fix(xlsx): speed up by detecting the true last non-empty row/column (#2404)

* Update msexcel_backend.py

Fix #2307, Follow the instruction of https://github.com/docling-project/docling/issues/2307#issuecomment-3327248503.

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>

* Update msexcel_backend.py

Fix error

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>

* Fix linting issues

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>

* Add test files and data (Signed-off-by: Huangrui Chu <huangrui.chu.1999@gmail.com>)

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>

* resolve conflict with test_backend_msexecl; update the boundary

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>

* chore(xlsx): use a dataclass to represent a bounding rectangle in worksheets

Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com>

* chore(xlsx): increase parsing speed by iterating on 'sheet._cells'

Increase the parsing speed of the spreadsheet backend by iterating on 'sheets._cells'
since this is proportional to the number of created cells.
Rename test file to align it to other test files.

Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com>

---------

Signed-off-by: Richard (Huangrui) Chu <65276824+HuangruiChu@users.noreply.github.com>
Signed-off-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com>
Co-authored-by: Cesar Berrospi Ramis <ceb@zurich.ibm.com>

tests/data/groundtruth/docling_v2/xlsx_04_inflated.xlsx.itxt

@@ -0,0 +1,11 @@
+item-0 at level 0: unspecified: group _root_
+  item-1 at level 1: section: group sheet: Sheet1
+    item-2 at level 2: table with [7x3]
+  item-3 at level 1: section: group sheet: Sheet2
+    item-4 at level 2: table with [9x4]
+    item-5 at level 2: table with [5x3]
+    item-6 at level 2: table with [5x3]
+  item-7 at level 1: section: group sheet: Sheet3
+    item-8 at level 2: table with [7x3]
+    item-9 at level 2: table with [7x3]
+    item-10 at level 2: picture

tests/data/groundtruth/docling_v2/xlsx_04_inflated.xlsx.md

@@ -0,0 +1,53 @@
+|   first  |   second  |   third |
+|----------|-----------|---------|
+|        1 |         5 |       9 |
+|        2 |         4 |       6 |
+|        3 |         3 |       3 |
+|        4 |         2 |       0 |
+|        5 |         1 |      -3 |
+|        6 |         0 |      -6 |
+
+|   col-1 |   col-2 |   col-3 |   col-4 |
+|---------|---------|---------|---------|
+|       1 |       2 |       3 |       4 |
+|       2 |       4 |       6 |       8 |
+|       3 |       6 |       9 |      12 |
+|       4 |       8 |      12 |      16 |
+|       5 |      10 |      15 |      20 |
+|       6 |      12 |      18 |      24 |
+|       7 |      14 |      21 |      28 |
+|       8 |      16 |      24 |      32 |
+
+|   col-1 |   col-2 |   col-3 |
+|---------|---------|---------|
+|       1 |       2 |       3 |
+|       2 |       4 |       6 |
+|       3 |       6 |       9 |
+|       4 |       8 |      12 |
+
+|   col-1 |   col-2 |   col-3 |
+|---------|---------|---------|
+|       1 |       2 |       3 |
+|       2 |       4 |       6 |
+|       3 |       6 |       9 |
+|       4 |       8 |      12 |
+
+| first    | header   | header   |
+|----------|----------|----------|
+| first    | second   | third    |
+| 1        | 2        | 3        |
+| 3        | 4        | 5        |
+| 3        | 6        | 7        |
+| 8        | 9        | 9        |
+| 10       | 9        | 9        |
+
+| first (f)   | header (f)   | header (f)   |
+|-------------|--------------|--------------|
+| first (f)   | second       | third        |
+| 1           | 2            | 3            |
+| 3           | 4            | 5            |
+| 3           | 6            | 7            |
+| 8           | 9            | 9            |
+| 10          | 9            | 9            |
+
+<!-- image -->

tests/test_backend_msexcel.py

@@ -2,6 +2,7 @@ import logging
 from pathlib import Path
 
 import pytest
+from openpyxl import load_workbook
 
 from docling.backend.msexcel_backend import MsExcelDocumentBackend
 from docling.datamodel.base_models import InputFormat
@@ -113,3 +114,67 @@ def test_chartsheet(documents) -> None:
     assert doc.groups[1].name == "sheet: Duck Chart"
     assert doc.pages[2].size.height == 0
     assert doc.pages[2].size.width == 0
+
+
+def test_inflated_rows_handling(documents) -> None:
+    """Test that files with inflated max_row are handled correctly.
+
+    xlsx_04_inflated.xlsx has inflated max_row (1,048,496) but only 7 rows of actual data.
+    This test verifies that our backend correctly identifies true data bounds.
+    """
+    # First, verify the file has inflated max_row using openpyxl directly
+    path = next(item for item in get_excel_paths() if item.stem == "xlsx_04_inflated")
+
+    wb = load_workbook(path)
+    ws = wb.active
+    reported_max_row = ws.max_row
+
+    # Assert that openpyxl reports inflated max_row
+    assert reported_max_row > 100000, (
+        f"xlsx_04_inflated.xlsx should have inflated max_row (expected >100k, got {reported_max_row:,}). "
+        f"This test file is designed to verify proper handling of Excel files with inflated row counts."
+    )
+
+    _log.info(
+        f"xlsx_04_inflated.xlsx - Openpyxl reported max_row: {reported_max_row:,}"
+    )
+
+    # Now test that our backend handles it correctly
+    in_doc = InputDocument(
+        path_or_stream=path,
+        format=InputFormat.XLSX,
+        filename=path.stem,
+        backend=MsExcelDocumentBackend,
+    )
+    backend = MsExcelDocumentBackend(in_doc=in_doc, path_or_stream=path)
+
+    # Verify backend detects correct number of pages (should be 4, like test-01)
+    page_count = backend.page_count()
+    assert page_count == 4, (
+        f"Backend should detect 4 pages (same as test-01), got {page_count}"
+    )
+
+    # Verify converted document has correct pages
+    doc = next(item for path, item in documents if path.stem == "xlsx_04_inflated")
+    assert len(doc.pages) == 4, f"Document should have 4 pages, got {len(doc.pages)}"
+
+    # Verify page sizes match expected dimensions (same as test-01)
+    # These should reflect actual data, not inflated row counts
+    assert doc.pages.get(1).size.as_tuple() == (3.0, 7.0), (
+        f"Page 1 should be 3x7 cells, got {doc.pages.get(1).size.as_tuple()}"
+    )
+    assert doc.pages.get(2).size.as_tuple() == (9.0, 18.0), (
+        f"Page 2 should be 9x18 cells, got {doc.pages.get(2).size.as_tuple()}"
+    )
+    assert doc.pages.get(3).size.as_tuple() == (13.0, 36.0), (
+        f"Page 3 should be 13x36 cells, got {doc.pages.get(3).size.as_tuple()}"
+    )
+    assert doc.pages.get(4).size.as_tuple() == (0.0, 0.0), (
+        f"Page 4 should be 0x0 cells (empty), got {doc.pages.get(4).size.as_tuple()}"
+    )
+
+    _log.info(
+        f"✓ Successfully handled inflated max_row: "
+        f"reported {reported_max_row:,} rows, "
+        f"correctly processed as {page_count} pages with proper dimensions"
+    )