fix(schema 5.1.0): throw error if obsm is not type np.ndarray (#859)

Co-authored-by: nayib-jose-gloria <[email protected]>
Co-authored-by: Nayib Gloria <[email protected]>

cellxgene_schema_cli/cellxgene_schema/validate.py

@@ -1003,6 +1003,7 @@ def _validate_obsm(self):
 
             regex_pattern = r"^[a-zA-Z][a-zA-Z0-9_.-]*$"
 
+            unknown_key = False  # an unknown key does not match 'spatial' or 'X_{suffix}'
             if key.startswith("X_"):
                 obsm_with_x_prefix += 1
                 if key.lower() == "x_spatial":
@@ -1021,19 +1022,35 @@ def _validate_obsm(self):
                     f"not be available in Explorer"
                 )
                 issue_list = self.warnings
+                unknown_key = True
 
             if not isinstance(value, np.ndarray):
-                issue_list.append(
+                self.errors.append(
                     f"All embeddings have to be of 'numpy.ndarray' type, " f"'adata.obsm['{key}']' is {type(value)}')."
                 )
                 # Skip over the subsequent checks that require the value to be an array
                 continue
 
-            if len(value.shape) < 2 or value.shape[0] != self.adata.n_obs or value.shape[1] < 2:
-                issue_list.append(
-                    f"All embeddings must have as many rows as cells, and at least two columns."
-                    f" 'adata.obsm['{key}']' has shape of '{value.shape}'."
+            if len(value.shape) < 2:
+                self.errors.append(
+                    f"All embeddings must at least two dimensions. 'adata.obsm['{key}']' has a shape length of '{len(value.shape)}'."
                 )
+            else:
+                if value.shape[0] != self.adata.n_obs:
+                    self.errors.append(
+                        f"All embeddings must have as many rows as cells. 'adata.obsm['{key}']' has rows='{value.shape[0]}'."
+                    )
+
+                if unknown_key and value.shape[1] < 1:
+                    self.errors.append(
+                        f"All unspecified embeddings must have at least one column. 'adata.obsm['{key}']' has columns='{value.shape[1]}'."
+                    )
+
+                if not unknown_key and value.shape[1] < 2:
+                    self.errors.append(
+                        f"All 'X_' and 'spatial' embeddings must have at least two columns. 'adata.obsm['{key}']' has columns='{value.shape[1]}'."
+                    )
+
             if not (np.issubdtype(value.dtype, np.integer) or np.issubdtype(value.dtype, np.floating)):
                 issue_list.append(
                     f"adata.obsm['{key}'] has an invalid data type. It should be "

cellxgene_schema_cli/tests/test_schema_compliance.py

@@ -2057,7 +2057,10 @@ def test_obsm_values_warn_start_with_X(self, validator_with_adata):
             "WARNING: Dataframe 'var' only has 4 rows. Features SHOULD NOT be filtered from expression matrix.",
             "WARNING: Embedding key in 'adata.obsm' harmony is not 'spatial' nor does it start with 'X_'. "
             "Thus, it will not be available in Explorer",
-            "WARNING: All embeddings have to be of 'numpy.ndarray' type, 'adata.obsm['harmony']' is <class 'pandas.core.frame.DataFrame'>').",
+            "WARNING: Validation of raw layer was not performed due to current errors, try again after fixing current errors.",
+        ]
+        assert validator.errors == [
+            "ERROR: All embeddings have to be of 'numpy.ndarray' type, 'adata.obsm['harmony']' is <class 'pandas.core.frame.DataFrame'>')."
         ]
 
     def test_obsm_values_suffix_is_forbidden(self, validator_with_adata):
@@ -2080,13 +2083,14 @@ def test_obsm_values_key_start_with_number(self, validator_with_adata):
         validator.adata.obsm["3D"] = pd.DataFrame(validator.adata.obsm["X_umap"], index=validator.adata.obs_names)
         validator.validate_adata()
         assert validator.errors == [
-            "ERROR: Embedding key in 'adata.obsm' 3D does not match the regex pattern ^[a-zA-Z][a-zA-Z0-9_.-]*$."
+            "ERROR: Embedding key in 'adata.obsm' 3D does not match the regex pattern ^[a-zA-Z][a-zA-Z0-9_.-]*$.",
+            "ERROR: All embeddings have to be of 'numpy.ndarray' type, 'adata.obsm['3D']' is <class "
+            "'pandas.core.frame.DataFrame'>').",
         ]
         assert validator.warnings == [
             "WARNING: Dataframe 'var' only has 4 rows. Features SHOULD NOT be filtered from expression matrix.",
             "WARNING: Embedding key in 'adata.obsm' 3D is not 'spatial' nor does it start with 'X_'. "
             "Thus, it will not be available in Explorer",
-            "WARNING: All embeddings have to be of 'numpy.ndarray' type, 'adata.obsm['3D']' is <class 'pandas.core.frame.DataFrame'>').",
             "WARNING: Validation of raw layer was not performed due to current errors, try again after fixing current errors.",
         ]
 
@@ -2136,9 +2140,23 @@ def test_obsm_shape_one_column(self, validator_with_visium_assay, key):
         validator.adata.obsm[key] = numpy.delete(validator.adata.obsm[key], 0, 1)
         validator.validate_adata()
         assert validator.errors == [
-            "ERROR: All embeddings must have as many rows as cells, and "
-            f"at least two columns. 'adata.obsm['{key}']' has shape "
-            "of '(2, 1)'."
+            "ERROR: All 'X_' and 'spatial' embeddings must have at least two columns. "
+            f"'adata.obsm['{key}']' has columns='1'."
+        ]
+
+    def test_obsm_shape_zero_column_with_unknown_key(self, validator_with_adata):
+        """
+        embeddings that are not 'X_' or 'spatial' that are ndarrays must have at least one column
+        """
+        # Makes 0 column array
+        validator = validator_with_adata
+        n_obs = validator_with_adata.adata.n_obs
+        validator.adata.obsm["unknown"] = numpy.zeros((n_obs, 0))
+        validator.validate_adata()
+        assert validator.errors == [
+            "ERROR: The size of the ndarray stored for a 'adata.obsm['unknown']' MUST NOT " "be zero.",
+            "ERROR: All unspecified embeddings must have at least one column. "
+            "'adata.obsm['unknown']' has columns='0'.",
         ]
 
     def test_obsm_shape_same_rows_and_columns(self, validator_with_adata):
@@ -2166,7 +2184,9 @@ def test_obsm_size_zero(self, validator_with_adata):
         validator.adata = save_and_read_adata(adata)
         validator.validate_adata()
         assert validator.errors == [
-            "ERROR: The size of the ndarray stored for a 'adata.obsm['badsize']' MUST NOT be zero.",
+            "ERROR: The size of the ndarray stored for a 'adata.obsm['badsize']' MUST NOT " "be zero.",
+            "ERROR: All unspecified embeddings must have at least one column. "
+            "'adata.obsm['badsize']' has columns='0'.",
         ]