Add export_options to export_to_file operator

The current implementation of the export_to_file operator doesn't provide any mechanism for overriding or customizing the behavior of the file output. Instead the operator simply calls one of a number of Pandas.DataFrame.to_* functions with default functions that SDK users are locked in to.

This modification allows for the provision of a configuration dictionary that enables the customizing or overriding of the file write behavior by passing the parameters to the FileType.create_from_dataframe implementation.

python-sdk/src/astro/files/base.py

@@ -114,18 +114,19 @@ def is_pattern(self) -> bool:
         """
         return not pathlib.PosixPath(self.path).suffix
 
-    def create_from_dataframe(self, df: pd.DataFrame, store_as_dataframe: bool = True) -> None:
+    def create_from_dataframe(self, df: pd.DataFrame, store_as_dataframe: bool = True, export_options: dict | None = None) -> None:
         """Create a file in the desired location using the values of a dataframe.
 
         :param store_as_dataframe: Whether the data should later be deserialized as a dataframe or as a file containing
             delimited data (e.g. csv, parquet, etc.).
         :param df: pandas dataframe
+        :param export_options: additional arguments to pass to the underlying write functionality
         """
 
         self.is_dataframe = store_as_dataframe
 
         with self.location.get_stream() as stream:
-            self.type.create_from_dataframe(stream=stream, df=df)
+            self.type.create_from_dataframe(stream=stream, df=df, **export_options)
 
     @property
     def openlineage_dataset_namespace(self) -> str:

python-sdk/src/astro/files/types/base.py

@@ -27,11 +27,12 @@ def export_to_dataframe(self, stream, **kwargs) -> pd.DataFrame:
         raise NotImplementedError
 
     @abstractmethod
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:
         """Write file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the underlying write functionality
         """
         raise NotImplementedError
 

python-sdk/src/astro/files/types/csv.py

@@ -38,13 +38,15 @@ def export_to_dataframe(
         return PandasDataframe.from_pandas_df(df)
 
     # We need skipcq because it's a method overloading so we don't want to make it a static method
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:  # skipcq PYL-R0201
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:  # skipcq PYL-R0201
         """Write csv file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the pandas `to_csv` function
         """
-        df.to_csv(stream, index=False)
+
+        df.to_csv(stream, **dict(index=False, **kwargs))
 
     @property
     def name(self):

python-sdk/src/astro/files/types/excel.py

@@ -37,10 +37,11 @@ def export_to_dataframe(
         return PandasDataframe.from_pandas_df(df)
 
     # We need skipcq because it's a method overloading so we don't want to make it a static method
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:  # skipcq PYL-R0201
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:  # skipcq PYL-R0201
         """Write Excel file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the pandas `to_excel` function
         """
-        df.to_excel(stream, index=False)
+        df.to_excel(stream, **dict(index=False, **kwargs))

python-sdk/src/astro/files/types/json.py

@@ -42,13 +42,14 @@ def export_to_dataframe(
         return PandasDataframe.from_pandas_df(df)
 
     # We need skipcq because it's a method overloading so we don't want to make it a static method
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:  # skipcq PYL-R0201
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:  # skipcq PYL-R0201
         """Write json file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the pandas `to_json` function
         """
-        df.to_json(stream, orient="records")
+        df.to_json(stream, **dict(orient="records", **kwargs))
 
     @property
     def name(self):

python-sdk/src/astro/files/types/ndjson.py

@@ -39,13 +39,14 @@ def export_to_dataframe(
         return PandasDataframe.from_pandas_df(df)
 
     # We need skipcq because it's a method overloading so we don't want to make it a static method
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:  # skipcq PYL-R0201
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:  # skipcq PYL-R0201
         """Write ndjson file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the pandas `to_json` function
         """
-        df.to_json(stream, orient="records", lines=True)
+        df.to_json(stream, **dict(orient="records", lines=True, **kwargs))
 
     @property
     def name(self):

python-sdk/src/astro/files/types/parquet.py

@@ -57,13 +57,14 @@ def _convert_remote_file_to_byte_stream(stream) -> io.IOBase:
         return remote_obj_buffer
 
     # We need skipcq because it's a method overloading so we don't want to make it a static method
-    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper) -> None:  # skipcq PYL-R0201
+    def create_from_dataframe(self, df: pd.DataFrame, stream: io.TextIOWrapper, **kwargs) -> None:  # skipcq PYL-R0201
         """Write parquet file to one of the supported locations
 
         :param df: pandas dataframe
         :param stream: file stream object
+        :param kwargs: additional arguments to pass to the pandas `to_parquet` method
         """
-        df.to_parquet(stream)
+        df.to_parquet(stream, **kwargs)
 
     @property
     def name(self):

python-sdk/src/astro/sql/operators/export_to_file.py

@@ -21,20 +21,23 @@ class ExportToFileOperator(AstroSQLBaseOperator):
     :param input_data: Table to convert to file
     :param output_file: File object containing the path to the file and connection id.
     :param if_exists: Overwrite file if exists. Default False.
+    :param export_options: Additional options to pass to the file export functions.
     """
 
-    template_fields = ("input_data", "output_file")
+    template_fields = ("input_data", "output_file", "export_options")
 
     def __init__(
         self,
         input_data: BaseTable | pd.DataFrame,
         output_file: File,
         if_exists: ExportExistsStrategy = "exception",
+        export_options: dict | None = None,
         **kwargs,
     ) -> None:
         self.output_file = output_file
         self.input_data = input_data
         self.if_exists = if_exists
+        self.export_options = export_options or {}
         self.kwargs = kwargs
         datasets = {"output_datasets": self.output_file}
         if isinstance(input_data, Table):
@@ -57,7 +60,7 @@ def execute(self, context: Context) -> File:  # skipcq PYL-W0613
             raise ValueError(f"Expected input_table to be Table or dataframe. Got {type(self.input_data)}")
         # Write file if overwrite == True or if file doesn't exist.
         if self.if_exists == "replace" or not self.output_file.exists():
-            self.output_file.create_from_dataframe(df, store_as_dataframe=False)
+            self.output_file.create_from_dataframe(df, store_as_dataframe=False, export_options=self.export_options)
             return self.output_file
         else:
             raise FileExistsError(f"{self.output_file.path} file already exists.")
@@ -144,7 +147,8 @@ def export_to_file(
     output_file: File,
     if_exists: ExportExistsStrategy = "exception",
     task_id: str | None = None,
-    **kwargs: Any,
+    export_options: dict | None = None,
+    **kwargs,
 ) -> XComArg:
     """Convert ExportToFileOperator into a function. Returns XComArg.
 
@@ -170,6 +174,7 @@ def export_to_file(
     :param input_data: Input table / dataframe
     :param if_exists: Overwrite file if exists. Default "exception"
     :param task_id: task id, optional
+    :param export_options: Additional options to pass to the file export functions.
     """
 
     task_id = task_id or get_unique_task_id("export_to_file")
@@ -179,5 +184,6 @@ def export_to_file(
         output_file=output_file,
         input_data=input_data,
         if_exists=if_exists,
+        export_options=export_options,
         **kwargs,
     ).output

python-sdk/tests/sql/operators/test_export_file.py

@@ -37,6 +37,25 @@ def make_df():
     assert df.equals(pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4]}))
 
 
+def test_save_dataframe_to_local_with_options(sample_dag):
+    @aql.dataframe
+    def make_df():
+        return pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4]})
+
+    with sample_dag:
+        df = make_df()
+        aql.export_to_file(
+            input_data=df,
+            output_file=File(path="/tmp/saved_df.csv"),
+            if_exists="replace",
+            export_options={"header": None},
+        )
+    test_utils.run_dag(sample_dag)
+
+    df = pd.read_csv("/tmp/saved_df.csv")
+    assert df.equals(pd.DataFrame(data={"0": [1, 2], "1": [3, 4]}))
+
+
 @pytest.mark.parametrize("database_table_fixture", [{"database": Database.SQLITE}], indirect=True)
 def test_save_temp_table_to_local(sample_dag, database_table_fixture):
     _, test_table = database_table_fixture