Raising meaningful warnings/errors for interpolate_bads, when supplie…

mne/channels/channels.py

@@ -827,6 +827,7 @@ def interpolate_bads(
         origin="auto",
         method=None,
         exclude=(),
+        on_bad_position="warn",
         verbose=None,
     ):
         """Interpolate bad MEG and EEG channels.
@@ -872,6 +873,10 @@ def interpolate_bads(
         exclude : list | tuple
             The channels to exclude from interpolation. If excluded a bad
             channel will stay in bads.
+        on_bad_position : "raise" | "warn" | "ignore"
+            What to do when one or more sensor positions are invalid (zero or NaN).
+            If ``"warn"`` or ``"ignore"``, channels with invalid positions will be
+            filled with :data:`~numpy.nan`.
         %(verbose)s
 
         Returns
@@ -896,6 +901,23 @@ def interpolate_bads(
 
         _check_preload(self, "interpolation")
         _validate_type(method, (dict, str, None), "method")
+
+        invalid_chs = []
+        for ch in self.info["bads"]:
+            loc = self.info["chs"][self.ch_names.index(ch)]["loc"][:3]
+            if np.allclose(loc, 0.0, atol=1e-16) or np.isnan(loc).any():
+                invalid_chs.append(ch)
+
+        if invalid_chs:
+            msg = (
+                f"Channel(s) {invalid_chs} have invalid sensor position(s). "
+                "Interpolation cannot proceed correctly. If you want to continue "
+                "despite missing positions, set on_bad_position='warn' or 'ignore', "
+                "which outputs all NaN values (np.nan) for the interpolated "
+                "channel(s)."
+            )
+            _on_missing(on_bad_position, msg)
+
         method = _handle_default("interpolation_method", method)
         ch_types = self.get_channel_types(unique=True)
         # figure out if we have "mag" for "meg", "hbo" for "fnirs", ... to filter the

mne/channels/tests/test_interpolation.py

@@ -420,6 +420,30 @@ def test_nan_interpolation(raw):
     raw.interpolate_bads(method="nan", reset_bads=False)
     assert raw.info["bads"] == ch_to_interp
 
+    store = raw.info["chs"][1]["loc"]
+    # for on_bad_position="raise"
+    with pytest.raises(ValueError, match="have invalid sensor position"):
+        raw.info["bads"] = ch_to_interp
+        raw.info["chs"][1]["loc"] = np.full(12, np.nan)
+        # DOES NOT interpolates at all. So raw.info["bads"] remains as is
+        raw.interpolate_bads(on_bad_position="raise")
+    assert np.isnan(raw.info["chs"][1]["loc"]).any()
+
+    # for on_bad_position="warn"
+    with pytest.warns(RuntimeWarning, match="have invalid sensor position"):
+        # this DOES the interpolation BUT with a warning
+        # so raw.info["bad"] will be empty again,
+        # and interpolated channel with be all np.nan
+        raw.interpolate_bads(on_bad_position="warn")
+    assert np.isnan(raw.info["chs"][1]["loc"]).any()
+
+    # for on_bad_position="ignore"
+    raw.info["bads"] = ch_to_interp
+    raw.interpolate_bads(on_bad_position="ignore")
+    assert np.isnan(raw.info["chs"][1]["loc"]).any()
+
+    raw.info["chs"][1]["loc"] = store
+
     # make sure other channels are untouched
     raw.drop_channels(ch_to_interp)
     good_chs = raw.get_data()