Fixing test_extra_test_iss1541

pypdf/_writer.py

@@ -2194,7 +2194,7 @@ def clean(content: ContentStream) -> None:
                     i += 1
                 else:
                     i += 1
-            content.get_data()  # this ensures ._data is rebuilt
+            content.get_data()  # this ensures ._data is rebuilt from the .operations
 
         try:
             d = cast(dict, cast(DictionaryObject, page["/Resources"])["/XObject"])

pypdf/generic/_data_structures.py

@@ -972,6 +972,22 @@ def set_data(self, data: bytes) -> None:  # deprecated
 
 
 class ContentStream(DecodedStreamObject):
+    """
+    In order to be fast, this datastructure can contain either:
+    * raw data in ._data
+    * parsed stream operations in ._operations
+
+    At any time, ContentStream object can either have one or both of those fields defined,
+    and zero or one of those fields set to None.
+
+    Those fields are "rebuilt" lazily, when accessed:
+    * when .get_data() is called, if ._data is None, it is rebuilt from ._operations
+    * when .operations is called, if ._operations is None, it is rebuilt from ._data
+
+    On the other side, those fields can be invalidated:
+    * when .set_data() is called, ._operations is set to None
+    * when .operations is set, ._data is set to None
+    """
     def __init__(
         self,
         stream: Any,
@@ -1211,6 +1227,14 @@ def operations(self, operations: List[Tuple[Any, Any]]) -> None:
         self._operations = operations
         self._data = b""
 
+    # This overrides the parent method:
+    def write_to_stream(
+        self, stream: StreamType, encryption_key: Union[None, str, bytes] = None
+    ) -> None:
+        if not self._data and self._operations:
+            self.get_data()  # this ensures ._data is rebuilt for ContentStream
+        super().write_to_stream(stream, encryption_key)
+
 
 def read_object(
     stream: StreamType,