[kemo]

Summary

• Adds getCoordinatesInRange() and getCoordinatesOutsideRange() methods to Cells.php that filter cells using the internal index array
• Refactors insertNewBefore() in ReferenceHelper.php from a single loop over ALL cells to two targeted passes: one for affected cells (move) and one for outside cells (formula update only)
• Reduces the work from O(all cells) to O(affected cells) + O(formula cells outside range)

Changes

• src/PhpSpreadsheet/Collection/Cells.php: Added getCoordinatesInRange() and getCoordinatesOutsideRange()
• src/PhpSpreadsheet/ReferenceHelper.php: Two-pass refactor of insertNewBefore()
• tests/PhpSpreadsheetTests/ReferenceHelper6Test.php: 14 new tests covering row/column insert, formula updates, edge cases
• tests/PhpSpreadsheetTests/Benchmark/ReferenceHelperBenchmark.php: Benchmark with 10,000 cells, indexed vs full iteration comparison

Test plan

• Verify inserting rows correctly updates formula references
• Verify inserting columns correctly updates formula references
• Verify edge cases: insert at row 1, insert at last column, Z/AA boundary
• Run benchmark: vendor/bin/phpunit --group benchmark --filter ReferenceHelperBenchmark --stderr

[kemo]

Benchmark Results (reference-helper-cell-index)

Targeted benchmark: insert rows/columns into worksheets of increasing size, with formulas.

Benchmark	Master	Branch	Delta
10k cells, insert rows	807ms	798ms	-1%
10k cells, insert cols	825ms	816ms	-1%
60k cells, insert rows	4,886ms	4,965ms	+2%
60k cells, insert cols	5,567ms	5,382ms	-3%
100k cells, insert rows	8,098ms	7,963ms	-2%
100k cells, insert cols	8,728ms	8,231ms	-6%

Marginal improvement (2–6%) at larger sheet sizes. Within noise for smaller sheets. Memory usage unchanged.

[oleibman]

Thank you for the summaries. For purposes of prioritization, I am going to put the PRs with less demonstrable improvement, including this one, in draft status for now. This does not mean that I will not return to them.

[oleibman]

There is one uncovered line in the changes to Cells. I do not consider this a showstopper, especially since I am fairly sure that the test in question will never be true. But the lack of coverage is easily and safely eliminated, so we may as well do that.