[stevevanhooser]

Problem

ndi.setup.daq.reader.mfdaq.stimulus.vhlabvisspike2 was a bare subclass of the CED Spike2 reader. The mk1–mk3 and e1–e3 stimulus channels were therefore looked up as channel numbers in spike2data.smr. MATLAB instead decodes them from the epoch directory.

Change

This ports +ndi/+setup/+daq/+reader/+mfdaq/+stimulus/vhlabvisspike2.m, covering epochclock, getchannelsepoch, samplerate and readevents_epochsamples_native:

• It reads stimtimes.txt and stims.mat (MTI2) through vhlib.StimDecode.read_stimtimes_txt / getstimscript and uses vlt.data.dropnan. These come from vhlab-library-python and vhlab-toolbox-python, which are already dependencies, so nothing is reimplemented. vhlab-NewStim-python is not needed because only the stimid and startStopTimes fields of MTI2 are read. ndi.daq.metadatareader.NewStimStims reads stimulus parameters from the same file and does not overlap with this.
• It applies stimtimes2stimtimes_mapping.txt and mti2stimtimes_mapping.txt when they are present. Each lists 1-based indices to keep, and NaN entries are dropped.
• When the trigger counts differ it raises a ValueError with MATLAB's wording.
• When a trial's Spike2 id differs from mod(mti.stimid,256) it raises VHLabVisSpike2StimulusMismatchError. This is a ValueError subclass whose .identifier is ndi:setup:daq:reader:mfdaq:stimulus:vhlabvisspike2:StimulusMismatch. The message follows vhlab importer: robust handling of legacy cell intervals and cell names VH-Lab/NDI-matlab#1114 (9fe55861a): Epoch t00006: stimulus number N does not line up …, followed by the path and a pointer to the mapping files.
• Stimulus on/off/setup/clear times are computed with timeshift = stimtimes(i) - startStopTimes(2), and results are limited to the window t0 <= t <= t1.

Bridge

INTERFACE UPDATE: this adds a new src/ndi/setup/daq/reader/mfdaq/stimulus/ndi_matlab_python_bridge.yaml with a vhlabvisspike2 class entry, examined against NDI-matlab fe64a9f53. It records:

• the dependency on #1114;
• that matlab_branch is deliberately not set, because the file is already on main. When #1114 merges, the drift check will flag the entry for re-examination;
• one MATLAB quirk that is not copied: MATLAB indexes marker and event results inconsistently ({i} vs {end+1}).

Tests

tests/test_setup_vhlabvisspike2.py builds synthetic epoch directories containing stimtimes.txt and a minimal stims.mat. It covers:

• records that agree, including an id ≥ 256;
• the time window;
• a count mismatch;
• an id mismatch, checking that the message names the epoch;
• a mapping file that fixes a run stopped early (Spike2 has 3 triggers, stims.mat has 5);
• a stimtimes mapping file containing NaN.

black --check, ruff check and the full pytest tests/ all pass. The bridge tests also pass when run against a full NDI-matlab checkout.

Follow-up

VH-Lab/NDI-matlab#1114 has not merged yet. Once it does:

• add a note to the setup entry in the root bridge (src/ndi/ndi_matlab_python_bridge.yaml) saying that #1114's other changes are all in +ndi/+setup/+conv/+vhlab, which has no Python port;
• re-sync this entry's hash.

---

Generated by Claude Code