Fix duplicate probe_id field in ProbeGroup.to_numpy(complete=True)

[alejoe91]

Problem

ProbeGroup.to_numpy(complete=True) crashes with

ValueError: field 'probe_id' occurs more than once

when a probe has a contact annotation named probe_id. Since 0.4.0, to_numpy always adds its own probe_id field and then also adds every contact annotation (with complete=True), so probe_id appears twice.

How such annotations appear: version mismatch

These annotations come from files written and read by different probeinterface versions:

1. Written with probeinterface >= 0.4.0: a recording's contact_vector property (probegroup.to_numpy(complete=True)) includes a probe_id column.
2. Read with probeinterface < 0.4.0: Probe.from_numpy doesn't know probe_id as a structural field, so it turns the column into a contact annotation. A SortingAnalyzer saved at this point stores the probe with that annotation.
3. Loaded with probeinterface >= 0.4.0: to_numpy(complete=True), called for example by SortingAnalyzer.get_channel_locations(), crashes.

We hit this with Open Ephys data that was compressed with spikeinterface/probeinterface 0.4.x and sorted with spikeinterface 0.104.7/probeinterface 0.3.x. The resulting analyzers can't be loaded with probeinterface 0.4.x.

Fix

to_numpy now skips any probe fields named probe_index or probe_id, so the probegroup's own values are used. Stale annotations stay on the probe but are no longer exported.

Minimal reproducer

import probeinterface as pi
from probeinterface import generate_linear_probe

probe = generate_linear_probe(num_elec=4)
probe.set_device_channel_indices(range(4))
probe.annotate_contacts(probe_id=["0"] * 4)
probegroup = pi.ProbeGroup()
probegroup.add_probe(probe)
probegroup.to_numpy(complete=True)  # ValueError on main

Tests

Added test_to_numpy_with_probe_id_contact_annotation. It fails on main and passes with this fix.

🤖 Generated with Claude Code