feat: use tell() instead of __len__

Author: felixscherz

pyiceberg/avro/file.py

@@ -290,7 +290,7 @@ def write_block(self, objects: List[D]) -> None:
         self.encoder.write(self.sync_bytes)
 
     def __len__(self) -> int:
-        """Returns the total number number of bytes written."""
+        """Return the total number number of bytes written."""
         if self.closed:
             return len(self.output_file)
-        return len(self.output_stream)
+        return self.output_stream.tell()

pyiceberg/io/__init__.py

@@ -125,8 +125,8 @@ def __exit__(
         """Perform cleanup when exiting the scope of a 'with' statement."""
 
     @abstractmethod
-    def __len__(self) -> int:
-        """Returns the total number number of bytes written to the stream."""
+    def tell(self) -> int:
+        """Return the total number number of bytes written to the stream."""
 
 
 class InputFile(ABC):

pyiceberg/manifest.py

@@ -778,11 +778,10 @@ def add_entry(self, entry: ManifestEntry) -> ManifestWriter:
         return self
 
     def __len__(self) -> int:
-        """Returns the total number number of bytes written."""
+        """Return the total number number of bytes written."""
         return len(self._writer)
 
 
-
 class RollingManifestWriter:
     closed: bool
     _supplier: Generator[ManifestWriter, None, None]
@@ -833,8 +832,7 @@ def _should_roll_to_new_file(self) -> bool:
         if not self._current_writer:
             return False
         return (
-            self._current_file_rows >= self._target_number_of_rows
-            or len(self._current_writer) >= self._target_file_size_in_bytes
+            self._current_file_rows >= self._target_number_of_rows or len(self._current_writer) >= self._target_file_size_in_bytes
         )
 
     def _close_current_writer(self) -> None:

tests/utils/test_manifest.py

@@ -516,13 +516,6 @@ def test_rolling_manifest_writer(
         spec_id=demo_manifest_file.partition_spec_id,
     )
 
-    # The tests are using `PyArrowFileIO` where `OutputStream` is implemented as `pyarrow.lib.BufferedOutputStream`
-    # this is just to show the tests passing if `pyarrow.lib.BufferedOutputStream` would implement the
-    # new `OutputStream` protocol that includes a `__len__` method
-    from pyiceberg.avro.file import AvroOutputFile
-    AvroOutputFile.__len__ = lambda self: self.output_stream.tell()
-
-
     with TemporaryDirectory() as tmpdir:
 
         def supplier() -> Generator[ManifestWriter, None, None]: