3 releases (breaking)

0.3.1 Jan 20, 2024
0.2.0 Jan 15, 2024
0.1.0 Jan 14, 2024

#1309 in Parser implementations

Download history 2689/week @ 2024-07-22 2910/week @ 2024-07-29 3172/week @ 2024-08-05 2745/week @ 2024-08-12 2754/week @ 2024-08-19 2706/week @ 2024-08-26 2803/week @ 2024-09-02 2581/week @ 2024-09-09 2739/week @ 2024-09-16 2940/week @ 2024-09-23 2438/week @ 2024-09-30 2990/week @ 2024-10-07 2842/week @ 2024-10-14 2595/week @ 2024-10-21 2869/week @ 2024-10-28 2151/week @ 2024-11-04

10,844 downloads per month

MIT/Apache

42KB
1K SLoC

BOML

A dependency-free, (almost) zero-copy TOML parser for Rust.

Quick Demo

Parsing

BOML uses the Toml struct to parse a TOML file and load its root table. This will either return the successfully-parsed TOML, or detailed information about a syntax error. If parsing was successful, the Toml struct can essentially be used like a HashMap to get values from the root table.

use boml::prelude::*;

fn parse_cargo_toml() {
	let source = include_str!("../Cargo.toml");
  let toml = Toml::parse(source).unwrap();
  // If you prefer, `Toml::new()` will do the same thing:
  // let toml = Toml::new(source).unwrap();

  // Get the package table from the `Cargo.toml` file
  let package = toml.get("package").unwrap();
}

Types

BOML stores TOML data in a TomlValue enum, with variants for each type. In the example above, the package variable is just storing a TomlValue, which could be any TOML type and isn't very useful to us. We want package to be a table. BOML provides several ergonomic ways of doing this:

// Tables provide `get_<type>` methods, which will return the value as that type if it matches.
// If this method fails, it provides detailed error information, described below.
let package = toml.get_table("package").unwrap();
// All tables provide the `get_<type>` methods, so we can use them to get values from package, too
let name = package.get_string("name").unwrap();
assert_eq!(name, "boml");

// `TomlValue`s can be converted to one of their enum variants - this works similarly to the `.ok()` and
// `.err()` methods on `Result`s.
let package = toml.get("package").unwrap().table().unwrap();
let name = package.get("name").unwrap().string().unwrap();
assert_eq!(name, "boml");

// If you're really dedicated to boilerplate, you can also manually unwrap the enum variant.
let Some(TomlValue::Table(package)) = toml.get("package") else {
  panic!("I love boilerplate");
};

BOML also provides a TomlValueType enum, allowing you to determine a value's type at runtime. The .ty() method on TomlValues gives you this information:

let package = toml.get("package").unwrap();
assert_eq!(package.ty(), TomlValueType::Table);

Error Handling

There are 2 sources of errors in BOML: A parsing error, or an error from one of the get_<type> methods. These use the TomlError and TomlGetError types, respectively.

TomlError, the parsing error type, stores the span of text where the parsing error occurred, and a TomlErrorKind which describes the type of error at that span.

TomlGetError is an error from one of the get_<type> methods in tables. It occurs when there's no value for the provided key (InvalidKey) or when the types aren't the same (TypeMismatch - could happen if, for example, you try to get a String value with get_table). A TypeMismatch error stores the actual TOML value and its type, so you can attempt to still use it if possible.

Status/To-Do

BOML can parse everything in TOML except for the date/time/date-time types. Its original goal was just to parse Rust config files, like Cargo.toml, for bargo.

BOML also may parse what is technically invalid TOML as valid TOML. It's current goal is to just parse TOML, so extra cases that are technically not valid TOML may not get caught.

You can test BOML against the official TOML test suite by running the toml_test test (cargo t toml_test -- --show-output). This test currently skips tests for the time types, and only prints warnings (instead of failing) if an invalid test passes. It also skips tests that don't have a valid UTF-8 encoding (since Rust strings require UTF-8). With those exceptions in place, BOML is able to pass the toml-test suite.

Why "(almost) zero-copy"?

TOML has 2 kinds of strings: basic strings, and literal strings. Literal strings are just strings BOML can read from the file, but basic strings can have escapes (\n, for example, gets replaced with the newline character). Processing these escapes requires copying the string, and then replacing the escapes with their actual characters.

BOML will only copy and format a string if the string is a basic string (surrounded by ") and actually contains escapes. Literal strings (surrounded by ') and basic strings without escapes are not copied.

Whatsitstandfor

Yes.

No runtime deps