std.csv

parseWithHeader returns Result<Array<Map<string, string>>, string>. A repeated header name returns Err instead of choosing a first or last column.

Header uniqueness is checked before converting data rows. Repeated names are rejected when cell values are equal and when the input contains a header without data rows.

The header names are the parsed CSV field values. Equality uses those string values without an additional trimming or case-folding step.

Use csv.parse when columns must remain positional. Positional parsing preserves every field and does not assign header names.

stringify accepts arrays of string rows and returns text. It preserves row order and cell text. Rows may have different field counts. Commas separate fields and every row ends with LF, including the last row. An empty array of rows becomes empty text.

A field containing a comma, quotation mark, CR or LF is enclosed in quotation marks. An embedded quotation mark is doubled. Embedded CR and LF remain inside the quoted field. A sole empty field is quoted to distinguish it from a row with no fields. Empty fields in a larger row need no quotation marks. Other whitespace and Unicode text remain unchanged.

stringifyWithHeader(rows, columns) -> Result<string, string> writes a header using the supplied column names and then projects every map onto that ordered list. An absent key supplies an empty cell. Keys outside the list are ignored. A header is emitted even when there are no data rows.

Column names use exact string equality. A repeated name, including a repeated empty name, returns a nonempty Err identifying that name and its zero based repeated index. Validation happens before row projection. A unique empty name is permitted. An empty column list writes an empty header row and one empty row for each input row.

Positional serialization has no value error for a well typed input. Named serialization uses Result for invalid columns. Incorrect row or cell types follow ordinary checking diagnostics. Both functions evaluate supplied arguments once from left to right and perform no I/O or mutation. Their collection arguments follow value semantics.

Named parsing and named writing require the same unique header names. Positional serialization remains available when a row has no naming schema. CSV parsing has its separate malformed input and field count rules.