diff --git a/pypath/formats/sqldump.py b/pypath/formats/sqldump.py
index de64a135e..6b8342116 100644
--- a/pypath/formats/sqldump.py
+++ b/pypath/formats/sqldump.py
@@ -23,7 +23,7 @@
 Processing MySQL dump files.
 """
 
-from typing import IO, Iterable
+from typing import Generator, IO, Iterable
 import os
 import collections
 import re
@@ -49,7 +49,7 @@ def tables(
         con_param: dict | None = None,
         source_id: str | tuple | None = None,
         **kwargs
-    ) -> tuple[dict, dict] | dict[str, pd.DataFrame] | sqlite3.Connection:
+    ) -> dict[str, list[tuple] | pd.DataFrame] | sqlite3.Connection:
     """
     From a SQL dump, retrieve tables as data frames or as lists of tuples.
 
@@ -91,17 +91,16 @@ def tables(
     else:
 
         contents = {}
-        headers = {}
 
     for table in tables:
 
-        header, content = table_from_cache(table, source_id)
+        content = table_from_cache(table, source_id)
 
         if not content:
 
             seek(sqldump, table)
-            header, content = one_table(sqldump)
-            table_to_cache(header, content, table, source_id)
+            content = list(one_table(sqldump))
+            table_to_cache(content, table, source_id)
 
             if not content:
 
@@ -109,7 +108,11 @@ def tables(
 
         if return_df or return_sqlite:
 
-            content = pd.DataFrame.from_records(content, columns = header)
+            content = pd.DataFrame.from_records(
+                content,
+                columns = content[0]._fields,
+            )
+
             _log(
                 f'Data frame of table `{table}`: '
                 f'{content.shape[0]} records, {common.df_memory_usage(content)}.'
@@ -130,7 +133,6 @@ def tables(
         else:
 
             contents[table] = content
-            headers[table] = None if return_df else header
 
         tables_done.add(table)
 
@@ -138,13 +140,17 @@ def tables(
 
             break
 
-    return contents if return_df or return_sqlite else (headers, contents)
+    if not return_sqlite and len(contents) == 1:
+
+        contents = common.first(contents.values())
+
+    return contents
 
 
 def table_from_cache(
         table: str,
         source_id: tuple,
-    ) -> tuple[list, list]:
+    ) -> list[tuple]:
     """
     Retrieve table contents from cache.
     """
@@ -161,14 +167,18 @@ def table_from_cache(
                 f'Cache path: `{cache_path}`.'
             )
 
-            return pickle.load(fp)
+            name, fields, content = pickle.load(fp)
 
-    return [], []
+            record = collections.namedtuple(name, fields)
+            content = [record(*r) for r in content]
+
+            return content
+
+    return []
 
 
 def table_to_cache(
-        header: dict,
-        content: dict,
+        content: list[tuple],
         table: str,
         source_id: tuple,
     ) -> None:
@@ -182,30 +192,37 @@ def table_to_cache(
     with open(cache_path, 'wb') as fp:
 
         _log(f'Saving table `{table}` to cache. Cache path: `{cache_path}`.')
-        pickle.dump((header, content), fp)
+        pickle.dump(
+            (
+                content[0].__class__.__name__,
+                content[0]._fields,
+                [tuple(r) for r in content]
+            ),
+            fp,
+        )
 
 
-def one_table(sqldump: IO) -> tuple[list[str], list[tuple]]:
+def one_table(sqldump: IO) -> Generator[tuple]:
     """
     Retrieve one table from a MySQL dump.
     """
 
     parser = sqlparse.parsestream(sqldump)
-    header = []
-    content = []
-    inserts = 0
+    n_inserts = 0
+    n_records = 0
+    record = None
     table_name = the_table_name = None
 
     for statement in parser:
 
         if statement.get_type() == 'INSERT':
 
-            inserts += 1
+            n_inserts += 1
             sublists = statement.get_sublists()
             table_info = next(sublists)
             table_name = table_info.get_name()
 
-            if inserts == 1:
+            if n_inserts == 1:
 
                 the_table_name = table_name
                 _log(f'Processing table: `{table_name}`')
@@ -214,28 +231,30 @@ def one_table(sqldump: IO) -> tuple[list[str], list[tuple]]:
 
                 break
 
-            header = [
-                col.get_name()
-                for col in table_info.get_parameters()
-            ]
+            if record is None:
 
-            content.extend(
-                tuple(
+                record = collections.namedtuple(
+                    f'Ramp{table_name.capitalize()}',
+                    [col.get_name() for col in table_info.get_parameters()],
+                )
+                globals()[record.__class__.__name__] = record
+
+            for rec in next(sublists).get_sublists():
+
+                n_records += 1
+
+                yield record(*(
                     s.value.strip('"\'')
                     for s in next(rec.get_sublists()).get_identifiers()
-                )
-                for rec in next(sublists).get_sublists()
-            )
+                ))
 
     if table_name:
 
         _log(
-            f'Processed {len(content)} records in {inserts} INSERT statements '
+            f'Processed {n_records} records in {n_inserts} INSERT statements '
             f'from table `{the_table_name}`.'
         )
 
-    return header, content
-
 
 def get_source_id(
         sqldump: str | IO,