fix(hnsw): address @igorls's #1173 review

Restore three `_pin_hnsw_threads` tests that the previous integrity-gate
commit deleted. The function is still live code on develop (defined at
chroma.py:207, called from chroma.py:705 + mcp_server.py), so the
deletion left the import unused (ruff F401) and dropped coverage on a
function unrelated to this PR's scope. Restored verbatim from main.

Plus three nits @igorls flagged:

- **Thread-safety doc**: `_quarantined_paths` mutation is lock-free;
  documented that idempotency of `quarantine_stale_hnsw` is the safety
  property (concurrent same-palace calls produce a benign redundant
  rename attempt that no-ops, no need for a lock).
- **Pickle protocol assumption**: `_segment_appears_healthy` requires
  PROTO ≥ 2 (`0x80`). Documented; matches what chromadb writes today,
  and a future protocol-0/1 emission would conservatively quarantine
  + lazy-rebuild rather than mis-classify as healthy.
- Class-level vs module-level scope: keeping class-level — the
  conftest reset is the controlled case, and module-level wouldn't
  remove the foot-gun, just relocate it. Conftest reset documented in
  the existing comment is the right pattern for test isolation.

Style nit (`Path(marker).touch()` vs `open(marker, "a").close()`)
deferred — that pattern lives in #1177's territory, not #1173's.

37/37 tests pass on the PR branch.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
jp
2026-04-26 13:20:14 -07:00
parent 74ff5e6b98
commit 942aae3ed5
2 changed files with 63 additions and 0 deletions
+49
View File
@@ -578,3 +578,52 @@ def test_make_client_quarantines_each_palace_independently(tmp_path, monkeypatch
assert calls == [palace_a, palace_b]
# ── _pin_hnsw_threads (per-process retrofit, separate from this PR's gate) ──
def test_pin_hnsw_threads_retrofits_legacy_collection(tmp_path):
"""Legacy collections (created without num_threads) get the retrofit applied."""
palace_path = tmp_path / "legacy-palace"
palace_path.mkdir()
client = chromadb.PersistentClient(path=str(palace_path))
col = client.create_collection(
"mempalace_drawers",
metadata={"hnsw:space": "cosine"}, # no num_threads — legacy
)
assert col.configuration_json.get("hnsw", {}).get("num_threads") is None
_pin_hnsw_threads(col)
assert col.configuration_json["hnsw"]["num_threads"] == 1
def test_pin_hnsw_threads_swallows_all_errors():
"""Retrofit never raises even when collection.modify explodes."""
class _ExplodingCollection:
def modify(self, *args, **kwargs):
raise RuntimeError("boom")
_pin_hnsw_threads(_ExplodingCollection()) # must not raise
def test_get_collection_applies_retrofit_on_existing_palace(tmp_path):
"""ChromaBackend.get_collection(create=False) applies the retrofit."""
palace_path = tmp_path / "palace"
palace_path.mkdir()
# Simulate a legacy palace: create collection without num_threads
bootstrap_client = chromadb.PersistentClient(path=str(palace_path))
bootstrap_client.create_collection("mempalace_drawers", metadata={"hnsw:space": "cosine"})
del bootstrap_client # drop reference so a fresh client reopens cleanly
wrapper = ChromaBackend().get_collection(
str(palace_path),
collection_name="mempalace_drawers",
create=False,
)
assert wrapper._collection.configuration_json["hnsw"]["num_threads"] == 1