std.toml

import std.toml

config = toml.parse(text)
text   = toml.stringify(value)
config = toml.read("config.toml")
toml.write("config.toml", value)

Identical in shape to std.yaml and std.json - parse, stringify, read, write - so switching format is switching the module name.

What it maps to

TOML tables become maps, arrays become lists, and TOML's date-time types arrive as strings.

A top-level map is required for stringify, because TOML has no non-table root. Serializing a bare list or number is an error rather than something invented.

When to use it

TOML is a good fit for configuration a person edits: comments survive, the syntax is unambiguous, and its strictness catches typos that YAML would silently accept as a different value.

That last row matters: ecko.json beside your script is loaded automatically and readable as defaults.<key>, which for a project's own configuration is less work than any of these.

Errors

Malformed input raises { kind: "parse", format: "toml" }, with path from toml.read. Catch it around any file a user can edit, and report the message - it names the line.

Comments

Parsing discards comments, so a read-modify-write cycle loses them. For a config file a person maintains, prefer writing it by hand and only reading it in code.