table_checksum uses "SELECT SUM(hashtext(t::text)) ..." (#337)

It fixes a problem with large tables

ATTENTION: new checksum algorithm is used!

Author: dmitry-lipetsk

src/node.py

@@ -2236,18 +2236,16 @@ def _table_checksum__use_cn(
         assert cursor is not None
 
         try:
-            cursor.execute("SELECT t::text FROM {} as t".format(
+            cursor.execute("SELECT SUM(hashtext(t::text)) FROM {} as t".format(
                 __class__._delim_sql_ident(table)
             ))
 
-            while True:
-                row = cursor.fetchone()
-                if row is None:
-                    break
-                assert type(row) in [list, tuple]  # noqa: E721
-                assert len(row) == 1
-                sum += hash(row[0])
-                continue
+            row = cursor.fetchone()
+            assert row is not None
+            assert type(row) in [list, tuple]  # noqa: E721
+            assert len(row) == 1
+            v = row[0]
+            sum += int(v if v is not None else 0)
         finally:
             cursor.close()
 

tests/test_testgres_common.py

@@ -2102,7 +2102,7 @@ def test_node__table_checksum(
 
                 with cn.connection.cursor() as cursor:
                     assert cursor is not None
-                    cursor.execute("SELECT t::text FROM \"t\" as t;")
+                    cursor.execute("SELECT hashtext(t::text) FROM \"t\" as t;")
 
                     checksum1 = 0
                     record_count = 0
@@ -2113,7 +2113,7 @@ def test_node__table_checksum(
                         assert type(row) in [list, tuple]  # noqa: E721
                         assert len(row) == 1
                         record_count += 1
-                        checksum1 += hash(row[0])
+                        checksum1 += int(row[0])
                         pass
 
                     assert record_count == table_checksum_test_data.record_count
@@ -2162,7 +2162,7 @@ def test_node__pgbench_table_checksums__one_table(
 
                 with cn.connection.cursor() as cursor:
                     assert cursor is not None
-                    cursor.execute("SELECT t::text FROM \"t\" as t;")
+                    cursor.execute("SELECT hashtext(t::text) FROM \"t\" as t;")
 
                     checksum1 = 0
                     record_count = 0
@@ -2173,7 +2173,7 @@ def test_node__pgbench_table_checksums__one_table(
                         assert type(row) in [list, tuple]  # noqa: E721
                         assert len(row) == 1
                         record_count += 1
-                        checksum1 += hash(row[0])
+                        checksum1 += int(row[0])
                         pass
 
                     assert record_count == table_checksum_test_data.record_count