[ENG-8488] Fix inconsistent Excel rendering by enforcing binary mode for .xls/.xlsx (#401)

Author: sh-andriy

mfr/extensions/tabular/render.py

@@ -11,6 +11,7 @@
 
 logger = logging.getLogger(__name__)
 
+BINARY_EXCEL_EXTS = {'.xls', '.xlsx'}
 
 class TabularRenderer(extension.BaseRenderer):
 
@@ -30,8 +31,14 @@ def render(self):
                 extension=self.metadata.ext,
             )
 
-        with open(self.file_path, errors='replace') as fp:
-            sheets, size, nbr_rows, nbr_cols = self._render_grid(fp, self.metadata.ext)
+        ext = (self.metadata.ext or '').lower()
+        if ext in BINARY_EXCEL_EXTS:
+            open_kwargs = {'mode': 'rb'}
+        else:
+            open_kwargs = {'errors': 'replace'}
+
+        with open(self.file_path, **open_kwargs) as fp:
+            sheets, size, nbr_rows, nbr_cols = self._render_grid(fp, ext)
 
         # Force GC
         gc.collect()

mfr/extensions/tabular/utilities.py

@@ -90,41 +90,28 @@ def sav_to_csv(fp):
 
 def to_bytes(fp):
     """
-    Return *exactly* the original bytes of the Excel file and rewind *fp*.
-    Handles both binary and text wrappers that WaterButler may give us.
+    Return exactly the original bytes and rewind fp.
+    Requires a binary file-like object or a bytes object.
     """
-    try:
-        fp.seek(0)
-    except Exception:
-        pass
-
-    raw = fp.read()
-    if isinstance(raw, bytes):
-        try:
-            fp.seek(0)
-        except Exception:
-            pass
-        return raw
+    if isinstance(fp, (bytes, bytearray, memoryview)):
+        return bytes(fp)
 
-    if hasattr(fp, "buffer"):
-        buf = fp.buffer
+    if hasattr(fp, "read"):
         try:
-            buf.seek(0)
+            if hasattr(fp, "seek"):
+                fp.seek(0)
         except Exception:
             pass
-        data = buf.read()
+        raw = fp.read()
         try:
-            buf.seek(0)
+            if hasattr(fp, "seek"):
+                fp.seek(0)
         except Exception:
             pass
-    else:
-        data = raw.encode("utf-8", "surrogateescape")
+        if isinstance(raw, (bytes, bytearray, memoryview)):
+            return bytes(raw)
 
-    try:
-        fp.seek(0)
-    except Exception:
-        pass
-    return data
+    raise TypeError("Expected binary file-like object; got text/str")
 
 
 def _extract_rows(fields, raw_rows):