continue |
| 577 | entry_ts = int(entry.timestamp.timestamp()) |
| 578 | if entry_ts < cutoff_ts: |
| 579 | expired += 1 |
| 580 | else: |
| 581 | kept_lines.append(line) |
| 582 | kept += 1 |
| 583 | |
| 584 | if dry_run: |
| 585 | return expired, kept |
| 586 | |
| 587 | if not kept_lines: |
| 588 | try: |
| 589 | p.unlink() |
| 590 | except OSError as exc: |
| 591 | logger.warning("⚠️ symlog expire: could not remove %s: %s", p, exc) |
| 592 | return expired, kept |
| 593 | |
| 594 | tmp_path = p.with_suffix(".lock") |
| 595 | try: |
| 596 | tmp_path.write_text("".join(kept_lines), encoding="utf-8") |
| 597 | os.replace(tmp_path, p) |
| 598 | except OSError as exc: |
| 599 | logger.warning("⚠️ symlog expire: atomic write failed for %s: %s", p, exc) |
| 600 | try: |
| 601 | tmp_path.unlink(missing_ok=True) |
| 602 | except OSError: |
| 603 | pass |
| 604 | raise |
| 605 | |
| 606 | return expired, kept |
| 607 | |
| 608 | |
| 609 | # --------------------------------------------------------------------------- |
| 610 | # Delete |
| 611 | # --------------------------------------------------------------------------- |
| 612 | |
| 613 | def delete_symlog_entry( |
| 614 | repo_root: pathlib.Path, |
| 615 | symbol_addr: str, |
| 616 | index: int | None, |
| 617 | ) -> tuple[int, int]: |
| 618 | """Remove one entry (by index) or all entries from a symbol journal. |
| 619 | |
| 620 | Args: |
| 621 | repo_root: Root of the Muse repository. |
| 622 | symbol_addr: ``"path/to/file.py::symbol_name"``. |
| 623 | index: 0-based index from newest (``@{0}`` = newest). |
| 624 | Pass ``None`` to delete **all** entries. |
| 625 | |
| 626 | Returns: |
| 627 | ``(deleted_count, remaining_count)``. |
| 628 | |
| 629 | Raises: |
| 630 | FileNotFoundError: Journal file does not exist (only when index is not None). |
| 631 | IndexError: *index* is out of range; args are ``(index, total)``. |
| 632 | |
| 633 | The write is atomic when entries remain. An empty journal is deleted |
| 634 | rather than left as a zero-byte stub. |
| 635 | """ |
| 636 | p = symlog_path(repo_root, symbol_addr) |
| 637 | |
| 638 | if not p.exists(): |
| 639 | if index is None: |
| 640 | return 0, 0 # --all on missing log is graceful |
| 641 | raise FileNotFoundError(p) |
| 642 | |
| 643 | raw = p.read_text(encoding="utf-8", errors="replace") |
| 644 | all_lines = [ln for ln in raw.splitlines(keepends=True) if ln.strip()] |
| 645 | total = len(all_lines) |
| 646 | |
| 647 | if index is None: |
| 648 | try: |
| 649 | p.unlink() |
| 650 | except OSError as exc: |
| 651 | logger.warning("⚠️ symlog delete --all: could not remove %s: %s", p, exc) |
| 652 | return total, 0 |
| 653 | |
| 654 | if index < 0 or index >= total: |
| 655 | raise IndexError(index, total) |
| 656 | |
| 657 | # index is 0-based from newest; lines are stored oldest-first. |
| 658 | line_pos = total - 1 - index |
| 659 | kept_lines = all_lines[:line_pos] + all_lines[line_pos + 1:] |
| 660 | |
| 661 | if not kept_lines: |
| 662 | try: |
| 663 | p.unlink() |
| 664 | except OSError as exc: |
| 665 | logger.warning("⚠️ symlog delete: could not remove %s: %s", p, exc) |
| 666 | return 1, 0 |
| 667 | |
| 668 | tmp_path = p.with_suffix(".lock") |
| 669 | try: |
| 670 | tmp_path.write_text("".join(kept_lines), encoding="utf-8") |
| 671 | os.replace(tmp_path, p) |
| 672 | except OSError as exc: |
| 673 | logger.warning("⚠️ symlog delete: atomic write failed for %s: %s", p, exc) |
| 674 | try: |
| 675 | tmp_path.unlink(missing_ok=True) |
| 676 | except OSError: |
| 677 | pass |
| 678 | raise |
| 679 | |
| 680 | return 1, len(kept_lines) |
| 681 | |
| 682 | |
| 683 | # --------------------------------------------------------------------------- |
| 684 | # Orphan pruning (muse#53 item 12) |
| 685 | # --------------------------------------------------------------------------- |
| 686 | |
| 687 | def _symbol_exists_in_head(repo_root: pathlib.Path, symbol_addr: str) -> bool: |
| 688 | """Return ``True`` if *symbol_addr*'s file and symbol both exist at HEAD. |
| 689 | |
| 690 | Never raises — any failure along the way (no HEAD commit, file removed, |
| 691 | parse failure) is treated as "doesn't exist", matching |
| 692 | :func:`extract_symbols`'s own fail-silent contract. |
| 693 | """ |
| 694 | if "::" not in symbol_addr: |
| 695 | return False |
| 696 | file_path, _, symbol_name = symbol_addr.partition("::") |
| 697 | if not symbol_name: |
| 698 | return False |
| 699 | try: |
| 700 | from muse.core.refs import get_head_commit_id, read_current_branch |
| 701 | from muse.core.commits import read_commit |
| 702 | from muse.core.snapshots import read_snapshot |
| 703 | |
| 704 | branch = read_current_branch(repo_root) |
| 705 | commit_id = get_head_commit_id(repo_root, branch) |
| 706 | if commit_id is None: |
| 707 | return False |
| 708 | commit_rec = read_commit(repo_root, commit_id) |
| 709 | if commit_rec is None: |
| 710 | return False |
| 711 | snap_rec = read_snapshot(repo_root, commit_rec.snapshot_id) |
| 712 | if snap_rec is None: |
| 713 | return False |
| 714 | file_obj_id = snap_rec.manifest.get(file_path) |
| 715 | if file_obj_id is None: |
| 716 | return False |
| 717 | symbols = extract_symbols(repo_root, file_obj_id, file_path) |
| 718 | return symbol_name in symbols |
| 719 | except Exception: |
| 720 | return False |
| 721 | |
| 722 | |
| 723 | def find_orphan_symlogs(repo_root: pathlib.Path) -> list[str]: |
| 724 | """Return symbol addresses with a journal whose symbol no longer exists at HEAD. |
| 725 | |
| 726 | A symbol is orphaned when either its file was deleted/renamed away, or |
| 727 | the symbol itself was removed from that file — in both cases the journal |
| 728 | under ``.muse/symlogs/`` survives (journals are append-only and never |
| 729 | auto-deleted on symbol removal) with no live symbol to show it for. |
| 730 | """ |
| 731 | return [ |
| 732 | addr for addr in list_symlog_symbols(repo_root) |
| 733 | if not _symbol_exists_in_head(repo_root, addr) |
| 734 | ] |
| 735 | |
| 736 | |
| 737 | def prune_orphan_symlogs( |
| 738 | repo_root: pathlib.Path, |
| 739 | *, |
| 740 | dry_run: bool = False, |
| 741 | ) -> list[str]: |
| 742 | """Delete (or report, with *dry_run*) every orphaned symlog journal. |
| 743 | |
| 744 | Reuses :func:`delete_symlog_entry` with ``index=None`` for the actual |
| 745 | removal — the same "delete the whole journal" path ``muse symlog delete |
| 746 | --all`` already uses and already has tested atomic-delete behavior for. |
| 747 | |
| 748 | Returns: |
| 749 | The list of symbol addresses pruned (or that would be pruned, under |
| 750 | ``dry_run``). |
| 751 | """ |
| 752 | orphans = find_orphan_symlogs(repo_root) |
| 753 | if dry_run: |
| 754 | return orphans |
| 755 | for addr in orphans: |
| 756 | try: |
| 757 | delete_symlog_entry(repo_root, addr, None) |
| 758 | except FileNotFoundError: |
| 759 | continue |
| 760 | return orphans |
| 761 | |
| 762 | |
| 763 | # --------------------------------------------------------------------------- |
| 764 | # Phase 2 — Symbol diff and commit integration |
| 765 | # --------------------------------------------------------------------------- |
| 766 | |
| 767 | @dataclass |
| 768 | class SymbolDiff: |
| 769 | """Result of comparing two symbol maps (old vs. new content IDs). |
| 770 | |
| 771 | ``renames`` holds ``(old_name, new_name)`` pairs where the same content ID |
| 772 | appears in both the deleted and created sets — same object, different name. |
| 773 | Renamed symbols are removed from ``created`` and ``deleted`` so callers |
| 774 | never double-count them. |
| 775 | """ |
| 776 | |
| 777 | created: set[str] |
| 778 | deleted: set[str] |
| 779 | modified: set[str] |
| 780 | renames: list[tuple[str, str]] |
| 781 | |
| 782 | |
| 783 | def compute_symbol_diff( |
| 784 | old_symbols: dict[str, str], |
| 785 | new_symbols: dict[str, str], |
| 786 | old_rename_keys: dict[str, str] | None = None, |
| 787 | new_rename_keys: dict[str, str] | None = None, |
| 788 | ) -> SymbolDiff: |
| 789 | """Diff two ``{symbol_name: content_id}`` maps. |
| 790 | |
| 791 | Rename detection: a name that disappeared in *old* whose rename key matches |
| 792 | that of a new name that didn't exist before is classified as a rename rather |
| 793 | than a delete + create. |
| 794 | |
| 795 | *old_rename_keys* / *new_rename_keys* are optional ``{name: key}`` maps |
| 796 | used exclusively for rename matching — typically ``body_hash`` values from |
| 797 | ``parse_symbols``, which are stable across pure renames. When not supplied, |
| 798 | the ``content_id`` from *old_symbols* / *new_symbols* is used as the key |
| 799 | instead (correct for unit tests with artificial IDs; rarely matches in |
| 800 | practice for real code because ``content_id`` encodes the function name). |
| 801 | |
| 802 | A content-ID change with a name change is always treated as delete + create. |
| 803 | """ |
| 804 | _old_rk = old_rename_keys if old_rename_keys is not None else old_symbols |
| 805 | _new_rk = new_rename_keys if new_rename_keys is not None else new_symbols |
| 806 | |
| 807 | old_names = set(old_symbols) |
| 808 | new_names = set(new_symbols) |
| 809 | |
| 810 | raw_deleted = old_names - new_names |
| 811 | raw_created = new_names - old_names |
| 812 | modified = {n for n in old_names & new_names if old_symbols[n] != new_symbols[n]} |
| 813 | |
| 814 | # Build an inverted map of rename_key → name for the *created* side. |
| 815 | created_by_rk: dict[str, str] = {} |
| 816 | for name in raw_created: |
| 817 | rk = _new_rk.get(name, "") |
| 818 | if not rk: |
| 819 | continue |
| 820 | # Only map when a single created name has this key — ambiguous keys |
| 821 | # cannot be deterministically paired with a deleted name. |
| 822 | if rk not in created_by_rk: |
| 823 | created_by_rk[rk] = name |
| 824 | else: |
| 825 | created_by_rk[rk] = "" # sentinel: ambiguous |
| 826 | |
| 827 | renames: list[tuple[str, str]] = [] |
| 828 | confirmed_deleted = set(raw_deleted) |
| 829 | confirmed_created = set(raw_created) |
| 830 | |
| 831 | for old_name in sorted(raw_deleted): |
| 832 | rk = _old_rk.get(old_name, "") |
| 833 | new_name = created_by_rk.get(rk) if rk else None |
| 834 | if new_name: # non-empty sentinel → unambiguous match |
| 835 | renames.append((old_name, new_name)) |
| 836 | confirmed_deleted.discard(old_name) |
| 837 | confirmed_created.discard(new_name) |
| 838 | |
| 839 | return SymbolDiff( |
| 840 | created=confirmed_created, |
| 841 | deleted=confirmed_deleted, |
| 842 | modified=modified, |
| 843 | renames=renames, |
| 844 | ) |
| 845 | |
| 846 | |
| 847 | def extract_symbols( |
| 848 | repo_root: pathlib.Path, |
| 849 | file_object_id: str, |
| 850 | file_path: str = "", |
| 851 | ) -> dict[str, str]: |
| 852 | """Read a file object from the store and return ``{symbol_name: content_id}``. |
| 853 | |
| 854 | Uses the code AST parser to extract top-level symbols. Returns an empty |
| 855 | dict for objects that are not present in the store, cannot be parsed, or |
| 856 | are non-code files. Never raises — parse failures are silently swallowed |
| 857 | because symlog writes are non-fatal. |
| 858 | |
| 859 | Args: |
| 860 | repo_root: Root of the Muse repository. |
| 861 | file_object_id: Content-addressed ID of the file object (``sha256:…``). |
| 862 | file_path: Repo-relative file path passed to ``parse_symbols`` so |
| 863 | symbol keys are properly namespaced (``"src/f.py::name"``). |
| 864 | """ |
| 865 | try: |
| 866 | from muse.core.object_store import read_object |
| 867 | source = read_object(repo_root, file_object_id) |
| 868 | if source is None: |
| 869 | return {} |
| 870 | from muse.plugins.code.ast_parser import parse_symbols |
| 871 | # parse_symbols returns {full_addr: {content_id: …, …}} |
| 872 | # Keys are "file.py::name" — we want just the name part. |
| 873 | raw = parse_symbols(source, file_path) |
| 874 | result: dict[str, str] = {} |
| 875 | for addr, sym in raw.items(): |
| 876 | cid = sym.get("content_id", "") |
| 877 | if not cid: |
| 878 | continue |
| 879 | name = addr.split("::")[-1] if "::" in addr else addr |
| 880 | result[name] = cid |
| 881 | return result |
| 882 | except Exception: |
| 883 | return {} |
| 884 | |
| 885 | |
| 886 | def _extract_body_hashes( |
| 887 | repo_root: pathlib.Path, |
| 888 | file_object_id: str, |
| 889 | file_path: str = "", |
| 890 | ) -> dict[str, str]: |
| 891 | """Return ``{symbol_name: body_hash}`` for rename detection. |
| 892 | |
| 893 | ``body_hash`` hashes only the function body, not the name — two functions |
| 894 | with the same body but different names share the same ``body_hash``, making |
| 895 | it the correct key for rename matching. Falls back to ``content_id`` if |
| 896 | ``body_hash`` is absent from the symbol record. |
| 897 | """ |
| 898 | try: |
| 899 | from muse.core.object_store import read_object |
| 900 | source = read_object(repo_root, file_object_id) |
| 901 | if source is None: |
| 902 | return {} |
| 903 | from muse.plugins.code.ast_parser import parse_symbols |
| 904 | raw = parse_symbols(source, file_path) |
| 905 | result: dict[str, str] = {} |
| 906 | for addr, sym in raw.items(): |
| 907 | key = sym.get("body_hash", "") or sym.get("content_id", "") |
| 908 | if not key: |
| 909 | continue |
| 910 | name = addr.split("::")[-1] if "::" in addr else addr |
| 911 | result[name] = key |
| 912 | return result |
| 913 | except Exception: |
| 914 | return {} |
| 915 | |
| 916 | |
| 917 | def _write_symlogs( |
| 918 | repo_root: pathlib.Path, |
| 919 | parent_commit_id: str | None, |
| 920 | new_snapshot_id: str, |
| 921 | new_commit_id: str, |
| 922 | author: str, |
| 923 | commit_message: str, |
| 924 | ) -> None: |
| 925 | """Write symlog entries for every symbol that changed in this commit. |
| 926 | |
| 927 | Called by ``muse commit`` after the commit record is finalised. Wrapped |
| 928 | in a broad ``try/except`` at the call site — a failure here must never |
| 929 | abort a commit. |
| 930 | |
| 931 | Algorithm: |
| 932 | |
| 933 | 1. Build ``old_manifest`` from the parent commit (empty for initial commit). |
| 934 | 2. For each file whose object ID changed, extract old and new symbol maps. |
| 935 | 3. Diff the two maps to classify changes as created/modified/deleted/renamed. |
| 936 | 4. Append the appropriate symlog entry for each changed symbol. |
| 937 | """ |
| 938 | from muse.core.object_store import read_object |
| 939 | from muse.core.commits import read_commit |
| 940 | from muse.core.snapshots import read_snapshot |
| 941 | |
| 942 | # Load parent manifest. |
| 943 | old_manifest: dict[str, str] = {} |
| 944 | if parent_commit_id is not None: |
| 945 | parent_rec = read_commit(repo_root, parent_commit_id) |
| 946 | if parent_rec is not None: |
| 947 | snap_rec = read_snapshot(repo_root, parent_rec.snapshot_id) |
| 948 | if snap_rec is not None: |
| 949 | old_manifest = dict(snap_rec.manifest) |
| 950 | |
| 951 | # Load new manifest. |
| 952 | new_snap_rec = read_snapshot(repo_root, new_snapshot_id) |
| 953 | if new_snap_rec is None: |
| 954 | return |
| 955 | new_manifest: dict[str, str] = dict(new_snap_rec.manifest) |
| 956 | |
| 957 | first_line = (commit_message.splitlines()[0] if commit_message else "").strip() |
| 958 | |
| 959 | # Find files whose object IDs differ. |
| 960 | all_paths = set(old_manifest) | set(new_manifest) |
| 961 | for file_path in sorted(all_paths): |
| 962 | old_obj = old_manifest.get(file_path, "") |
| 963 | new_obj = new_manifest.get(file_path, "") |
| 964 | if old_obj == new_obj: |
| 965 | continue # file unchanged at object level |
| 966 | |
| 967 | old_syms = extract_symbols(repo_root, old_obj, file_path) if old_obj else {} |
| 968 | new_syms = extract_symbols(repo_root, new_obj, file_path) if new_obj else {} |
| 969 | |
| 970 | if not old_syms and not new_syms: |
| 971 | continue # non-code file |
| 972 | |
| 973 | old_bodies = _extract_body_hashes(repo_root, old_obj, file_path) if old_obj else {} |
| 974 | new_bodies = _extract_body_hashes(repo_root, new_obj, file_path) if new_obj else {} |
| 975 | diff = compute_symbol_diff(old_syms, new_syms, old_bodies, new_bodies) |
| 976 | |
| 977 | # Created |
| 978 | for name in sorted(diff.created): |
| 979 | append_symlog( |
| 980 | repo_root, |
| 981 | f"{file_path}::{name}", |
| 982 | old_content_id=NULL_CONTENT_ID, |
| 983 | new_content_id=new_syms[name], |
| 984 | commit_id=new_commit_id, |
| 985 | author=author, |
| 986 | operation=f"symbol-created: {name}", |
| 987 | ) |
| 988 | |
| 989 | # Modified |
| 990 | for name in sorted(diff.modified): |
| 991 | append_symlog( |
| 992 | repo_root, |
| 993 | f"{file_path}::{name}", |
| 994 | old_content_id=old_syms[name], |
| 995 | new_content_id=new_syms[name], |
| 996 | commit_id=new_commit_id, |
| 997 | author=author, |
| 998 | operation=f"symbol-modified: {first_line}", |
| 999 | ) |
| 1000 | |
| 1001 | # Deleted |
| 1002 | for name in sorted(diff.deleted): |
| 1003 | append_symlog( |
| 1004 | repo_root, |
| 1005 | f"{file_path}::{name}", |
| 1006 | old_content_id=old_syms[name], |
| 1007 | new_content_id=NULL_CONTENT_ID, |
| 1008 | commit_id=new_commit_id, |
| 1009 | author=author, |
| 1010 | operation=f"symbol-deleted: {name}", |
| 1011 | ) |
| 1012 | |
| 1013 | # Renamed — two entries per rename |
| 1014 | for old_name, new_name in diff.renames: |
| 1015 | new_addr = f"{file_path}::{new_name}" |
| 1016 | old_addr = f"{file_path}::{old_name}" |
| 1017 | # Terminal entry on the old path |
| 1018 | append_symlog( |
| 1019 | repo_root, |
| 1020 | old_addr, |
| 1021 | old_content_id=old_syms[old_name], |
| 1022 | new_content_id=NULL_CONTENT_ID, |
| 1023 | commit_id=new_commit_id, |
| 1024 | author=author, |
| 1025 | operation=f"symbol-renamed-to: {new_addr}", |
| 1026 | ) |
| 1027 | # Born-from entry on the new path |
| 1028 | append_symlog( |
| 1029 | repo_root, |
| 1030 | new_addr, |
| 1031 | old_content_id=NULL_CONTENT_ID, |
| 1032 | new_content_id=new_syms[new_name], |
| 1033 | commit_id=new_commit_id, |
| 1034 | author=author, |
| 1035 | operation=f"symbol-born-from: {old_addr}", |
| 1036 | ) |
| 1037 | |
| 1038 | |
| 1039 | # --------------------------------------------------------------------------- |
| 1040 | # Phase 5 — @{N} ref resolution |
| 1041 | # --------------------------------------------------------------------------- |
| 1042 | |
| 1043 | #: Regex matching ``<addr>@{N}`` or ``<addr>@{<timespec>}`` — anchored at both |
| 1044 | #: ends. Group 2 is disambiguated (digit index vs relative-time spec) inside |
| 1045 | #: ``resolve_symlog_addr`` / ``_resolve_symlog_index`` (muse#53 item 6). |
| 1046 | _SYMLOG_REF_RE = re.compile(r"^(.+)@\{([^}]+)\}$") |
| 1047 | |
| 1048 | #: Matches "N.unit.ago" relative-time specs, e.g. "1.hour.ago", "2.days.ago". |
| 1049 | _SYMLOG_TIME_AGO_RE = re.compile( |
| 1050 | r"^(?P<num>\d+)\.(?P<unit>second|minute|hour|day|week)s?\.ago$" |
| 1051 | ) |
| 1052 | |
| 1053 | _SYMLOG_TIME_UNIT_SECONDS = { |
| 1054 | "second": 1, "minute": 60, "hour": 3600, "day": 86_400, "week": 604_800, |
| 1055 | } |
| 1056 | |
| 1057 | |
| 1058 | def _parse_symlog_time_cutoff(inner: str) -> datetime.datetime | None: |
| 1059 | """Parse a relative-time ``@{...}`` spec into an absolute UTC cutoff. |
| 1060 | |
| 1061 | Supported forms (muse#53 item 6, mirrors the reflog equivalent): |
| 1062 | |
| 1063 | - ``N.unit.ago`` — unit is second/minute/hour/day/week, singular or |
| 1064 | plural (e.g. ``"1.hour.ago"``, ``"2.days.ago"``). |
| 1065 | - ``yesterday`` — start of today (UTC midnight). |
| 1066 | |
| 1067 | Returns ``None`` for anything that isn't a recognized time spec. |
| 1068 | """ |
| 1069 | now = datetime.datetime.now(datetime.timezone.utc) |
| 1070 | if inner == "yesterday": |
| 1071 | return now.replace(hour=0, minute=0, second=0, microsecond=0) |
| 1072 | m = _SYMLOG_TIME_AGO_RE.match(inner) |
| 1073 | if m is None: |
| 1074 | return None |
| 1075 | num = int(m.group("num")) |
| 1076 | unit_seconds = _SYMLOG_TIME_UNIT_SECONDS[m.group("unit")] |
| 1077 | return now - datetime.timedelta(seconds=num * unit_seconds) |
| 1078 | |
| 1079 | |
| 1080 | def _resolve_symlog_index(inner: str, entries: "list[SymlogEntry]") -> int: |
| 1081 | """Resolve the inner content of ``@{...}`` to a 0-based index into *entries*. |
| 1082 | |
| 1083 | *entries* must already be newest-first (the ``read_symlog`` contract). |
| 1084 | Digit-only *inner* resolves by position; anything else is tried as a |
| 1085 | relative-time spec, resolving to the first (newest) entry at or before |
| 1086 | the computed cutoff. |
| 1087 | |
| 1088 | Raises: |
| 1089 | IndexError: ``(index, total)`` for an out-of-range digit index, or |
| 1090 | ``(-1, total)`` when a valid time spec has no qualifying entry. |
| 1091 | ValueError: *inner* is neither a digit index nor a recognized |
| 1092 | relative-time spec. |
| 1093 | """ |
| 1094 | total = len(entries) |
| 1095 | if inner.isdigit(): |
| 1096 | index = int(inner) |
| 1097 | if index >= total: |
| 1098 | raise IndexError(index, total) |
| 1099 | return index |
| 1100 | |
| 1101 | cutoff = _parse_symlog_time_cutoff(inner) |
| 1102 | if cutoff is None: |
| 1103 | raise ValueError( |
| 1104 | f"{inner!r} inside @{{...}} is neither an index nor a recognized " |
| 1105 | "relative-time spec (e.g. '1.hour.ago', 'yesterday')" |
| 1106 | ) |
| 1107 | for idx, entry in enumerate(entries): |
| 1108 | if entry.timestamp <= cutoff: |
| 1109 | return idx |
| 1110 | raise IndexError(-1, total) |
| 1111 | |
| 1112 | |
| 1113 | @dataclass(frozen=True) |
| 1114 | class SymlogResolution: |
| 1115 | """Result of resolving a ``<addr>@{N}`` symlog reference. |
| 1116 | |
| 1117 | ``followed_from`` is non-None only when ``--follow`` crossed a rename |
| 1118 | boundary — it holds the prior symbol address that contributed this entry. |
| 1119 | ``content_id`` equals ``new_content_id`` of the resolved ``SymlogEntry``. |
| 1120 | """ |
| 1121 | |
| 1122 | symbol: str # base symbol address (without @{N}) |
| 1123 | index: int # 0-based index (0 = newest) |
| 1124 | content_id: str # new_content_id at this index |
| 1125 | commit_id: str # commit that caused this entry |
| 1126 | operation: str # lifecycle description |
| 1127 | timestamp: datetime.datetime # UTC-aware entry timestamp |
| 1128 | followed_from: str | None = None # prior addr when follow crossed a boundary |
| 1129 | author: str = "" # author/agent_id recorded on the entry (muse#53 item 9) |
| 1130 | |
| 1131 | |
| 1132 | def resolve_symlog_addr( |
| 1133 | spec: str, |
| 1134 | repo_root: pathlib.Path, |
| 1135 | *, |
| 1136 | follow: bool = False, |
| 1137 | ) -> "SymlogResolution | None": |
| 1138 | """Resolve a ``<addr>@{N}`` symlog reference to a :class:`SymlogResolution`. |
| 1139 | |
| 1140 | Args: |
| 1141 | spec: Full spec string, e.g. ``"billing.py::compute_total@{2}"``, |
| 1142 | or a relative-time spec (muse#53 item 6), e.g. |
| 1143 | ``"billing.py::compute_total@{1.hour.ago}"``. |
| 1144 | repo_root: Root of the Muse repository. |
| 1145 | follow: When ``True``, traverse rename chains so indices beyond the |
| 1146 | primary symbol's log are resolved from prior symbol addresses. |
| 1147 | |
| 1148 | Returns: |
| 1149 | A :class:`SymlogResolution` on success, or ``None`` if *spec* does not |
| 1150 | match the ``@{...}`` pattern (so callers can use ``if result is None:`` |
| 1151 | to fall through to normal processing). |
| 1152 | |
| 1153 | Raises: |
| 1154 | FileNotFoundError: No journal exists for the resolved symbol address. |
| 1155 | IndexError: ``(index, total)`` for an out-of-range digit index, |
| 1156 | or ``(-1, total)`` when a time spec has no |
| 1157 | qualifying entry. |
| 1158 | ValueError: *addr* is not a valid symbol address (missing |
| 1159 | ``::``), or the ``@{...}`` content is neither a |
| 1160 | digit index nor a recognized relative-time spec. |
| 1161 | """ |
| 1162 | m = _SYMLOG_REF_RE.match(spec) |
| 1163 | if m is None: |
| 1164 | return None |
| 1165 | |
| 1166 | addr = m.group(1) |
| 1167 | inner = m.group(2) |
| 1168 | |
| 1169 | # Validate that the addr portion is a proper symbol address. |
| 1170 | # This will raise ValueError if :: is missing (propagated to caller). |
| 1171 | p = symlog_path(repo_root, addr) |
| 1172 | if not p.exists(): |
| 1173 | raise FileNotFoundError(p) |
| 1174 | |
| 1175 | if follow: |
| 1176 | primary = read_symlog(repo_root, addr, limit=100_000, follow=False) |
| 1177 | all_entries = read_symlog(repo_root, addr, limit=100_000, follow=True) |
| 1178 | index = _resolve_symlog_index(inner, all_entries) |
| 1179 | entry = all_entries[index] |
| 1180 | followed_from: str | None = None |
| 1181 | if index >= len(primary) and primary: |
| 1182 | followed_from = primary[-1].born_from |
| 1183 | return SymlogResolution( |
| 1184 | symbol=addr, |
| 1185 | index=index, |
| 1186 | content_id=entry.new_content_id, |
| 1187 | commit_id=entry.commit_id, |
| 1188 | operation=entry.operation, |
| 1189 | timestamp=entry.timestamp, |
| 1190 | followed_from=followed_from, |
| 1191 | author=entry.author, |
| 1192 | ) |
| 1193 | |
| 1194 | entries = read_symlog(repo_root, addr, limit=100_000, follow=False) |
| 1195 | index = _resolve_symlog_index(inner, entries) |
| 1196 | entry = entries[index] |
| 1197 | return SymlogResolution( |
| 1198 | symbol=addr, |
| 1199 | index=index, |
| 1200 | content_id=entry.new_content_id, |
| 1201 | commit_id=entry.commit_id, |
| 1202 | operation=entry.operation, |
| 1203 | timestamp=entry.timestamp, |
| 1204 | followed_from=None, |
| 1205 | author=entry.author, |
| 1206 | ) |
| 1207 | |
| 1208 | |
| 1209 | def resolve_symbol_body( |
| 1210 | repo_root: pathlib.Path, |
| 1211 | symbol_addr: str, |
| 1212 | commit_id: str, |
| 1213 | ) -> "dict | None": |
| 1214 | """Return symbol body info for *symbol_addr* at *commit_id*. |
| 1215 | |
| 1216 | Reconstructs the symbol by reading: commit record → snapshot manifest → |
| 1217 | file object → AST parse → slice by ``lineno``/``end_lineno``. Symbol |
| 1218 | ``content_id`` values are AST-computed hashes not stored as raw objects, |
| 1219 | so this walk is the only correct reconstruction path. |
| 1220 | |
| 1221 | Returns a dict with keys ``source``, ``kind``, ``lineno``, ``end_lineno``, |
| 1222 | ``qualified_name``, ``file_path`` on success, or ``None`` on any failure. |
| 1223 | """ |
| 1224 | if "::" not in symbol_addr: |
| 1225 | return None |
| 1226 | file_path, _, symbol_name = symbol_addr.partition("::") |
| 1227 | if not symbol_name: |
| 1228 | return None |
| 1229 | |
| 1230 | try: |
| 1231 | from muse.core.commits import read_commit |
| 1232 | from muse.core.object_store import read_object |
| 1233 | from muse.core.snapshots import read_snapshot |
| 1234 | from muse.plugins.code.ast_parser import adapter_for_path |
| 1235 | |
| 1236 | commit_rec = read_commit(repo_root, commit_id) |
| 1237 | if commit_rec is None: |
| 1238 | return None |
| 1239 | snap_rec = read_snapshot(repo_root, commit_rec.snapshot_id) |
| 1240 | if snap_rec is None: |
| 1241 | return None |
| 1242 | file_obj_id = snap_rec.manifest.get(file_path) |
| 1243 | if file_obj_id is None: |
| 1244 | return None |
| 1245 | raw = read_object(repo_root, file_obj_id) |
| 1246 | if raw is None: |
| 1247 | return None |
| 1248 | |
| 1249 | adapter = adapter_for_path(file_path) |
| 1250 | tree = adapter.parse_symbols(raw, file_path) |
| 1251 | |
| 1252 | # Prefer exact qualified_name match, fall back to bare name. |
| 1253 | found = next( |
| 1254 | (r for r in tree.values() if r["qualified_name"] == symbol_name), |
| 1255 | None, |
| 1256 | ) |
| 1257 | if found is None: |
| 1258 | found = next( |
| 1259 | (r for r in tree.values() |
| 1260 | if r["name"] == symbol_name and r["kind"] != "import"), |
| 1261 | None, |
| 1262 | ) |
| 1263 | if found is None: |
| 1264 | return None |
| 1265 | |
| 1266 | lineno = found["lineno"] |
| 1267 | end_lineno = found["end_lineno"] |
| 1268 | text = raw.decode("utf-8", errors="replace") |
| 1269 | lines = text.splitlines() |
| 1270 | # lineno is 1-based inclusive; end_lineno is 1-based inclusive. |
| 1271 | source = "\n".join(lines[lineno - 1:end_lineno]) |
| 1272 | return { |
| 1273 | "source": source, |
| 1274 | "kind": found["kind"], |
| 1275 | "lineno": lineno, |
| 1276 | "end_lineno": end_lineno, |
| 1277 | "qualified_name": found["qualified_name"], |
| 1278 | "file_path": file_path, |
| 1279 | } |
| 1280 | except Exception: |
| 1281 | return None |
File History
1 commit
sha256:986277e88b9de700cfb77ed3385e7b344df08fbb801c9615906da7df9c45c91d
feat: muse#53 item 12 — muse symlog prune-orphans
Sonnet 5
patch
1 day ago