feat: handle different schema for GetTableSchema/ExecuteSchema

adbc_drivers_validation/model.py

@@ -406,6 +406,10 @@ class SelectQuery:
     bind_schema_path: Path | None = None
     #: Data for the bind parameters.
     bind_path: Path | None = None
+    #: Schema of the result set (when not executing a query).  For some
+    #: databases and some situations, this is different than the schema you get
+    #: when you actually execute the query.
+    catalog_schema_path: Path | None = None
 
     def setup_query(self) -> str | None:
         if not self.setup_query_path:
@@ -434,6 +438,11 @@ def query(self) -> str:
     def expected_schema(self) -> pyarrow.Schema:
         return try_txtcase(self.expected_schema_path, query_schema, ["expected_schema"])
 
+    def catalog_schema(self) -> pyarrow.Schema:
+        if not self.catalog_schema_path:
+            return self.expected_schema()
+        return try_txtcase(self.catalog_schema_path, query_schema, ["catalog_schema"])
+
     def expected_result(self) -> pyarrow.Table:
         return try_txtcase(
             self.expected_path, query_table, ["expected"], self.expected_schema()
@@ -548,6 +557,8 @@ def merge(
                         "expected_schema_path": parent.query.expected_schema_path,
                         "expected_path": parent.query.expected_path,
                     }
+                    if parent.query.catalog_schema_path:
+                        params["catalog_schema_path"] = parent.query.catalog_schema_path
                     if parent.query.setup_query_path:
                         params["setup_query_path"] = parent.query.setup_query_path
                     # TODO: we also want to test with ExecuteQuery so perhaps
@@ -588,6 +599,7 @@ def merge(
             in {
                 "query_path",
                 "expected_schema_path",
+                "catalog_schema_path",
                 "expected_path",
                 "setup_query_path",
                 "bind_query_path",

adbc_drivers_validation/tests/query.py

@@ -123,9 +123,8 @@ def test_execute_schema(
     ) -> None:
         subquery = query.query
         assert isinstance(subquery, model.SelectQuery)
-
         sql = subquery.query()
-        expected_schema = subquery.expected_schema()
+        expected_schema = subquery.catalog_schema()
 
         _setup_query(driver, conn, query)
 
@@ -142,16 +141,16 @@ def test_get_table_schema(
         query: model.Query,
     ) -> None:
         subquery = query.query
-
-        expected_schema = subquery.expected_schema()
+        assert isinstance(subquery, model.SelectQuery)
+        expected_schema = subquery.catalog_schema()
 
         with setup_connection(query, conn):
             _setup_query(driver, conn, query)
 
             table_name = None
             md = query.metadata()
             table_name = md.setup.drop
-            if not table_name and isinstance(subquery, model.SelectQuery):
+            if not table_name:
                 # XXX: rather hacky, but extract the table name from the SELECT query
                 # that would normally be executed
                 query_str = subquery.query().split()

adbc_drivers_validation/txtcase.py

@@ -92,7 +92,7 @@ def get_part(self, part: str, schema: pyarrow.Schema | None = None) -> typing.An
 
         if part == "metadata":
             return tomllib.loads(value)
-        if part in {"bind_schema", "expected_schema", "input_schema"}:
+        if part in {"bind_schema", "expected_schema", "catalog_schema", "input_schema"}:
             return arrowjson.loads_schema(value)
         elif part in {"bind_query", "query", "setup_query"}:
             return value