fix: enforce strict validation for terms-of-use metadata (#766)

Author: adilraza99

malariagen_data/anoph/base.py

@@ -564,7 +564,10 @@ def _sample_set_has_unrestricted_use(self, *, sample_set: str):
         )
 
         if "unrestricted_use" not in release_manifest_df.columns:
-            return False
+            raise ValueError(
+                f"Column 'unrestricted_use' missing from manifest for sample set '{sample_set}'. "
+                "This indicates a data integrity issue in the release manifest."
+            )
 
         sample_set_records_srs = release_manifest_df.loc[
             release_manifest_df["sample_set"] == sample_set, "unrestricted_use"
@@ -833,22 +836,15 @@ def lookup_terms_of_use_info(self, sample_set: base_params.sample_set) -> dict:
                 "terms_of_use_url",
                 "unrestricted_use",
             ]
-            placeholder_values = {
-                "terms_of_use_expiry_date": "2099-12-31",
-                "terms_of_use_url": float("nan"),
-                "unrestricted_use": False,
-            }
-            available_cols = [c for c in expected_cols if c in df_sample_sets.columns]
-            if available_cols:
-                lookup = df_sample_sets[available_cols].to_dict(orient="index")
-                missing_cols = set(expected_cols) - set(available_cols)
-                if missing_cols:
-                    for ss in lookup:
-                        for mc in missing_cols:
-                            lookup[ss][mc] = placeholder_values[mc]
-            else:
-                lookup = {ss: dict(placeholder_values) for ss in df_sample_sets.index}
-            self._cache_sample_set_to_terms_of_use_info = lookup
+            missing_cols = [c for c in expected_cols if c not in df_sample_sets.columns]
+            if missing_cols:
+                raise ValueError(
+                    f"Terms-of-use columns missing from manifest: {missing_cols}. "
+                    "This indicates a data integrity issue in the release manifest."
+                )
+            self._cache_sample_set_to_terms_of_use_info = df_sample_sets[
+                expected_cols
+            ].to_dict(orient="index")
         try:
             return self._cache_sample_set_to_terms_of_use_info[sample_set]
         except KeyError as e:

tests/anoph/test_base.py

@@ -293,14 +293,8 @@ def test_lookup_terms_of_use_info_missing_columns(ag3_sim_fixture):
         )
 
         sample_set = "1177-VO-ML-LEHMANN-VMF00004"
-        info = api.lookup_terms_of_use_info(sample_set)
-        assert isinstance(info, dict)
-        assert "terms_of_use_expiry_date" in info
-        assert "terms_of_use_url" in info
-        assert "unrestricted_use" in info
-        assert info["terms_of_use_expiry_date"] == "2099-12-31"
-        assert pd.isna(info["terms_of_use_url"])
-        assert info["unrestricted_use"] is False
+        with pytest.raises(ValueError, match="Terms-of-use columns missing"):
+            api.lookup_terms_of_use_info(sample_set)
     finally:
         for mp, bp in zip(manifest_paths, backups):
             shutil.move(bp, mp)
@@ -333,8 +327,8 @@ def test_sample_set_has_unrestricted_use_missing_column(ag3_sim_fixture):
         )
 
         sample_set = "1177-VO-ML-LEHMANN-VMF00004"
-        result = api._sample_set_has_unrestricted_use(sample_set=sample_set)
-        assert result is False
+        with pytest.raises(ValueError, match="unrestricted_use.*missing"):
+            api._sample_set_has_unrestricted_use(sample_set=sample_set)
     finally:
         for mp, bp in zip(manifest_paths, backups):
             shutil.move(bp, mp)