[neilcawse]

What changed

Three XLSX round-trip losses, all in crates/xlsx:

• _xlpm. LET/LAMBDA parameters. Excel stores LET/LAMBDA parameter names as _xlpm.name in the file. The writer already adds the prefix, but read_formula (sheet_read.rs) used the raw <f> text, so a saved LET(x,1,x+1) came back as _xlfn.LET(_xlpm.x,1,_xlpm.x+1). Cell, shared and array formulas now go through fmla::from_file, the conversion CF, data validation and names already use. When the formula parses, the stored text is the parser/printer form with _xlfn./_xlpm. removed. When it doesn't parse, the text is kept as written. One side effect: cell formula text read from a file is now in printer form (for example, spacing may change) rather than byte-for-byte as stored.
• Line breaks in table column names. table_columns (sheet_write.rs) replaced \r/\n with spaces in column names. Excel allows a line break in a table header, and the header cell still had the break, so the column name and its header cell disagreed after a save. The name is now written unchanged (XML escaping already handles it).
• Leading backslash in table names. table_name turned a leading \ into _ (or fell back to TableN). Excel's rules for names in formulas allow a name to start with a letter, _ or \, so \_Prime.1 is now kept. Structured references such as COUNTA(\_Prime.1[Name]) already parsed.

Tests

• tests::roundtrip::let_lambda_parameters_drop_xlpm_on_read: LET, an immediately called LAMBDA and an array MAP/LAMBDA formula are written with _xlpm. in the sheet XML and read back with bare parameter names.
• tests::roundtrip::table_newline_column_and_backslash_name_survive: a table named \_Prime.1 with a Handle\nBars column keeps both after a round trip, along with the header cell value and a structured reference formula to it.

Commands run:

cargo fmt --all -- --check
cargo clippy -p gridcraft-xlsx -p gridcraft-engine --all-targets -- -D warnings
cargo test -p gridcraft-xlsx -p gridcraft-engine

Before this change, the gridcraft-cli check round trip (#416, not needed by this PR) reported 3 of 41 local corpus files differing after a save. With the change, it reported 0. That was measured on an older base, before rebasing onto current main.

Docs

• docs/file-format-parity.md: header and a history row for this change. The header and the history table overlap other XLSX PRs that touch this file, so a rebase may be needed.

Not done

• No change to how other formula sources (chart series, pivot formulas) are read. They were not part of the round-trip differences found.

Unverified: needs confirmation in Excel

• That Excel opens a table header with a line break, written as-is in tableColumn@name, without a repair prompt. The header cell and the column name now match, which is what Excel itself writes, but this has not been opened in Excel.
• That Excel accepts a table name starting with \ in table@name/displayName.

[echelon]

Merged — thank you for the contribution! If you haven't already, please join our Discord: https://discord.gg/artcraft