npc-sync 0.2.7__tar.gz → 0.2.8__tar.gz

{npc_sync-0.2.7/src/npc_sync.egg-info → npc_sync-0.2.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: npc-sync
-Version: 0.2.7
+Version: 0.2.8
 Summary: Repackaging of AllenSDK.SyncDataset with extra functionality, compatible with data in the cloud.
 Author-email: Ben Hardcastle <ben.hardcastle@alleninstitue.org>
 License: MIT

{npc_sync-0.2.7 → npc_sync-0.2.8}/pyproject.toml

@@ -16,7 +16,7 @@ dependencies = [
     "scipy>=1.13.1",
     "npc-stim>=0.1.23",
 ]
-version = "0.2.7"
+version = "0.2.8"
 classifiers = [
     "Development Status :: 3 - Alpha",
     "Programming Language :: Python :: 3",

{npc_sync-0.2.7 → npc_sync-0.2.8}/src/npc_sync/sync.py

@@ -68,6 +68,11 @@ REPLAY_SLOT_SAMPLES = 10_000
 acquisition buffers. Buffer sizes seen so far (50,000 and 10,000 samples) are multiples
 of this. Buffer size is not recorded in the sync file metadata."""
 
+REPLAY_MAX_BUFFER_SAMPLES = 50_000
+"""Largest acquisition buffer (0.5 s at 100 kHz) considered when looking for replayed
+buffers. Identical slots separated by more than this are coincidental matches (e.g. quiet
+or periodic stretches of real data) and are ignored."""
+
 SUSPICIOUS_INTERVAL_THRESHOLD = 0.3 / FRAME_RATE
 DEFINITE_SHORT_INTERVAL_THRESHOLD = 0.05 / FRAME_RATE
 # at 60 fps we should never have diode-flip intervals this small: consider them anomalous
@@ -175,6 +180,7 @@ def find_replayed_buffers(
     times: npt.ArrayLike,
     bits: npt.ArrayLike,
     slot_samples: int = REPLAY_SLOT_SAMPLES,
+    max_buffer_samples: int = REPLAY_MAX_BUFFER_SAMPLES,
 ) -> tuple[tuple[tuple[int, int], ...], int | None]:
     """Find acquisition buffers in a sync event list that are replayed copies of a later
     buffer.
@@ -197,6 +203,8 @@ def find_replayed_buffers(
         Bit state (all lines) of each event.
     slot_samples : int
         Fingerprint slot width in samples. Must divide the buffer size.
+    max_buffer_samples : int
+        Identical slots further apart than this are not treated as replays.
 
     Returns
     -------
@@ -238,6 +246,10 @@ def find_replayed_buffers(
         return (), None
     earlier = slots[:-1][same]
     lags = slots[1:][same] - earlier
+    plausible = lags <= max_buffer_samples // slot_samples
+    earlier, lags = earlier[plausible], lags[plausible]
+    if lags.size == 0:
+        return (), None
     values, counts = np.unique(lags, return_counts=True)
     lag = int(values[np.argmax(counts)])
     if len(values) > 1:

{npc_sync-0.2.7 → npc_sync-0.2.8/src/npc_sync.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: npc-sync
-Version: 0.2.7
+Version: 0.2.8
 Summary: Repackaging of AllenSDK.SyncDataset with extra functionality, compatible with data in the cloud.
 Author-email: Ben Hardcastle <ben.hardcastle@alleninstitue.org>
 License: MIT

{npc_sync-0.2.7 → npc_sync-0.2.8}/tests/test_replay.py

@@ -94,3 +94,13 @@ def test_known_sessions(session: str) -> None:
     if not paths:
         pytest.skip(f"local sync file not available for {session}")
     assert npc_sync.SyncDataset(paths[0]).invalid_intervals == KNOWN[session]
+
+
+def test_ignores_identical_slots_beyond_max_buffer_size() -> None:
+    rng = np.random.default_rng(1)
+    rel = np.sort(rng.choice(10_000, 50, replace=False))
+    state = rng.integers(0, 2**32, 50)
+    # same slot content 31 slots (310,000 samples) apart: not a plausible buffer
+    times = np.concatenate([rel, rel + 310_000])
+    bits = np.concatenate([state, state])
+    assert npc_sync.sync.find_replayed_buffers(times, bits) == ((), None)