fix(xlsx): a blank cell should not rewrite the rest of its column

packages/markitdown-ocr/tests/test_xlsx_converter.py

@@ -95,24 +95,24 @@ def test_xlsx_image_middle(svc: MockOCRService) -> None:
         "## Revenue\n"
         "| Q1 Report | Unnamed: 1 |\n"
         "| --- | --- |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| Revenue | $50,000 |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
         "| Profit Margin | 40% |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}\n\n"
         "## Expenses\n"
         "| Expense Breakdown | Unnamed: 1 |\n"
         "| --- | --- |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| Expenses | $30,000 |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
         "| Savings | $5,000 |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}"
@@ -133,12 +133,12 @@ def test_xlsx_image_end(svc: MockOCRService) -> None:
         "| Total Revenue | $500,000 |\n"
         "| Total Expenses | $300,000 |\n"
         "| Net Profit | $200,000 |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| Signature: | NaN |\n\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "| Signature: |  |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}\n\n"
         "## Budget\n"
@@ -147,12 +147,12 @@ def test_xlsx_image_end(svc: MockOCRService) -> None:
         "| Marketing | $100,000 |\n"
         "| R&D | $150,000 |\n"
         "| Operations | $50,000 |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| NaN | NaN |\n"
-        "| Approved: | NaN |\n\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "|  |  |\n"
+        "| Approved: |  |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}"
     )
@@ -170,10 +170,10 @@ def test_xlsx_multiple_images(svc: MockOCRService) -> None:
         "| Dashboard |\n"
         "| --- |\n"
         "| Status: Active |\n"
-        "| NaN |\n"
-        "| NaN |\n"
-        "| NaN |\n"
-        "| NaN |\n"
+        "|  |\n"
+        "|  |\n"
+        "|  |\n"
+        "|  |\n"
         "| Performance Summary |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}\n\n"
@@ -204,27 +204,27 @@ def test_xlsx_complex_layout(svc: MockOCRService) -> None:
         "## Complex Report\n"
         "| Annual Report 2024 | Unnamed: 1 |\n"
         "| --- | --- |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| Month | Sales |\n"
         "| Jan | 1000 |\n"
         "| Feb | 1200 |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| Total | 2200 |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}\n\n"
         f"{_OCR_BLOCK}\n\n"
         "## Customers\n"
         "| Customer Metrics | Unnamed: 1 |\n"
         "| --- | --- |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| New Customers | 250 |\n"
         "| Retention Rate | 92% |\n\n"
         "### Images in this sheet:\n\n"
         f"{_OCR_BLOCK}\n\n"
         "## Regions\n"
         "| Regional Breakdown | Unnamed: 1 |\n"
         "| --- | --- |\n"
-        "| NaN | NaN |\n"
+        "|  |  |\n"
         "| Region | Revenue |\n"
         "| North | $800K |\n"
         "| South | $600K |\n\n"

packages/markitdown-ocr/tests/test_xlsx_inheritance.py

@@ -288,10 +288,35 @@ def fixed_read(*args: Any, **kwargs: Any):
     assert "Future shared fix" in result and "<native>" not in result
     assert result.count("[Image OCR]") == 2
     repair.assert_called_once()
-    assert calls == [{"sheet_name": None, "engine": "openpyxl"}] * 2
+    assert calls == [{"sheet_name": None, "engine": "openpyxl", "dtype": object}] * 2
     service.extract_text.assert_called_once()
 
 
+def test_blank_cell_stays_blank_and_keeps_its_column_native() -> None:
+    workbook = openpyxl.Workbook()
+    sheet = workbook.active
+    sheet.title = "Cells"
+    sheet.append(["Units", "Shipped"])
+    sheet.append([12, True])
+    sheet.append([None, None])
+    sheet.append([7, False])
+    sheet.add_image(SheetImage(io.BytesIO(_RED)), "D1")
+    stream = io.BytesIO()
+    workbook.save(stream)
+    workbook.close()
+
+    assert _convert(XlsxConverterWithOCR(_service()), stream.getvalue()) == (
+        "## Cells\n"
+        "| Units | Shipped |\n"
+        "| --- | --- |\n"
+        "| 12 | True |\n"
+        "|  |  |\n"
+        "| 7 | False |\n\n"
+        "### Images in this sheet:\n\n"
+        "*[Image OCR]  \nrecognized  \n[End OCR]*"
+    )
+
+
 def test_reported_ocr_error_warns_once_and_keeps_native_output() -> None:
     service = Mock(
         extract_text=Mock(

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

@@ -1,3 +1,4 @@
+import datetime
 import io
 import re
 import sys
@@ -52,18 +53,60 @@ 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", dtype=object
+            )
         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", dtype=object
+            )
         yield sheets, repaired_stream if repaired_stream is not None else file_stream
     finally:
         if repaired_stream is not None:
             repaired_stream.close()
 
 
+def _format_cell(value: Any) -> str:
+    """Render one cell the way the spreadsheet holds it.
+
+    openpyxl hands back a ``datetime`` for a date cell, and ``str()`` on it
+    appends a midnight time that is not in the spreadsheet.
+    """
+    if isinstance(value, datetime.datetime):
+        if value.time() == datetime.time.min:
+            return value.date().isoformat()
+        return value.isoformat(sep=" ")
+    return str(value)
+
+
+def _format_float(value: float) -> str:
+    """Render a fractional number with the 15 significant digits Excel shows."""
+    return format(value, ".15g")
+
+
+def sheet_to_html(sheet: Any) -> str:
+    """Render one sheet as an HTML table.
+
+    ``na_rep=""`` keeps an empty cell empty: the default writes the string
+    ``NaN`` into it, which reads as a value rather than as a blank. The
+    formatters are what keep the remaining cells rendered as themselves --
+    pandas applies ``na_rep`` to the blanks, ``float_format`` to fractional
+    numbers, and the formatter to everything else, without re-inferring a
+    dtype for the column. A float needs ``float_format`` because with
+    ``index=False`` pandas renders it with its own display precision and
+    never reaches ``formatters``.
+    """
+    return sheet.to_html(
+        index=False,
+        na_rep="",
+        formatters=[_format_cell] * len(sheet.columns),
+        float_format=_format_float,
+    )
+
+
 def _rename_show_zeroes_attribute(data: bytes) -> bytes:
     return _SHEET_VIEW_START_TAG.sub(
         lambda match: _SHOW_ZEROES_ATTRIBUTE.sub(rb"showZeros\1", match.group(0)),
@@ -150,7 +193,7 @@ def convert(
 
             for s in sheets:
                 md_content += f"## {s}\n"
-                html_content = sheets[s].to_html(index=False)
+                html_content = sheet_to_html(sheets[s])
                 md_content += (
                     self._html_converter.convert_string(
                         html_content, **kwargs
@@ -237,11 +280,13 @@ 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", dtype=object
+        )
         md_content = ""
         for s in sheets:
             md_content += f"## {s}\n"
-            html_content = sheets[s].to_html(index=False)
+            html_content = sheet_to_html(sheets[s])
             md_content += (
                 self._html_converter.convert_string(
                     html_content, **kwargs