0.104.0 bug fixes PRs (#4479)

Co-authored-by: hayleybounds <hayleybounds@gmail.com>
Co-authored-by: Chris Halcrow <57948917+chrishalcrow@users.noreply.github.com>

Author: 3 people

src/spikeinterface/extractors/cbin_ibl.py

@@ -106,14 +106,7 @@ def __init__(self, folder_path=None, load_sync_channel=False, stream_name="ap",
             else:
                 self.set_probe(probe, in_place=True)
 
-            # load num_channels_per_adc depending on probe type
-            ptype = probe.annotations["probe_type"]
-
-            if ptype in [21, 24]:  # NP2.0
-                num_channels_per_adc = 16
-            else:  # NP1.0
-                num_channels_per_adc = 12
-            sample_shifts = get_neuropixels_sample_shifts_from_probe(probe, num_channels_per_adc)
+            sample_shifts = get_neuropixels_sample_shifts_from_probe(probe)
             self.set_property("inter_sample_shift", sample_shifts)
 
         self._kwargs = {

src/spikeinterface/metrics/quality/quality_metrics.py

@@ -77,6 +77,20 @@ def _handle_backward_compatibility_on_load(self):
             if "peak_sign" in self.params["metric_params"]["amplitude_median"]:
                 del self.params["metric_params"]["amplitude_median"]["peak_sign"]
 
+        # TODO: update this once `main_channel_index` PR is merged
+        # global peak_sign used to find appropriate channels for pca metric computation
+        # If not found, use a "peak_sign" set by any metric
+        global_peak_sign_from_params = self.params.get("peak_sign")
+        if global_peak_sign_from_params is None:
+            for metric_params in self.params["metric_params"].values():
+                if "peak_sign" in metric_params:
+                    global_peak_sign_from_params = metric_params["peak_sign"]
+                    break
+            # If still not found, use <0.104.0 default, "neg"
+            if global_peak_sign_from_params is None:
+                global_peak_sign_from_params = "neg"
+            self.params["peak_sign"] = global_peak_sign_from_params
+
     def _set_params(
         self,
         metric_names: list[str] | None = None,

src/spikeinterface/metrics/template/template_metrics.py

@@ -169,6 +169,11 @@ def _handle_backward_compatibility_on_load(self):
             if "waveform_ratios" not in self.params["metric_names"]:
                 self.params["metric_names"].append("waveform_ratios")
 
+        # If original analyzer doesn't have "peaks_data" or "main_channel_templates",
+        # then we can't save this tmp data (important for merges/splits)
+        if "peaks_data" not in self.data:
+            self.tmp_data_to_save = []
+
     def _set_params(
         self,
         metric_names: list[str] | None = None,
@@ -234,12 +239,12 @@ def _prepare_data(self, sorting_analyzer, unit_ids):
             unit_ids = sorting_analyzer.unit_ids
         peak_sign = self.params["peak_sign"]
         upsampling_factor = self.params["upsampling_factor"]
-        min_thresh_detect_peaks_troughs = self.params["min_thresh_detect_peaks_troughs"]
-        edge_exclusion_ms = self.params.get("edge_exclusion_ms", 0.1)
+        min_thresh_detect_peaks_troughs = self.params.get("min_thresh_detect_peaks_troughs", 0.3)
+        edge_exclusion_ms = self.params.get("edge_exclusion_ms", 0.09)
         min_peak_trough_distance_ratio = self.params.get("min_peak_trough_distance_ratio", 0.2)
         min_extremum_distance_samples = self.params.get("min_extremum_distance_samples", 3)
         sampling_frequency = sorting_analyzer.sampling_frequency
-        if self.params["upsampling_factor"] > 1:
+        if upsampling_factor > 1:
             sampling_frequency_up = upsampling_factor * sampling_frequency
         else:
             sampling_frequency_up = sampling_frequency
@@ -249,7 +254,7 @@ def _prepare_data(self, sorting_analyzer, unit_ids):
             m in get_multi_channel_template_metric_names() for m in self.params["metrics_to_compute"]
         )
 
-        operator = self.params["template_operator"]
+        operator = self.params.get("template_operator", "average")
         extremum_channel_indices = get_template_extremum_channel(
             sorting_analyzer, peak_sign=peak_sign, outputs="index", operator=operator
         )
@@ -312,7 +317,7 @@ def _prepare_data(self, sorting_analyzer, unit_ids):
             # multi_channel_templates is a list of 2D arrays of shape (n_times, n_channels)
             tmp_data["multi_channel_templates"] = multi_channel_templates
             tmp_data["channel_locations_multi"] = channel_locations_multi
-            tmp_data["depth_direction"] = self.params["depth_direction"]
+            tmp_data["depth_direction"] = self.params.get("depth_direction", "y")
 
         # Add peaks_info and preprocessed templates to self.data for storage in extension
         columns = []

src/spikeinterface/postprocessing/tests/test_valid_unit_periods.py

@@ -34,8 +34,8 @@ def test_user_defined_periods(self):
                 periods[idx]["unit_index"] = unit_index
                 period_start = num_samples // 4
                 period_duration = num_samples // 2
-                periods[idx]["start_sample_index"] = period_start
-                periods[idx]["end_sample_index"] = period_start + period_duration
+                periods[idx]["start_sample_index"] = period_start - unit_index * 10
+                periods[idx]["end_sample_index"] = period_start + period_duration + unit_index * 10
                 periods[idx]["segment_index"] = segment_index
 
         sorting_analyzer = self._prepare_sorting_analyzer(
@@ -48,8 +48,17 @@ def test_user_defined_periods(self):
             minimum_valid_period_duration=1,
         )
         # check that valid periods correspond to user defined periods
-        ext_periods = ext.get_data(outputs="numpy")
-        np.testing.assert_array_equal(ext_periods, periods)
+        ext_periods_numpy = ext.get_data(outputs="numpy")
+        np.testing.assert_array_equal(ext_periods_numpy, periods)
+
+        # check that `numpy` and `by_unit` outputs are the same
+        ext_periods_by_unit = ext.get_data(outputs="by_unit")
+        for segment_index in range(num_segments):
+            for unit_index, unit_id in enumerate(unit_ids):
+                periods_numpy_seg0 = ext_periods_numpy[ext_periods_numpy["segment_index"] == segment_index]
+                periods_numpy_unit = periods_numpy_seg0[periods_numpy_seg0["unit_index"] == unit_index]
+                period = [(periods_numpy_unit["start_sample_index"][0], periods_numpy_unit["end_sample_index"][0])]
+                assert period == ext_periods_by_unit[segment_index][unit_id]
 
     def test_user_defined_periods_as_arrays(self):
         unit_ids = self.sorting.unit_ids

src/spikeinterface/postprocessing/valid_unit_periods.py

@@ -548,12 +548,12 @@ def _get_data(self, outputs: str = "by_unit"):
             for segment_index in range(self.sorting_analyzer.get_num_segments()):
                 segment_mask = good_periods_array["segment_index"] == segment_index
                 periods_dict = {}
-                for unit_index in unit_ids:
-                    periods_dict[unit_index] = []
+                for unit_index, unit_id in enumerate(unit_ids):
+                    periods_dict[unit_id] = []
                     unit_mask = good_periods_array["unit_index"] == unit_index
                     good_periods_unit_segment = good_periods_array[segment_mask & unit_mask]
                     for start, end in good_periods_unit_segment[["start_sample_index", "end_sample_index"]]:
-                        periods_dict[unit_index].append((start, end))
+                        periods_dict[unit_id].append((start, end))
                 good_periods.append(periods_dict)
 
         return good_periods