Preserve literal NA values and blank cells in Excel conversions

packages/markitdown/src/markitdown/converters/_xlsx_converter.py

@@ -52,12 +52,22 @@ def _read_xlsx_sheets(
     repaired_stream = None
     try:
         try:
-            sheets = pd.read_excel(file_stream, sheet_name=None, engine="openpyxl")
+            sheets = pd.read_excel(
+                file_stream,
+                sheet_name=None,
+                engine="openpyxl",
+                keep_default_na=False,
+            )
         except TypeError as exc:
             if "showZeroes" not in str(exc):
                 raise
             repaired_stream = _repair_sheetview_show_zeroes(file_stream, start_pos)
-            sheets = pd.read_excel(repaired_stream, sheet_name=None, engine="openpyxl")
+            sheets = pd.read_excel(
+                repaired_stream,
+                sheet_name=None,
+                engine="openpyxl",
+                keep_default_na=False,
+            )
         yield sheets, repaired_stream if repaired_stream is not None else file_stream
     finally:
         if repaired_stream is not None:
@@ -237,7 +247,9 @@ def convert(
                 _xls_dependency_exc_info[2]
             )
 
-        sheets = pd.read_excel(file_stream, sheet_name=None, engine="xlrd")
+        sheets = pd.read_excel(
+            file_stream, sheet_name=None, engine="xlrd", keep_default_na=False
+        )
         md_content = ""
         for s in sheets:
             md_content += f"## {s}\n"

packages/markitdown/tests/test_excel_literal_values.py

@@ -0,0 +1,66 @@
+import io
+import zipfile
+from pathlib import Path
+
+import pytest
+from openpyxl import Workbook
+
+from markitdown import MarkItDown
+
+ROWS = [
+    ["token", "blank", "amount"],
+    ["NA", None, 1],
+    ["NULL", None, 2],
+    ["None", None, 3],
+    ["n/a", None, 4],
+    ["nan", "present", 5],
+]
+
+
+@pytest.mark.parametrize("file_format", ["xlsx", "xlsx_legacy", "xls"])
+def test_excel_preserves_literal_na_values(file_format):
+    if file_format == "xls":
+        # The legacy workbook contains ROWS, including real blank cells.
+        stream = io.BytesIO(
+            Path(__file__)
+            .with_name("test_files")
+            .joinpath("literal_na.xls")
+            .read_bytes()
+        )
+    else:
+        workbook = Workbook()
+        workbook.active.title = "Data"
+        for row in ROWS:
+            workbook.active.append(row)
+        stream = io.BytesIO()
+        workbook.save(stream)
+        stream.seek(0)
+
+        if file_format == "xlsx_legacy":
+            repaired_input = io.BytesIO()
+            with zipfile.ZipFile(stream) as source:
+                with zipfile.ZipFile(repaired_input, "w") as target:
+                    for item in source.infolist():
+                        data = source.read(item.filename)
+                        if item.filename == "xl/worksheets/sheet1.xml":
+                            data = data.replace(
+                                b"<sheetView ", b'<sheetView showZeroes="0" ', 1
+                            )
+                        target.writestr(item, data)
+            stream = repaired_input
+            stream.seek(0)
+
+    result = MarkItDown().convert_stream(
+        stream, file_extension=".xls" if file_format == "xls" else ".xlsx"
+    )
+
+    assert result.markdown == (
+        "## Data\n"
+        "| token | blank | amount |\n"
+        "| --- | --- | --- |\n"
+        "| NA |  | 1 |\n"
+        "| NULL |  | 2 |\n"
+        "| None |  | 3 |\n"
+        "| n/a |  | 4 |\n"
+        "| nan | present | 5 |"
+    )