std.data.csv - CSV parsing

std.data.csv provides dialect-aware whole-table and streaming CSV parsing and rendering. Parsed documents are structured Csv values; quoting, delimiters, row shape, and contextual errors are handled by the package.

data.csv.parse(string text) -> result<Csv, CsvError> - parse rows without headers (each row is a list of strings).
data.csv.parse_with_headers(string text) -> result<Csv, CsvError> - parse rows and preserve the first line as header names.
data.csv.parse_with_options(string text, int delimiter, int quote, bool has_headers, bool trim, bool flexible) -> result<Csv, CsvError> - configure the dialect explicitly; delimiter and quote are ASCII byte values.

The resulting Csv value is a first-class type in the stdlib (see csv.stringify for output). Use table.stringify_with(delimiter, quote) when emitting a non-default dialect. Both parsing and writing reject invalid control-byte delimiters and quotes; malformed rows return contextual errors rather than being dropped. Whole-table and byte-reader inputs must be valid UTF-8 and are bounded to 16 MiB before parsing.

All fallible CSV operations return CsvError. Its kind is the typed csv.CsvErrorKind enum (Invalid, Parse, Type, Limit, or Io), while detail remains textual. Compare or match the enum for programmatic handling and use error.message() for display.

Example:

Mux

For bounded incremental processing, use the typed stream handles. A reader consumes one record at a time, and a writer can either retain encoded bytes or forward each record to an existing io.Writer:

Mux

Use an existing stream when the CSV is part of a larger pipeline:

Mux

CsvReader.from_reader(reader, has_headers) pulls bytes from the supplied reader as records are requested and leaves that reader open. It enforces the same 16 MiB input limit as the byte constructor. CsvWriter.from_writer forwards encoded records to the supplied writer and leaves it open; its bytes() method returns an error because output is not retained by the CSV handle. Call flush() when the destination needs an explicit flush.

CsvReader.from_bytes(bytes, has_headers) and CsvWriter.from_config(delimiter, quote) are explicit parameterized constructors; new() remains the no-argument default. Rows are list<string>, and malformed records remain typed errors.