[stevevanhooser]

Summary

Mirrors VH-Lab/NDI-matlab#1127. CI pairs this PR with that PR's branch, which has the same name.

1. New vhlab DAQ system vhneuropixelsGLX_nidq. ndi_common/daq_systems/vhlab/vhneuropixelsGLX_nidq.json is identical to the MATLAB file. It reads the SpikeGLX NI-DAQ stream (*.nidq.bin / *.nidq.meta) with the NDR neuropixelsGLX reader. That lets you name NI-DAQ digital lines in epochprobemap.txt; for example, an optogenetic fiber on XD4 is vhneuropixelsGLX_nidq:dimp5 (SpikeGLX XDn is channel di(n+1)). vhneuropixelsGLX reads the IMEC .ap.bin, whose digital lines are the IMEC sync bits, so those lines weren't reachable before. A blank vhlab session now has ten DAQ systems.

2. Digital-edge "on" events were one sample early. In ndi_daq_reader_mfdaq._read_derived_events (the non-ingested read for dep, den, dimp, dimn), each on-edge was np.where(...)[0], which is the last sample before the transition. MATLAB's 1 + find(...), this port's off-edges, and its ingested read all use the first sample after it. A line going high at 1.000 s at 1 kHz reported 0.999 s. This adds the missing 1 +.

The MATLAB PR also fixes dimn never returning its off-edges. This port's dimn branches were already correct, so there was nothing to port for that.

Changes

• src/ndi/ndi_common/daq_systems/vhlab/vhneuropixelsGLX_nidq.json: new.
• src/ndi/daq/mfdaq.py: the two on-edge lines in _read_derived_events.
• Bridge entries synced to NDI-matlab dfba35572, the head of the paired branch, each with a decision_log note:
  • mfdaq_reader (+ndi/+daq/+reader/mfdaq.m) was at 5c5784c6a. The only commit since then is the dimn fix, which needed nothing here; the on-edge fix above is what this port changes.
  • setup (+ndi/+setup/) was at fd7ddbf06. The only net change is a comment in vhlab.m.
• tests/test_daq_derived_digital_events.py: new. Checks dep, den, dimp and dimn against a reader with pulses at 1.0–1.1 s and 3.0–3.1 s. dimp and dimn fail without the fix (0.999 instead of 1.0).
• tests/test_vhneuropixelsglx_nidq.py: new. The system loads with the NDR neuropixelsGLX reader, the epochdir navigator and the .nidq file patterns.
• tests/test_vhprairieview.py: the blank-session count goes from 9 to 10.
• CHANGELOG.md: Added and Fixed entries.

Merge order

Merge VH-Lab/NDI-matlab#1127 first, as a merge commit rather than a squash. The two bridge hashes point at dfba35572, which exists only on the paired NDI-matlab branch until then. Once this PR is on main, its CI falls back to NDI-matlab main. If dfba35572 isn't there, because the MATLAB PR hasn't merged or was squashed, the bridge-hash checks will fail.

Testing

Run locally:

• black --check and ruff check pass.
• pytest tests/ passes (5,547 tests). The cloud tests were left out because they need cloud credentials.
• tests/test_matlab_bridge_hashes.py passes with NDI_BRIDGE_CHECK_STRICT=1 against the paired NDI-matlab branch, as in CI.
• The symmetry make_artifacts and read_artifacts suites pass, apart from two failures that fail the same way on main here: a blocked dataset download and the optional pyabf package missing.

🤖 Generated with Claude Code

https://claude.ai/code/session_011v7R5uyvdiE4Ex6Q7UkurQ