Port the vhlab-library-matlab functions NDI-matlab depends on: 7 missing, 2 already done #6

Companion to #4 (the bridge convention). That issue is about recording the mapping; this one is the concrete work list it would have made obvious.

NDI-python is porting NDI-matlab, and NDI-matlab calls into vhlab-library-matlab directly. Every such call is a function that has to exist here before the corresponding NDI-python port can complete. One is blocking right now (Waltham-Data-Science/NDI-python#134).

How this list was derived

Not from the existing table alone. NDI-matlab/src/ndi/docs/developer_notes/vhlabLibraryMatlab.md lists eleven names, but it mixes genuine calls with file-format conventions and does not say what is already ported, so I re-derived it:

1. Collected the 892 function names defined in vhlab-library-matlab.
2. Dropped the 32 also defined in NDI-matlab or vhlab-toolbox-matlab, so a name that resolves elsewhere is not attributed to the library.
3. Scanned NDI-matlab's .m sources for word-boundary call sites, separating free calls from obj.method(...) calls.
4. Hand-filtered the result. The raw scan flagged delete, extract, parameters, reference, tuning_curve, thresholds, spikedata, clock and setparameters — all MATLAB builtins, NDI's own function definitions, or local variables, not library calls.
5. Checked what is already ported here by symbol, not filename — which matters, since findassociate and friends live inside vhlib/md/, not in files of their own.

Tier 1 — free functions, port these

Start with neural_response_significance — it is a leaf and it unblocks a port today

ndi.app.oridirtuning.calculate_oridir_indexes cannot be finished without it (Waltham-Data-Science/NDI-python#134; the tuning-curve half shipped in #145 with the index half explicitly deferred to this).

It has no library dependencies at all — checked, not assumed. It needs colvec (already in vhlab-toolbox-python) and anova1 (MATLAB Statistics Toolbox → scipy.stats.f_oneway). The whole function is ~15 lines:

function [sigp,sigpb] = neural_response_significance(resp)
% SIGP  - P value of an ANOVA across all stimulus conditions.
% SIGPB - the same, including the blank if available; identical to SIGP if not.

sigp is a one-way ANOVA over the groups in resp.ind; sigpb adds resp.blankind as one more group. oridirtuning.m:206 maps them as visual_response_anova_p = sigpb, across_stimuli_anova_p = sigp.

Tier 2 — method calls, so the class is what is needed

These appear as obj.method(...), so porting the bare function name would be wrong.

vhlib/md/measureddata.py already has MeasuredData with associate, findassociate, getassociate, numassociates, disassociate, associates2struct and update — but not get_data, get_intervals or set_intervals. Adding those three to the existing class is likely the smallest useful next step after Tier 1.

Both get_data and getparameters have several MATLAB definitions across different @class folders, so the porting decision needs to name which class NDI-matlab's call site actually reaches. All three call sites are in NDI-matlab/src/ndi/+ndi/+setup/+conv/+vhlab/importMeasuredDataCells.m and +ndi/+daq/+metadatareader/NewStimStims.m.

A name to be careful with: getstimscript

vhlabLibraryMatlab.md lists it, and my scan excluded it — because it is defined in both vhlab-library-matlab/StimulusDecoding/ and vhlab-toolbox-matlab. Whichever is on the MATLAB path first wins, so which one NDI-matlab actually calls is genuinely ambiguous from the source. That needs a human answer before porting, and it is exactly the kind of thing a bridge decision_log (#4) exists to record once.

Transitive dependencies

BFS from the nine call sites pulls in 17 more library functions, almost all in two clusters:

• @measureddata and its analysis objects — associate, disassociate, findassociate, getassociate, numassociates, set_intervals, spikedata, contextmenu, repairhandles, verifywhere, delete. Four of these are already ported inside vhlib/md/.
• The old spike importers — cksmultipleunit, cksmeasureddata, get_data_par, loaddata, vhintan_loadcell.

Worth noting before anyone starts on Tier 2: the second cluster lives under OldSpikeImporters/, and vhintan_importcells reaches it. Whether that whole lineage should be ported at all, or whether NDI-matlab's two call sites are themselves legacy, is a decision rather than a task — NDI-matlab/+ndi/+gui/gui.m (the vhintan_* caller) and importMeasuredDataCells.m are both conversion/legacy paths.

Not calls — conventions only

vhlabLibraryMatlab.md also lists testdirinfo, vhintan_channelgrouping, vhintan_intan2spike2time and vhspike2_channelgrouping. Its own Comment column marks these "Reference to file format convention" — NDI-matlab names them in strings or comments rather than calling them. They need no port, and are listed here so the next person does not re-derive that.

Suggested order

1. neural_response_significance — leaf, ~15 lines, unblocks NDI-python #134.
2. get_intervals / get_data / set_intervals on the existing MeasuredData.
3. Decide on getstimscript (which repo's definition is authoritative).
4. Decide whether the OldSpikeImporters / vhintan_* lineage is in scope at all.

If #4 lands first, each of these gets a bridge entry as it is ported, and the "already done" column above stops needing to be re-derived by scanning.

🤖 Generated with Claude Code

https://claude.ai/code/session_01CqrNhfGnhKGcQvNiP2LNXK