Fix dtype of quality metrics before and after merging

src/spikeinterface/qualitymetrics/quality_metric_calculator.py

@@ -125,7 +125,13 @@ def _merge_extension_data(
         all_unit_ids = new_sorting_analyzer.unit_ids
         not_new_ids = all_unit_ids[~np.isin(all_unit_ids, new_unit_ids)]
 
+        # this creates a new metrics dictionary, but the dtype for everything will be
+        # object
         metrics = pd.DataFrame(index=all_unit_ids, columns=old_metrics.columns)
+        # we can iterate through the columns and convert them back to numbers with
+        # pandas.to_numeric. coerce allows us to keep the nan values.
+        for column in metrics.columns:
+            metrics[column] = pd.to_numeric(metrics[column], errors="coerce")
 
         metrics.loc[not_new_ids, :] = old_metrics.loc[not_new_ids, :]
         metrics.loc[new_unit_ids, :] = self._compute_metrics(

src/spikeinterface/qualitymetrics/tests/test_quality_metric_calculator.py

@@ -48,6 +48,34 @@ def test_compute_quality_metrics(sorting_analyzer_simple):
     assert "isolation_distance" in metrics.columns
 
 
+def test_merging_quality_metrics(sorting_analyzer_simple):
+
+    sorting_analyzer = sorting_analyzer_simple
+
+    metrics = compute_quality_metrics(
+        sorting_analyzer,
+        metric_names=None,
+        qm_params=dict(isi_violation=dict(isi_threshold_ms=2)),
+        skip_pc_metrics=False,
+        seed=2205,
+    )
+
+    # sorting_analyzer_simple has ten units
+    new_sorting_analyzer = sorting_analyzer.merge_units([[0, 1]])
+
+    new_metrics = new_sorting_analyzer.get_extension("quality_metrics").get_data()
+
+    # we should copy over the metrics after merge
+    for column in metrics.columns:
+        assert column in new_metrics.columns
+
+    # 10 units vs 9 units
+    assert len(metrics.index) > len(new_metrics.index)
+
+    # dtype should be fine after merge but is cast from Float64->float64
+    assert np.float64 == new_metrics["snr"].dtype
+
+
 def test_compute_quality_metrics_recordingless(sorting_analyzer_simple):
 
     sorting_analyzer = sorting_analyzer_simple