Extra checks to handle potentially bad ssx data

src/dials/algorithms/scaling/outlier_rejection.py

@@ -80,7 +80,10 @@ def reject_outliers(reflection_table, experiment, method="standard", zmax=6.0):
 def determine_Esq_outlier_index_arrays(Ih_table, experiment, emax=10.0):
     # first calculate normalised intensities and set in the Ih_table.
     intensities = Ih_table.as_miller_array(experiment.crystal.get_unit_cell())
-    normalised_intensities = quasi_normalisation(intensities)
+    try:
+        normalised_intensities = quasi_normalisation(intensities)
+    except AssertionError:
+        return [np.array([], dtype=np.uint64)] * Ih_table.n_datasets
 
     sel = normalised_intensities.data() > (emax**2)
     n_e2_outliers = sel.count(True)

src/dials/algorithms/symmetry/__init__.py

@@ -153,7 +153,8 @@ def _normalise(self, method):
             normalise = self.ml_iso_normalisation
         elif method == "ml_aniso":
             normalise = self.ml_aniso_normalisation
-
+        bad_datasets = []
+        normalised_intensities = None
         for i in range(int(flex.max(self.dataset_ids) + 1)):
             logger.info("\n" + "-" * 80 + "\n")
             logger.info("Normalising intensities for dataset %i\n", i + 1)
@@ -171,12 +172,22 @@ def _normalise(self, method):
                     method,
                     exc_info=True,
                 )
-                return
-            if i == 0:
-                normalised_intensities = intensities
+                bad_datasets.append(i)
             else:
-                normalised_intensities = normalised_intensities.concatenate(intensities)
+                if not normalised_intensities:
+                    normalised_intensities = intensities
+                else:
+                    normalised_intensities = normalised_intensities.concatenate(
+                        intensities
+                    )
 
+        if bad_datasets:
+            sel = flex.bool(self.dataset_ids.size(), True)
+            for i in bad_datasets:
+                bad_sel = self.dataset_ids == i
+                sel.set_selected(bad_sel, False)
+            self.dataset_ids = self.dataset_ids.select(sel)
+        assert self.dataset_ids.size() == normalised_intensities.size()
         self.intensities = normalised_intensities.set_info(
             self.intensities.info()
         ).set_observation_type_xray_intensity()
@@ -243,11 +254,15 @@ def _ml_normalisation(intensities, aniso):
             normalisation = absolute_scaling.ml_aniso_absolute_scaling(
                 intensities, n_residues=n_residues
             )
+            if not normalisation.p_scale:
+                raise RuntimeError("Unsuccessful normalisation")
             u_star = normalisation.u_star
         else:
             normalisation = absolute_scaling.ml_iso_absolute_scaling(
                 intensities, n_residues=n_residues
             )
+            if not (normalisation.b_wilson and normalisation.p_scale):
+                raise RuntimeError("Unsuccessful normalisation")
             u_star = adptbx.b_as_u(
                 adptbx.u_iso_as_u_star(intensities.unit_cell(), normalisation.b_wilson)
             )