Fix up encoding and avoid encode-decode waste. Resolves #82 (#86)

* Fix up encoding and avoid encode-decode waste.  Resolves #82

* Fix up type hint.

sycamore/data/document.py

@@ -32,7 +32,7 @@ def text_representation(self, value: str) -> None:
         self.data["text_representation"] = value
 
     @property
-    def binary_representation(self) -> Optional[str]:
+    def binary_representation(self) -> Optional[bytes]:
         return self.data["binary_representation"]
 
     @binary_representation.setter

sycamore/transforms/partition.py

@@ -46,10 +46,17 @@ def element_in_left_col(e: Element) -> bool:
 class Partitioner(ABC):
     @staticmethod
     def to_element(dict: dict[str, Any]) -> Element:
+        text = dict.pop("text")
+        if isinstance(text, str):
+            binary = text.encode("utf-8")
+        else:
+            binary = text
+            text = str(binary, "utf-8")
+
         element = Element()
         element.type = dict.pop("type")
-        element.binary_representation = dict.pop("text")
-        element.text_representation = str(element.binary_representation)
+        element.binary_representation = binary
+        element.text_representation = text
         element.properties.update(dict.pop("metadata"))
         element.properties.update(dict)