deps(markitdown): relax magika pin to allow 1.x

packages/markitdown/pyproject.toml

@@ -28,7 +28,7 @@ dependencies = [
   "beautifulsoup4",
   "requests",
   "markdownify",
-  "magika~=0.6.1",
+  "magika>=0.6.1,<2",
   "charset-normalizer",
   "defusedxml",
 ]

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

@@ -149,24 +149,42 @@ def convert(
                 images = _XlsxImages(workbook_stream)
 
             for s in sheets:
-                md_content += f"## {s}\n"
-                html_content = sheets[s].to_html(index=False)
-                md_content += (
-                    self._html_converter.convert_string(
-                        html_content, **kwargs
-                    ).markdown.strip()
-                    + "\n\n"
-                )
+                # Build the sheet's two content channels separately: the cell table and
+                # any embedded images. A worksheet with no columns has no header row, so
+                # rendering its table would emit a heading followed by a one-column
+                # "table" with no delimiter row (`|\n|  |`), which is not valid Markdown.
+                # But that must not skip the image hook: an image-only worksheet (an
+                # embedded image and no cell values) still produces content through
+                # `_image_to_html`, and dropping it would lose that content silently.
+                table_md = ""
+                if not sheets[s].columns.empty:
+                    html_content = sheets[s].to_html(index=False)
+                    table_md = (
+                        self._html_converter.convert_string(
+                            html_content, **kwargs
+                        ).markdown.strip()
+                        + "\n\n"
+                    )
+                image_md = ""
                 if images is not None:
                     image_content = images.to_html(s, self._image_to_html, kwargs)
                     if image_content:
-                        md_content += (
+                        image_md = (
                             self._html_converter.convert_string(
                                 image_content, **kwargs
                             ).markdown.strip()
                             + "\n\n"
                         )
 
+                # Emit the heading only when the sheet actually produces content, so a
+                # genuinely empty sheet stays entirely absent rather than leaving a bare
+                # `## name` with nothing under it.
+                if not table_md and not image_md:
+                    continue
+                md_content += f"## {s}\n"
+                md_content += table_md
+                md_content += image_md
+
         return DocumentConverterResult(markdown=md_content.strip())
 
     def _image_to_html(
@@ -240,6 +258,10 @@ def convert(
         sheets = pd.read_excel(file_stream, sheet_name=None, engine="xlrd")
         md_content = ""
         for s in sheets:
+            # Same rule as the XLSX path: a sheet with no columns has no header row, so
+            # rendering it would emit a heading followed by broken table syntax.
+            if sheets[s].columns.empty:
+                continue
             md_content += f"## {s}\n"
             html_content = sheets[s].to_html(index=False)
             md_content += (

packages/markitdown/tests/test_xlsx.py

@@ -543,3 +543,77 @@ def test_xlsx_show_zeroes_rename_is_scoped_to_sheet_view_tags() -> None:
 
 if __name__ == "__main__":
     raise SystemExit(pytest.main([__file__]))
+
+
+# Worksheets with no columns
+
+
+def _workbook_with(sheets: list[tuple[str, list[list[Any]]]]) -> bytes:
+    workbook = openpyxl.Workbook()
+    workbook.remove(workbook.active)
+    for name, rows in sheets:
+        sheet = workbook.create_sheet(name)
+        for row in rows:
+            sheet.append(row)
+    stream = io.BytesIO()
+    workbook.save(stream)
+    return stream.getvalue()
+
+
+def test_completely_empty_sheet_is_skipped_instead_of_emitting_broken_table() -> None:
+    data = _workbook_with([("Empty", [])])
+
+    markdown = _convert(XlsxConverter(), data)
+
+    assert markdown == ""
+    assert "Empty" not in markdown
+    assert "|" not in markdown
+
+
+def test_sheet_with_a_header_but_no_data_rows_keeps_its_heading() -> None:
+    data = _workbook_with([("HeaderOnly", [["col"]])])
+
+    markdown = _convert(XlsxConverter(), data)
+
+    assert markdown == "## HeaderOnly\n| col |\n| --- |"
+
+
+def test_data_sheet_followed_by_an_empty_sheet() -> None:
+    data = _workbook_with([("HasData", [["col"], ["v"]]), ("Empty", [])])
+
+    markdown = _convert(XlsxConverter(), data)
+
+    assert markdown == "## HasData\n| col |\n| --- |\n| v |"
+    assert "Empty" not in markdown
+
+
+def test_empty_sheet_does_not_hide_the_next_sheet() -> None:
+    data = _workbook_with([("Empty", []), ("HasData", [["col"], ["v"]])])
+
+    markdown = _convert(XlsxConverter(), data)
+
+    assert markdown == "## HasData\n| col |\n| --- |\n| v |"
+
+
+def test_image_only_sheet_is_preserved_via_the_hook() -> None:
+    """An image-only worksheet (no cell values) must still reach the image hook.
+
+    `columns.empty` describes only the cell table; a sheet carrying an embedded
+    image and no data must not be skipped, or the `_image_to_html` output is lost.
+    """
+
+    def render(stream: BinaryIO, info: StreamInfo, **kwargs: Any) -> str:
+        return "<p>Quarterly revenue chart</p>"
+
+    workbook = openpyxl.Workbook()
+    sheet = workbook.active
+    sheet.title = "ChartOnly"
+    sheet.add_image(SheetImage(io.BytesIO(_RED)), "A1")
+    stream = io.BytesIO()
+    workbook.save(stream)
+    workbook.close()
+
+    markdown = _convert(_ImageConverter(render), stream.getvalue())
+
+    assert "Quarterly revenue chart" in markdown
+    assert "ChartOnly" in markdown