[chdre]

Table detection in MsExcelDocumentBackend called sheet.iter_rows(...) and sheet.cell(...) over the whole data bounds. In openpyxl both create a Cell for every position they visit and keep it in sheet._cells. A sheet with a few values in a large rectangle therefore grew by one cell object per empty position, about 200 MB per million cells, and that memory was not released.

This change reads sheet._cells directly in _find_data_tables and _find_table_bounds:

• Existing cells are visited in the same row-major order.
• A missing cell counts as having no value and no text.
• A missing merge anchor is still recognised, through a new _MergedCellIndex.contains_at.

The output does not change. _cells was already read directly in _find_true_data_bounds and _value_bounds.

Checklist

• Documentation has been updated, if necessary. (not needed, internal change)
• Examples have been added, if necessary. (not needed)
• Tests have been added, if necessary.

Review evidence

Measured with DocumentConverter().convert() on generated workbooks, median of 3, tracemalloc (Python allocations only), Windows, Intel Core Ultra 9 285HX.

Workbook	Cells in sheet after table detection	Peak memory	Wall time	Markdown
Sparse: values in A1 and AD33334 (before)	1,000,020	194.3 MB	4.1 s	identical
Sparse (after)	2	1.1 MB	0.5 s	identical
Dense, 5000 rows x 30 columns (before)	150,000	388.3 MB	12.4 s	identical
Dense (after)	150,000	388.3 MB	11.7 s	identical

Dense sheets are unchanged, as expected.

New tests in tests/test_backend_msexcel.py:

• test_find_data_tables_does_not_materialize_empty_cells: the sheet's cell count is unchanged by table detection, and peak memory stays under 50 MiB. It fails on main.
• test_sparse_merged_anchor_and_comment_only_cell_tables: output equality for merges and comments in a sparse sheet.
• test_missing_merge_anchor_inside_table_keeps_span: a merge whose anchor cell is missing keeps its span. It fails without contains_at.

Old and new _find_data_tables output was also compared on all workbooks in tests/data and on 300 random sheets (values, merges, comments, gap tolerance 0 to 3), with 0 differences.

Commands run:

• uv run pytest tests/test_backend_msexcel.py tests/test_backend_msexcel_charts.py: 43 passed, 2 skipped.
• uv run prek run --files ...: all hooks passed.

Not run:

• The full test suite.
• Ground truth was not regenerated. test_e2e_excel_conversions passes unchanged.

Not included:

• Lighter intermediate objects for ExcelCell and TableCell.
• Processing a sheet in row windows.
• The chart code that still calls sheet.cell(); it is already capped by MAX_CHART_TABLE_CELLS.

[github-actions]

✅ DCO Check Passed

Thanks @chdre, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[github-actions]

AI triage (advisory)

Checked commit 9840b009e2. The analysis read the diff only. It did not run code from this pull request.

The PR changes xlsx table detection in MsExcelDocumentBackend so it reads only existing cells from sheet._cells. This stops openpyxl from creating empty cells for sparse sheets. It also adds _MergedCellIndex.contains_at and new tests.

CI safety: no concern found. A maintainer must still approve CI.

1 related pull request(s) in the same code

PR	State	Signal	Assessment
#5015	open	changes the same code	related: Both PRs change msexcel_backend.py and its tests. This PR fixes memory use in table detection. The other PR reads images placed in cells, which is a different change.

_{The ai:* labels are advisory. They do not approve CI and they do not add tests:full. A new push runs this check again.}

[github-actions]

AI first review (advisory): no blocker found

This review covers the full diff at 7dbc2abc65. It read the code and did not run it.

I checked table detection in MsExcelDocumentBackend for sparse sheets, coordinate traversal order, and merged-cell fallback handling. The change avoids creating empty Cell objects in sheet._cells and keeps table extraction results intact.

_{A maintainer makes the merge decision. This review never approves the pull request.}

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
docling/backend/msexcel_backend.py	83.33%	3 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

AI first review (advisory): no blocker found

This review covers the changes since 7dbc2abc65 at 9840b009e2. It read the code and did not run it.

The incremental change adds one test (test_threaded_comment_fallback_without_xml) for the threaded-comment XML-absent fallback path. The test logic is correct: cells with values are already in _cells after load, comments are injected in-memory, and _parse_threaded_comments correctly returns empty because no XML part exists in the saved file. The earlier review found no blockers; the original change has no new behavioral code in this diff. The human reviewer should confirm that Codecov's 5 uncovered lines do not include critical branches from this PR's new code paths.

_{A maintainer makes the merge decision. This review never approves the pull request.}