gc.py
python
sha256:c0eba5ad689cec79f4a3fcdc4f5da78556cb4b8cb7b330f944634356c10379ed
chore: pivot to nightly channel — bump version to 0.2.0.dev…
Sonnet 5
patch
12 days ago
| 1 | """``muse gc`` — garbage-collect unreachable objects. |
| 2 | |
| 3 | Content-addressed storage accumulates blobs that no live commit can reach. |
| 4 | These orphaned objects are safe to delete. ``muse gc`` walks the full commit |
| 5 | graph from every live branch and tag, marks every referenced object as |
| 6 | reachable, then removes the rest. |
| 7 | |
| 8 | A grace period (``--grace-period``, default 30 s) prevents deleting objects |
| 9 | written by a concurrent ``muse commit`` that has not yet created its commit |
| 10 | record. |
| 11 | |
| 12 | Usage:: |
| 13 | |
| 14 | muse gc # remove unreachable objects (30 s grace) |
| 15 | muse gc --full # also prune orphaned commits + snapshots |
| 16 | muse gc --clear-symbol-cache # delete the symbol cache (forces re-parse) |
| 17 | muse gc --dry-run # show what would be removed, touch nothing |
| 18 | muse gc --verbose # print each removed object ID |
| 19 | muse gc --grace-period 0 # no grace — useful in tests / CI |
| 20 | muse gc --json # machine-readable output |
| 21 | |
| 22 | Exit codes:: |
| 23 | |
| 24 | 0 — success (even if nothing was collected) |
| 25 | 1 — bad arguments or internal error |
| 26 | """ |
| 27 | |
| 28 | import argparse |
| 29 | import json |
| 30 | import logging |
| 31 | import sys |
| 32 | from typing import TypedDict |
| 33 | |
| 34 | from muse.core.envelope import EnvelopeJson, make_envelope |
| 35 | from muse.core.paths import symbol_cache_path as _symbol_cache_path |
| 36 | from muse.core.errors import ExitCode |
| 37 | from muse.core.gc import _DEFAULT_GRACE_PERIOD_SECONDS, _DEFAULT_REFLOG_EXPIRE_DAYS, _DEFAULT_SYMLOG_EXPIRE_DAYS, run_gc, prune_stale_remote_refs |
| 38 | from muse.core.repo import require_repo |
| 39 | from muse.core.timing import start_timer |
| 40 | from muse.core.validation import sanitize_display |
| 41 | |
| 42 | logger = logging.getLogger(__name__) |
| 43 | |
| 44 | # --------------------------------------------------------------------------- |
| 45 | # Typed JSON schema |
| 46 | # --------------------------------------------------------------------------- |
| 47 | |
| 48 | class _GcJson(EnvelopeJson): |
| 49 | """JSON output for ``muse gc --json``. |
| 50 | |
| 51 | Inherits the 6 standard envelope fields from :class:`~muse.core.envelope.EnvelopeJson`. |
| 52 | GC warnings (symlink-skip notices, corrupt snapshot alerts, etc.) are surfaced |
| 53 | via the envelope ``warnings`` list rather than a separate field. |
| 54 | |
| 55 | Fields |
| 56 | ------ |
| 57 | status "ok" on success; "error" on bad args or internal error. |
| 58 | error Empty on success; error message on failure. |
| 59 | mode "conservative" (default) or "tight" (--full). |
| 60 | collected_count Total objects collected. |
| 61 | collected_bytes Total bytes freed. |
| 62 | reachable_count Total objects still reachable after GC. |
| 63 | commits_reachable Reachable commits. |
| 64 | commits_collected Commits collected. |
| 65 | commits_collected_bytes Bytes freed from commit objects. |
| 66 | snapshots_reachable Reachable snapshots. |
| 67 | snapshots_collected Snapshots collected. |
| 68 | snapshots_collected_bytes Bytes freed from snapshot objects. |
| 69 | grace_period_seconds Grace period used (objects newer than this are kept). |
| 70 | dry_run True when no objects were actually deleted. |
| 71 | full True when --full (tight mode) was used. |
| 72 | collected_ids Sorted list of all collected object IDs. |
| 73 | collected_commit_ids Sorted list of collected commit IDs. |
| 74 | collected_snapshot_ids Sorted list of collected snapshot IDs. |
| 75 | symbol_cache_cleared True when the symbol cache was cleared. |
| 76 | symbol_cache_bytes Bytes freed from the symbol cache. |
| 77 | """ |
| 78 | |
| 79 | status: str |
| 80 | error: str |
| 81 | mode: str |
| 82 | collected_count: int |
| 83 | collected_bytes: int |
| 84 | reachable_count: int |
| 85 | commits_reachable: int |
| 86 | commits_collected: int |
| 87 | commits_collected_bytes: int |
| 88 | snapshots_reachable: int |
| 89 | snapshots_collected: int |
| 90 | snapshots_collected_bytes: int |
| 91 | grace_period_seconds: int |
| 92 | dry_run: bool |
| 93 | full: bool |
| 94 | collected_ids: list[str] |
| 95 | collected_commit_ids: list[str] |
| 96 | collected_snapshot_ids: list[str] |
| 97 | symbol_cache_cleared: bool |
| 98 | symbol_cache_bytes: int |
| 99 | stale_remote_refs_collected: int |
| 100 | stale_remote_refs_bytes: int |
| 101 | reflog_expired: int |
| 102 | symlog_expired: int |
| 103 | |
| 104 | class _GcErrorJson(EnvelopeJson): |
| 105 | """JSON error output for ``muse gc --json`` on invalid arguments.""" |
| 106 | |
| 107 | status: str |
| 108 | error: str |
| 109 | |
| 110 | # --------------------------------------------------------------------------- |
| 111 | # Helpers |
| 112 | # --------------------------------------------------------------------------- |
| 113 | |
| 114 | def _fmt_bytes(n: int) -> str: |
| 115 | """Human-readable byte count (B / KiB / MiB).""" |
| 116 | if n < 1024: |
| 117 | return f"{n} B" |
| 118 | if n < 1024 * 1024: |
| 119 | return f"{n / 1024:.1f} KiB" |
| 120 | return f"{n / (1024 * 1024):.1f} MiB" |
| 121 | |
| 122 | # --------------------------------------------------------------------------- |
| 123 | # Registration |
| 124 | # --------------------------------------------------------------------------- |
| 125 | |
| 126 | def register(subparsers: "argparse._SubParsersAction[argparse.ArgumentParser]") -> None: |
| 127 | """Register the ``gc`` subcommand.""" |
| 128 | parser = subparsers.add_parser( |
| 129 | "gc", |
| 130 | help="Remove unreachable objects from the Muse object store.", |
| 131 | description=__doc__, |
| 132 | formatter_class=argparse.RawDescriptionHelpFormatter, |
| 133 | ) |
| 134 | parser.add_argument( |
| 135 | "--dry-run", "-n", |
| 136 | action="store_true", |
| 137 | help="Show what would be removed without deleting anything.", |
| 138 | ) |
| 139 | parser.add_argument( |
| 140 | "--verbose", "-v", |
| 141 | action="store_true", |
| 142 | help="Print each object ID that is (or would be) removed.", |
| 143 | ) |
| 144 | parser.add_argument( |
| 145 | "--grace-period", |
| 146 | type=int, |
| 147 | default=_DEFAULT_GRACE_PERIOD_SECONDS, |
| 148 | dest="grace_period", |
| 149 | metavar="SECONDS", |
| 150 | help=( |
| 151 | "Skip objects written within the last SECONDS seconds even if " |
| 152 | "unreachable (protects concurrent commits). Default: " |
| 153 | f"{_DEFAULT_GRACE_PERIOD_SECONDS}." |
| 154 | ), |
| 155 | ) |
| 156 | parser.add_argument( |
| 157 | "--json", "-j", action="store_true", dest="json_out", |
| 158 | help="Emit machine-readable JSON instead of human text.", |
| 159 | ) |
| 160 | parser.add_argument( |
| 161 | "--full", |
| 162 | action="store_true", |
| 163 | help=( |
| 164 | "Also prune orphaned commit and snapshot objects from " |
| 165 | ".muse/objects/sha256/, mirroring git prune. Reachability " |
| 166 | "is computed from live branch refs, not all files in the store." |
| 167 | ), |
| 168 | ) |
| 169 | parser.add_argument( |
| 170 | "--clear-symbol-cache", |
| 171 | action="store_true", |
| 172 | dest="clear_symbol_cache", |
| 173 | help=( |
| 174 | "Delete the symbol cache (.muse/cache/symbols.json). " |
| 175 | "Forces a full re-parse on the next ``muse code`` invocation. " |
| 176 | "Use this when the cache is stale — e.g. after installing a new " |
| 177 | "language grammar (tree-sitter-markdown, etc.)." |
| 178 | ), |
| 179 | ) |
| 180 | parser.set_defaults(func=run) |
| 181 | |
| 182 | # --------------------------------------------------------------------------- |
| 183 | # Command implementation |
| 184 | # --------------------------------------------------------------------------- |
| 185 | |
| 186 | def run(args: argparse.Namespace) -> None: |
| 187 | """Remove unreachable objects from the Muse object store. |
| 188 | |
| 189 | Identifies and removes blobs no longer referenced by any commit, snapshot, |
| 190 | branch, or tag. The reachability walk always completes before any deletion |
| 191 | occurs. Use ``--full`` to also prune orphaned commits and snapshots; |
| 192 | ``--dry-run`` to preview without writing. |
| 193 | |
| 194 | Agent quickstart |
| 195 | ---------------- |
| 196 | :: |
| 197 | |
| 198 | muse gc --json |
| 199 | muse gc --dry-run --json |
| 200 | muse gc --full --json |
| 201 | muse gc --grace-period 0 --json |
| 202 | |
| 203 | JSON fields |
| 204 | ----------- |
| 205 | status ``"ok"`` on success. |
| 206 | mode ``"tight"`` (``--full``) or ``"conservative"``. |
| 207 | collected_count Number of blobs deleted. |
| 208 | collected_bytes Bytes reclaimed from blobs. |
| 209 | reachable_count Blobs retained as reachable. |
| 210 | commits_collected Commits pruned (``--full`` only). |
| 211 | snapshots_collected Snapshots pruned (``--full`` only). |
| 212 | dry_run ``true`` if no objects were actually deleted. |
| 213 | |
| 214 | Exit codes |
| 215 | ---------- |
| 216 | 0 Success. |
| 217 | 1 Invalid arguments. |
| 218 | 2 Not inside a Muse repository. |
| 219 | 3 I/O error reading or writing the object store. |
| 220 | """ |
| 221 | elapsed = start_timer() |
| 222 | dry_run: bool = args.dry_run |
| 223 | verbose: bool = args.verbose |
| 224 | json_out: bool = args.json_out |
| 225 | grace_period: int = args.grace_period |
| 226 | full: bool = args.full |
| 227 | clear_symbol_cache: bool = args.clear_symbol_cache |
| 228 | |
| 229 | if grace_period < 0: |
| 230 | if json_out: |
| 231 | print(json.dumps(_GcErrorJson( |
| 232 | **make_envelope(elapsed, exit_code=ExitCode.USER_ERROR), |
| 233 | status="error", |
| 234 | error=f"--grace-period must be \u2265 0 (got {grace_period})", |
| 235 | ))) |
| 236 | else: |
| 237 | print( |
| 238 | f"❌ --grace-period must be ≥ 0 (got {grace_period}).", |
| 239 | file=sys.stderr, |
| 240 | ) |
| 241 | raise SystemExit(ExitCode.USER_ERROR) |
| 242 | |
| 243 | repo_root = require_repo() |
| 244 | |
| 245 | # --- Reflog TTL — read from config, fall back to 90 days ----------------- |
| 246 | from muse.cli.config import get_config_value as _get_config_value |
| 247 | _ttl_raw = _get_config_value("reflog.expire-days", repo_root) |
| 248 | try: |
| 249 | reflog_expire_days: int = int(_ttl_raw) if _ttl_raw is not None else _DEFAULT_REFLOG_EXPIRE_DAYS |
| 250 | except (ValueError, TypeError): |
| 251 | reflog_expire_days = _DEFAULT_REFLOG_EXPIRE_DAYS |
| 252 | |
| 253 | # --- Symlog TTL — read from config, fall back to 90 days ----------------- |
| 254 | _sl_ttl_raw = _get_config_value("symlog.expire-days", repo_root) |
| 255 | try: |
| 256 | symlog_expire_days: int = int(_sl_ttl_raw) if _sl_ttl_raw is not None else _DEFAULT_SYMLOG_EXPIRE_DAYS |
| 257 | except (ValueError, TypeError): |
| 258 | symlog_expire_days = _DEFAULT_SYMLOG_EXPIRE_DAYS |
| 259 | |
| 260 | # --- Symbol cache clear -------------------------------------------------- |
| 261 | symbol_cache_cleared = False |
| 262 | symbol_cache_bytes = 0 |
| 263 | if clear_symbol_cache: |
| 264 | cache_path = _symbol_cache_path(repo_root) |
| 265 | if cache_path.is_file(): |
| 266 | symbol_cache_bytes = cache_path.stat().st_size |
| 267 | if not dry_run: |
| 268 | cache_path.unlink() |
| 269 | symbol_cache_cleared = True |
| 270 | |
| 271 | result = run_gc( |
| 272 | repo_root, |
| 273 | dry_run=dry_run, |
| 274 | grace_period_seconds=grace_period, |
| 275 | full=full, |
| 276 | reflog_expire_days=reflog_expire_days, |
| 277 | symlog_expire_days=symlog_expire_days, |
| 278 | ) |
| 279 | |
| 280 | # --- Stale remote tracking ref pruning (--full only) --------------------- |
| 281 | if full: |
| 282 | from muse.cli.config import list_remotes |
| 283 | configured_names = {r["name"] for r in list_remotes(repo_root)} |
| 284 | prune_stale_remote_refs(repo_root, configured_names, result, dry_run=dry_run) |
| 285 | |
| 286 | if json_out: |
| 287 | print(json.dumps(_GcJson( |
| 288 | **make_envelope(elapsed, warnings=result.warnings), |
| 289 | status="ok", |
| 290 | error="", |
| 291 | mode="tight" if full else "conservative", |
| 292 | collected_count=result.collected_count, |
| 293 | collected_bytes=result.collected_bytes, |
| 294 | reachable_count=result.reachable_count, |
| 295 | commits_reachable=result.commits_reachable, |
| 296 | commits_collected=result.commits_collected, |
| 297 | commits_collected_bytes=result.commits_collected_bytes, |
| 298 | snapshots_reachable=result.snapshots_reachable, |
| 299 | snapshots_collected=result.snapshots_collected, |
| 300 | snapshots_collected_bytes=result.snapshots_collected_bytes, |
| 301 | grace_period_seconds=result.grace_period_seconds, |
| 302 | dry_run=result.dry_run, |
| 303 | full=result.full, |
| 304 | collected_ids=sorted(result.collected_ids), |
| 305 | collected_commit_ids=sorted(result.collected_commit_ids), |
| 306 | collected_snapshot_ids=sorted(result.collected_snapshot_ids), |
| 307 | symbol_cache_cleared=symbol_cache_cleared, |
| 308 | symbol_cache_bytes=symbol_cache_bytes, |
| 309 | stale_remote_refs_collected=result.stale_remote_refs_collected, |
| 310 | stale_remote_refs_bytes=result.stale_remote_refs_bytes, |
| 311 | reflog_expired=result.reflog_expired, |
| 312 | symlog_expired=result.symlog_expired, |
| 313 | ))) |
| 314 | return |
| 315 | |
| 316 | prefix = "[dry-run] " if dry_run else "" |
| 317 | action = "Would remove" if dry_run else "Removed" |
| 318 | |
| 319 | if verbose and result.collected_ids: |
| 320 | print(f"{prefix}Unreachable objects:") |
| 321 | for oid in sorted(result.collected_ids): |
| 322 | print(f" {sanitize_display(oid)}") |
| 323 | |
| 324 | print( |
| 325 | f"{prefix}{action} {result.collected_count} object(s) " |
| 326 | f"({_fmt_bytes(result.collected_bytes)}) " |
| 327 | f"in {result.duration_ms:.1f}ms " |
| 328 | f"[{result.reachable_count} reachable]" |
| 329 | ) |
| 330 | |
| 331 | if full: |
| 332 | print( |
| 333 | f"{prefix}{action} {result.commits_collected} commit(s) " |
| 334 | f"({_fmt_bytes(result.commits_collected_bytes)}) " |
| 335 | f"[{result.commits_reachable} reachable]" |
| 336 | ) |
| 337 | print( |
| 338 | f"{prefix}{action} {result.snapshots_collected} snapshot(s) " |
| 339 | f"({_fmt_bytes(result.snapshots_collected_bytes)}) " |
| 340 | f"[{result.snapshots_reachable} reachable]" |
| 341 | ) |
| 342 | if result.stale_remote_refs_collected: |
| 343 | print( |
| 344 | f"{prefix}{action} {result.stale_remote_refs_collected} stale remote " |
| 345 | f"tracking ref(s) ({_fmt_bytes(result.stale_remote_refs_bytes)})" |
| 346 | ) |
| 347 | |
| 348 | if symbol_cache_cleared: |
| 349 | cache_action = "Would clear" if dry_run else "Cleared" |
| 350 | print(f"{prefix}{cache_action} symbol cache ({_fmt_bytes(symbol_cache_bytes)})") |
| 351 | |
| 352 | if result.reflog_expired: |
| 353 | reflog_action = "Would expire" if dry_run else "Expired" |
| 354 | print(f"{prefix}{reflog_action} {result.reflog_expired} reflog entry/entries") |
| 355 | |
| 356 | if result.symlog_expired: |
| 357 | symlog_action = "Would expire" if dry_run else "Expired" |
| 358 | print(f"{prefix}{symlog_action} {result.symlog_expired} symlog entry/entries") |
File History
10 commits
sha256:c287f599c5429903a139eadf3c5db5d930520e57cb0c3c575d9570e953c3b2d6
chore: bump version to 0.2.0.dev2 — nightly.2
Sonnet 4.6
patch
11 days ago
sha256:8de4334a98c945aace420969d389ad678aa926d4ab4e886b2ac4c4241cb3bf2b
revert: keep pyproject.toml in canonical PEP 440 form
Sonnet 4.6
patch
15 days ago
sha256:50bb615c573aa4b928fca75b60f82ba2a659910066507cec6a95c412ae22ccb9
docs: add issue docs for push have-negotiation bug (#55) an…
Sonnet 4.6
18 days ago
sha256:d90d175cded68aae1d4ffcf4858917854195d0cd8ce1fe73cee4dbc02541cb74
chore: trigger push to surface null-OID paths
Sonnet 4.6
23 days ago
sha256:e452ad9a6ace6ccc6d875a35e06caf9da5576a970c1c36133b69a891ce5fefa8
chore: prebuild timing test
Sonnet 4.6
34 days ago
sha256:0008ab6695e3e064b3e236b24fd19e538fef6a588eb0d211622f4466d919c0b1
merge: pull staging/dev — advance to 0.2.0rc12
Sonnet 4.6
patch
35 days ago
sha256:9c33d61749fff814c5226d5386aa2af7064c2c02788594a25fdd709358132eea
fix: _PROPOSAL_PREFIX_RESOLVE_LIMIT 200 → 100 to match hub …
Sonnet 4.6
47 days ago
sha256:36c3cb3e76619d4c30a6d9bf81b5ec4ff148e30dcfed913e3114ca7b43b81c7e
fix: rename objects→blobs in push client and all stale test…
Sonnet 4.6
patch
50 days ago
sha256:c06a9b9b9fee26c68ea725b44d54b2c0a171301ce9de746d5b656617b4463a9a
fix: repair four test failures from post-migration audit
Sonnet 4.6
patch
56 days ago
sha256:1900655993c83c4107067375548a7be823e471d2515830842f1a12cba4bd3cdf
fix: unified object store migration — idempotent writes, JS…
Sonnet 4.6
minor
⚠
56 days ago