diff --git a/setup.py b/setup.py
index c545061..e584d54 100644
--- a/setup.py
+++ b/setup.py
@@ -15,7 +15,7 @@
 
 PYMONGO_REQUIRES = ("pymongo>=3.11.0",)
 
-PYMSSQL_REQUIRES = ("cython>=0.29.21", "pymssql==2.1.5")
+PYMSSQL_REQUIRES = ("cython>=0.29.21", "pymssql>=2.1.4")
 
 PSYCOPG2_REQUIRES = ("psycopg2-binary>=2.8.6",)
 
diff --git a/src/dsdk/mssql.py b/src/dsdk/mssql.py
index f198164..9b25536 100644
--- a/src/dsdk/mssql.py
+++ b/src/dsdk/mssql.py
@@ -101,13 +101,6 @@ def connect(self) -> Generator[Any, None, None]:
             con.close()
             logger.info(self.CLOSE)
 
-    @contextmanager
-    def cursor(self, con) -> Generator[Any, None, None]:
-        """Yield cursor that provides dicts."""
-        # Replace return type with ContextManager[Any] when mypy is fixed.
-        with con.cursor(as_dict=True) as cur:
-            yield cur
-
 
 class Mixin(BaseMixin):
     """Mixin."""
diff --git a/src/dsdk/persistor.py b/src/dsdk/persistor.py
index 4ed87b1..5d0bf65 100644
--- a/src/dsdk/persistor.py
+++ b/src/dsdk/persistor.py
@@ -56,8 +56,14 @@ def df_from_query(
         if parameters is None:
             parameters = {}
         cur.execute(query, parameters)
+        columns = (each[0] for each in cur.description)
         rows = cur.fetchall()
-        return DataFrame(rows)
+        if rows:
+            df = DataFrame(rows)
+            df.columns = columns
+        else:
+            df = DataFrame(columns=columns)
+        return df
 
     @classmethod
     def df_from_query_by_ids(  # pylint: disable=too-many-arguments
@@ -72,11 +78,19 @@ def df_from_query_by_ids(  # pylint: disable=too-many-arguments
         if parameters is None:
             parameters = {}
         dfs = []
+        columns = None
+        chunk = None
         for chunk in chunks(ids, size):
             cur.execute(query, {"ids": chunk, **parameters})
             rows = cur.fetchall()
-            dfs.append(DataFrame(rows))
-        return concat(dfs, ignore_index=True)
+            if rows:
+                dfs.append(DataFrame(rows))
+        if chunk is None:
+            raise ValueError("Parameter ids must not be empty")
+        columns = (each[0] for each in cur.description)
+        df = concat(dfs, ignore_index=True)
+        df.columns = columns
+        return df
 
     def __init__(self, sql: Namespace, tables: Tuple[str, ...]):
         """__init__."""
@@ -93,7 +107,7 @@ def check(self, cur, exceptions):
                 logger.info(self.EXTANT, statement)
                 cur.execute(statement)
                 for row in cur:
-                    n = row["n"]
+                    n, *_ = row
                     assert n == 1
                     continue
             except exceptions:
@@ -125,10 +139,11 @@ def connect(self) -> Generator[Any, None, None]:
         raise NotImplementedError()
 
     @contextmanager
-    def cursor(self, con):
+    def cursor(self, con):  # pylint: disable=no-self-use
         """Yield a cursor that provides dicts."""
         # Replace return type with ContextManager[Any] when mypy is fixed.
-        raise NotImplementedError()
+        with con.cursor() as cursor:
+            yield cursor
 
     def extant(self, table: str) -> str:
         """Return extant table sql."""
diff --git a/src/dsdk/postgres.py b/src/dsdk/postgres.py
index 63886f4..66f75f7 100644
--- a/src/dsdk/postgres.py
+++ b/src/dsdk/postgres.py
@@ -29,10 +29,7 @@
         OperationalError,
         connect,
     )
-    from psycopg2.extras import (
-        RealDictCursor,
-        execute_batch,
-    )
+    from psycopg2.extras import execute_batch
     from psycopg2.extensions import (
         register_adapter,
         ISQLQuote,
@@ -138,13 +135,6 @@ def connect(self) -> Generator[Any, None, None]:
             con.close()
             logger.info(self.CLOSE)
 
-    @contextmanager
-    def cursor(self, con) -> Generator[Any, None, None]:
-        """Yield a cursor that provides dicts."""
-        # Replace return type with ContextManager[Any] when mypy is fixed.
-        with con.cursor(cursor_factory=RealDictCursor) as cur:
-            yield cur
-
     @contextmanager
     def open_run(self, parent: Any) -> Generator[Run, None, None]:
         """Open batch."""
@@ -155,15 +145,21 @@ def open_run(self, parent: Any) -> Generator[Run, None, None]:
             cur.execute(sql.schema)
             cur.execute(sql.runs.open, columns)
             for row in cur:
-                run = Run(
-                    row["id"], row["microservice_id"], row["model_id"], parent,
-                )
-                parent.as_of = row["as_of"]
-                duration = row["duration"]
+                (
+                    id_,
+                    microservice_id,
+                    model_id,
+                    duration,
+                    as_of,
+                    time_zone,
+                    *_,
+                ) = row
+                run = Run(id_, microservice_id, model_id, parent,)
+                parent.as_of = as_of
                 parent.duration = Interval(
                     on=duration.lower, end=duration.upper
                 )
-                parent.time_zone = row["time_zone"]
+                parent.time_zone = time_zone
                 break
 
         yield run
@@ -181,7 +177,7 @@ def open_run(self, parent: Any) -> Generator[Run, None, None]:
                 )
             cur.execute(sql.runs.close, {"id": run.id})
             for row in cur:
-                duration = row["duration"]
+                _, _, _, duration, _, _, *_ = row
                 run.duration = Interval(on=duration.lower, end=duration.upper)
                 break
 
diff --git a/src/dsdk/utils.py b/src/dsdk/utils.py
index 1dbdb79..a7bdbe2 100644
--- a/src/dsdk/utils.py
+++ b/src/dsdk/utils.py
@@ -3,6 +3,7 @@
 
 from __future__ import annotations
 
+from contextlib import contextmanager
 from functools import wraps
 from json import dump as json_dump
 from json import load as json_load
@@ -10,8 +11,9 @@
 from pickle import dump as pickle_dump
 from pickle import load as pickle_load
 from sys import stdout
+from time import perf_counter_ns
 from time import sleep as default_sleep
-from typing import Any, Callable, Sequence
+from typing import Any, Callable, Generator, Sequence
 
 logger = getLogger(__name__)
 
@@ -69,6 +71,19 @@ def load_pickle_file(path: str) -> object:
         return pickle_load(fin)
 
 
+@contextmanager
+def profile(key: str) -> Generator[Any, None, None]:
+    """Profile."""
+    # Replace return type with ContextManager[Any] when mypy is fixed.
+    begin = perf_counter_ns()
+    logger.info('{"key": "%s.begin", "ns": "%s"}', key, begin)
+    yield
+    end = perf_counter_ns()
+    logger.info(
+        '{"key": "%s.end", "ns": "%s", "elapsed": "%s"}', key, end, end - begin
+    )
+
+
 def retry(
     exceptions: Sequence[Exception],
     retries: int = 60,
diff --git a/test/test_postgres.py b/test/test_postgres.py
index d1794be..23f7cbc 100644
--- a/test/test_postgres.py
+++ b/test/test_postgres.py
@@ -94,7 +94,8 @@ def test_cursor():
     with persistor.rollback() as cur:
         cur.execute("""select 1 as n""")
         for row in cur.fetchall():
-            assert row["n"] == 1
+            n, *_ = row
+            assert n == 1
 
 
 def test_open_run(