Fix dataset corruption after updates (JDASoftwareGroup#445)

When updating a dataset with a table name other than 'table', an additional table named
'table' is erroneously created. This corrupts the dataset. The issue was introduced after
deprecating the table name feature in the 4.0.0 release. The root cause is not passing the
table name as an argument within `partition_on` and `add_metapartition`, which leads to the
default table name "table" being used.

CHANGES.rst

@@ -2,6 +2,12 @@
 Changelog
 =========
 
+Kartothek 4.0.1 (2021-04-XX)
+============================
+
+* Fixed dataset corruption after updates when table names other than "table" are used (#445).
+
+
 Kartothek 4.0.0 (2021-03-17)
 ============================
 

kartothek/io_components/metapartition.py

@@ -471,6 +471,7 @@ def add_metapartition(
             schema=schema,
             partition_keys=metapartition.partition_keys or None,
             logical_conjunction=metapartition.logical_conjunction or None,
+            table_name=metapartition.table_name,
         )
 
         # Add metapartition information to the new object
@@ -1109,6 +1110,7 @@ def partition_on(self, partition_on: Union[str, Sequence[str]]):
                     f"{label}"
                 ),
                 partition_keys=partition_on,
+                table_name=self.table_name,
             )
             new_mp = new_mp.add_metapartition(tmp_mp, schema_validation=False)
         if self.indices:

tests/io_components/test_metapartition.py

@@ -1331,3 +1331,13 @@ def test_get_parquet_metadata_row_group_size(store):
         }
     )
     pd.testing.assert_frame_equal(actual, expected)
+
+
+def test_partition_on_keeps_table_name():
+    mp = MetaPartition(
+        label="label_1",
+        data=pd.DataFrame({"P": [1, 2, 1, 2], "L": [1, 1, 2, 2]}),
+        table_name="non-default-name",
+    )
+    repartitioned_mp = mp.partition_on(["P"])
+    assert repartitioned_mp.table_name == "non-default-name"