[cdeil]

Problem. <row> without r is valid (the attribute is optional in
ECMA-376 18.3.1.73) and means the row after the previous one. The SAX reader
left the row index at −1 and dropped every cell in such a row, even one placed
by its own r="A1"; an unindexed row after an indexed one reused the previous
row's index, so positional cells overwrote it. openpyxl reads
<row><c r="A1"><v>42</v></c></row> as A1 = 42; the library returned null and
dropped the value on save. Inherited from the original package.

A second effect of the same loop: ht, hidden, outlineLevel and
collapsed were applied as they were met, so a writer that puts them before
r attached them to the previous row.

Fix. The row index is inferred when r is absent, a cell carrying its own
reference is always placed by it, and the row's attributes are applied once
the index is known.

Validation. test/reader_test.dart, "Rows Without Explicit Indices":
omitted r with an explicit cell, an unindexed row after r="3" (A3, A4,
B4, matching openpyxl's reading of the same file), rows with no index at all,
and attributes written before r. All fail before, pass after; suite and
analyzer clean.

🤖 Generated with Claude Code