[DOP-21976] Add compression options for XML

Author: Ilyas Gasanov

docs/changelog/next_release/161.feature.rst

@@ -0,0 +1 @@
+Add compression options to XML file format

syncmaster/dto/transfers.py

@@ -36,15 +36,20 @@ class FileTransferDTO(TransferDTO):
 
     def __post_init__(self):
         if isinstance(self.file_format, dict):
-            self.file_format = self._get_format(self.file_format.copy())
+            self.file_format = self._get_file_format(self.file_format.copy())
         if isinstance(self.df_schema, str):
             self.df_schema = json.loads(self.df_schema)
 
-    def _get_format(self, file_format: dict):
+    def _get_file_format(self, file_format: dict) -> CSV | JSONLine | JSON | Excel | XML | ORC | Parquet:
         file_type = file_format.pop("type", None)
+        # XML at spark-xml has no "none" option https://github.com/databricks/spark-xml?tab=readme-ov-file#features
+        if file_type == "xml" and file_format.get("compression") == "none":
+            file_format.pop("compression")
+
         parser_class = self._format_parsers.get(file_type)
         if parser_class is not None:
             return parser_class.parse_obj(file_format)
+
         raise ValueError(f"Unknown file type: {file_type}")
 
 

syncmaster/schemas/v1/transfers/file_format.py

@@ -49,6 +49,14 @@ class CSVCompression(str, Enum):
     DEFLATE = "deflate"
 
 
+class XMLCompression(str, Enum):
+    NONE = "none"
+    BZIP2 = "bzip2"
+    GZIP = "gzip"
+    LZ4 = "lz4"
+    SNAPPY = "snappy"
+
+
 class CSV(BaseModel):
     type: CSV_FORMAT
     delimiter: str = ","
@@ -84,6 +92,7 @@ class XML(BaseModel):
     type: XML_FORMAT
     root_tag: str
     row_tag: str
+    compression: XMLCompression = XMLCompression.GZIP
 
 
 class ORC(BaseModel):

tests/test_integration/test_run_transfer/test_hdfs.py

@@ -229,8 +229,8 @@ async def test_run_transfer_hdfs_to_postgres(
             id="parquet",
         ),
         pytest.param(
-            ("xml", {}),
-            "without_compression",
+            ("xml", {"compression": "snappy"}),
+            "with_compression",
             id="xml",
         ),
     ],

tests/test_integration/test_run_transfer/test_s3.py

@@ -230,7 +230,7 @@ async def test_run_transfer_s3_to_postgres(
             id="parquet",
         ),
         pytest.param(
-            ("xml", {}),
+            ("xml", {"compression": "none"}),
             "without_compression",
             id="xml",
         ),

tests/test_unit/test_transfers/test_file_transfers/test_create_transfer.py

@@ -59,6 +59,7 @@
                 "type": "xml",
                 "root_tag": "data",
                 "row_tag": "record",
+                "compression": "lz4",
             },
             "options": {
                 "some": "option",
@@ -166,6 +167,7 @@ async def test_developer_plus_can_create_s3_transfer(
             "type": "xml",
             "root_tag": "data",
             "row_tag": "record",
+            "compression": "lz4",
         },
         "orc": {
             "type": "orc",
@@ -221,6 +223,7 @@ async def test_developer_plus_can_create_s3_transfer(
                 "type": "xml",
                 "root_tag": "data",
                 "row_tag": "record",
+                "compression": "bzip2",
             },
         },
         {
@@ -320,6 +323,7 @@ async def test_developer_plus_can_create_hdfs_transfer(
             "type": "xml",
             "root_tag": "data",
             "row_tag": "record",
+            "compression": "bzip2",
         },
         "orc": {
             "type": "orc",

tests/test_unit/test_transfers/test_file_transfers/test_read_transfer.py

@@ -41,6 +41,7 @@
                 "type": "xml",
                 "root_tag": "data",
                 "row_tag": "record",
+                "compression": "bzip2",
             },
             "options": {},
         },

tests/test_unit/test_transfers/test_file_transfers/test_update_transfer.py

@@ -41,6 +41,7 @@
                 "type": "xml",
                 "root_tag": "data",
                 "row_tag": "record",
+                "compression": "bzip2",
             },
             "options": {},
         },