differentiate between missing column and column with null/NaN values …

…for tissue_position

cellxgene_schema_cli/cellxgene_schema/validate.py

@@ -56,6 +56,7 @@
 ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE = f"{ERROR_SUFFIX_VISIUM} and {ERROR_SUFFIX_IS_SINGLE}"
 ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_FORBIDDEN = f"is only allowed for {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE}"
 ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_REQUIRED = f"is required for {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE}"
+ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_NOTNULL = f"cannot have missing or NaN values when {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE}"
 ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_IN_TISSUE_0 = f"{ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE} and in_tissue is 0"
 
 ERROR_SUFFIX_SPARSE_FORMAT = f"Please ensure it is either a dense array or one of the supported sparse matrix encodings ({','.join(SUPPORTED_SPARSE_MATRIX_TYPES)})"
@@ -1781,22 +1782,20 @@ def _validate_spatial_tissue_position(self, tissue_position_name: str, min: int,
         if not self._is_visium_and_is_single_true():
             return
 
-        # At this point, is_single is True and:
-        # - there's at least one row with Visum, tissue position column is required
-        # - for any Visium row, tissue position is required.
-        if (
-            tissue_position_name not in self.adata.obs
-            or (
-                (
-                    self.adata.obs["assay_ontology_term_id"]
-                    .apply(lambda t: is_ontological_descendant_of(ONTOLOGY_PARSER, t, ASSAY_VISIUM, False))
-                    .astype(bool)
-                )
-                & (self.adata.obs[tissue_position_name].isnull())
-            ).any()
-        ):
+        # visium rows require tissue_position columns
+        if tissue_position_name not in self.adata.obs:
+            # report column is required
             self.errors.append(f"obs['{tissue_position_name}'] {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_REQUIRED}.")
             return
+        elif ((
+                self.adata.obs["assay_ontology_term_id"]
+                .apply(lambda t: is_ontological_descendant_of(ONTOLOGY_PARSER, t, ASSAY_VISIUM, False))
+                .astype(bool)
+            )
+            & (self.adata.obs[tissue_position_name].isnull())).any():
+            # report column has bad values
+            self.errors.append(f"obs['{tissue_position_name}'] {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_NOTNULL}.")
+            return
 
         # Tissue position must be an int.
         obs_tissue_position = self.adata.obs.get(tissue_position_name)

cellxgene_schema_cli/tests/test_validate.py

@@ -17,6 +17,7 @@
     ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_FORBIDDEN,
     ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_IN_TISSUE_0,
     ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_REQUIRED,
+    ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_NOTNULL,
     SPATIAL_HIRES_IMAGE_MAX_DIMENSION_SIZE,
     SPATIAL_HIRES_IMAGE_MAX_DIMENSION_SIZE_VISIUM_11MM,
     Validator,
@@ -1104,6 +1105,17 @@ def test__validate_tissue_position_int_error(self, tissue_position_name):
         assert validator.errors
         assert f"obs['{tissue_position_name}'] must be of int type" in validator.errors[0]
 
+    @pytest.mark.parametrize("tissue_position_name", ["array_col", "array_row", "in_tissue"])
+    def test__validate_tissue_position_nan_error(self, tissue_position_name):
+        validator: Validator = Validator()
+        validator._set_schema_def()
+        validator.adata = adata_visium.copy()
+        validator.adata.obs[tissue_position_name] = np.nan
+
+        # Confirm tissue_position is identified as invalid.
+        validator._check_spatial_obs()
+        assert validator.errors[0] == f"obs['{tissue_position_name}'] {ERROR_SUFFIX_VISIUM_AND_IS_SINGLE_TRUE_NOTNULL}."
+
     @pytest.mark.parametrize("assay_ontology_term_id", ["EFO:0022857", "EFO:0022860", "EFO:0022859"])
     @pytest.mark.parametrize("tissue_position_name, min", [("array_col", 0), ("array_row", 0), ("in_tissue", 0)])
     def test__validate_tissue_position_int_min_error(self, assay_ontology_term_id, tissue_position_name, min):