Fix: Force index type to str (#28)

Author: adrien-berchet

data_validation_framework/task.py

@@ -295,10 +295,14 @@ def kwargs(self):
     def read_dataset(self):
         """Import the dataset to a :class:`pandas.DataFrame`.
 
+        Note that the index column is loaded as a string.
+
         This method can be overridden to load custom data (e.g. GeoDataFrame, etc.).
         The dataset should always be loaded from the path given by `self.dataset_df`.
         """
-        return pd.read_csv(self.dataset_df, index_col=self.input_index_col)
+        return pd.read_csv(
+            self.dataset_df, index_col=self.input_index_col, dtype={self.input_index_col: str}
+        )
 
     def pre_process(self, df, args, kwargs):
         """Method executed before applying the external function."""

tests/test_task.py

@@ -1274,7 +1274,9 @@ def check_exception(failed_task, exception):  # pylint: disable=unused-variable
         assert not luigi.build([failing_task], local_scheduler=True)
 
         assert failed_tasks == [str(failing_task)]
-        assert exceptions == [str(IndexError("The following index values are duplicated: [0, 1]"))]
+        assert exceptions == [
+            str(IndexError("The following index values are duplicated: ['0', '1']"))
+        ]
 
     def test_change_index(self, tmpdir, TestTask):
         dataset_df_path = str(tmpdir / "dataset.csv")