path: root/src/lib.rs

#![cfg_attr(not(feature = "std"), no_std)]

extern crate alloc;
#[cfg(not(feature = "std"))]
use alloc::collections::BTreeMap as Map;
use alloc::sync::Arc;
use core::fmt;
#[cfg(feature = "std")]
use std::collections::HashMap as Map;

/// File and line information
#[derive(Clone, Debug)]
pub struct Location {
	file: Arc<str>,
	line: u64,
}

impl fmt::Display for Location {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
		write!(f, "{}:{}", self.file, self.line)
	}
}

/// A string value, together with location information about where it is defined.
#[derive(Clone, Debug)]
struct Value {
	value: Box<str>,
	defined_at: Location,
}

#[derive(Clone, Debug, Default)]
pub struct Configuration {
	// wrap in an Arc for cheap cloning
	children: Arc<Map<Box<str>, Configuration>>,
	values: Arc<Map<Box<str>, Value>>,
}

#[non_exhaustive]
#[derive(Debug)]
pub enum Error {
	#[cfg(feature = "std")]
	IO(Box<str>, std::io::Error),
	IllegalCharacter(Location, char),
	InvalidUtf8(Location),
	BadInt(Location, Box<str>),
	BadUInt(Location, Box<str>),
	BadFloat(Location, Box<str>),
	BadBool(Location, Box<str>),
	UnmatchedLeftBrace(Location),
	InvalidKey(Location, Box<str>),
	InvalidValue(Location),
	InvalidLine(Location),
	StrayCharsAfterQuotedString(Location),
	UnterminatedString(Location, char),
}

impl fmt::Display for Error {
	fn fmt(&self, _f: &mut fmt::Formatter<'_>) -> fmt::Result {
		todo!()
	}
}

pub type Result<T> = std::result::Result<T, Error>;

fn parse_int(location: &Location, string: &str) -> Result<i64> {
	let bad_int = || Error::BadInt(location.clone(), string.into());
	if !string
		.bytes()
		.all(|c| c.is_ascii_hexdigit() || c == b'x' || c == b'X' || c == b'-' || c == b'+')
	{
		return Err(bad_int());
	}
	let signless = string.strip_prefix(['-', '+']).unwrap_or(string);
	let mut base = 10;
	let baseless = signless
		.strip_prefix("0x")
		.or_else(|| signless.strip_prefix("0X"))
		.unwrap_or_else(|| {
			base = 16;
			signless
		});
	for digit in baseless.bytes() {
		if base == 10 && !digit.is_ascii_digit() {
			return Err(bad_int());
		}
	}
	string.parse().map_err(|_| bad_int())
}

fn parse_uint(location: &Location, string: &str) -> Result<u64> {
	let bad_uint = || Error::BadUInt(location.clone(), string.into());
	if !string
		.bytes()
		.all(|c| c.is_ascii_hexdigit() || c == b'x' || c == b'X' || c == b'+')
	{
		return Err(bad_uint());
	}
	let signless = string.strip_prefix('+').unwrap_or(string);
	let mut base = 10;
	let baseless = signless
		.strip_prefix("0x")
		.or_else(|| signless.strip_prefix("0X"))
		.unwrap_or_else(|| {
			base = 16;
			signless
		});
	for digit in baseless.bytes() {
		if base == 10 && !digit.is_ascii_digit() {
			return Err(bad_uint());
		}
	}
	let val = signless.parse().map_err(|_| bad_uint())?;
	if val > i64::MAX as u64 {
		return Err(bad_uint());
	}
	Ok(val)
}

fn parse_float(location: &Location, string: &str) -> Result<f64> {
	let bad_float = || Error::BadFloat(location.clone(), string.into());
	if !string.bytes().all(|c| {
		c.is_ascii_digit() || c == b'.' || c == b'+' || c == b'-' || c == b'e' || c == b'E'
	}) {
		return Err(bad_float());
	}
	string.parse().map_err(|_| bad_float())
}

fn parse_bool(location: &Location, string: &str) -> Result<bool> {
	match string {
		"yes" | "on" | "true" => Ok(true),
		"no" | "off" | "false" => Ok(false),
		_ => Err(Error::BadBool(location.clone(), string.into())),
	}
}

fn parse_list(_location: &Location, _string: &str) -> Vec<String> {
	todo!()
}

/// Trait for reading configurations.
///
/// Ordinarily you won't need to implement this trait, since it is
/// already implemented by any `T: `[`std::io::BufRead`] (or else `&str` and `&[u8]`,
/// if the `std` feature is not enabled).
pub trait Read {
	/// Read up to the next line feed (or EOF), not including the line feed itself.
	///
	/// Puts the line in `line` and returns `Ok(true)`, if the end of file has been reached,
	/// `line` is unmodified and `Ok(false)` is returned.
	///
	/// You don't need to check for valid UTF-8 here — that is already done in the code which uses
	/// this trait.
	fn read_until_lf(&mut self, line: &mut Vec<u8>) -> Result<bool>;
}

#[cfg(feature = "std")]
impl<R: std::io::BufRead> Read for R {
	fn read_until_lf(&mut self, line: &mut Vec<u8>) -> Result<bool> {
		self.read_until(b'\n', line)
			.map_err(|e| Error::IO("read error".into(), e))?;
		if line.ends_with(b"\n") {
			line.pop();
			Ok(true)
		} else {
			Ok(!line.is_empty())
		}
	}
}

#[cfg(not(feature = "std"))]
impl Read for &str {
	fn read_until_lf(&mut self, line: &mut Vec<u8>) -> Result<bool> {
		match self.split_once('\n') {
			Some((pre, post)) => {
				*self = post;
				line.extend_from_slice(pre.as_bytes());
				Ok(true)
			}
			None => {
				if self.is_empty() {
					return Ok(false);
				}
				line.extend_from_slice(self.as_bytes());
				*self = "";
				Ok(true)
			}
		}
	}
}

#[cfg(not(feature = "std"))]
impl Read for &[u8] {
	fn read_until_lf(&mut self, line: &mut Vec<u8>) -> Result<bool> {
		match self.iter().position(|&c| c == b'\n') {
			Some(i) => {
				line.extend_from_slice(&self[..i]);
				*self = &self[i + 1..];
				Ok(true)
			}
			None => {
				if self.is_empty() {
					return Ok(false);
				}
				line.extend_from_slice(self);
				*self = b"";
				Ok(true)
			}
		}
	}
}

fn check_valid_key(location: &Location, s: &str) -> Result<()> {
	if s.bytes().all(|c| {
		c >= 0x80 || c.is_ascii_alphanumeric() || matches!(c, b'.' | b'_' | b'/' | b'*' | b'-')
	}) {
		Ok(())
	} else {
		Err(Error::InvalidKey(location.clone(), s.into()))
	}
}

/// Returns (unquoted value, new line number)
fn read_quoted_value(
	quoted: &str,
	reader: &mut dyn Read,
	start_location: &Location,
) -> Result<(String, u64)> {
	let delimiter: char = quoted.chars().next().unwrap();
	let mut unquoted = String::new();
	let mut line_number = start_location.line;
	let location = |line_number: u64| Location {
		file: start_location.file.clone(),
		line: line_number,
	};
	let mut line_buf = vec![];
	let mut first = true;
	loop {
		let line = if first {
			first = false;
			quoted
		} else {
			line_buf.truncate(0);
			if !reader.read_until_lf(&mut line_buf)? {
				break;
			}
			line_number += 1;
			line_buf.pop_if(|c| *c == b'\r');
			str::from_utf8(&line_buf).map_err(|_| Error::InvalidUtf8(location(line_number)))?
		};
		let mut chars = line.chars();
		while let Some(c) = chars.next() {
			if c == delimiter {
				if !chars.all(|c| c == ' ' || c == '\t') {
					return Err(Error::StrayCharsAfterQuotedString(location(line_number)));
				}
				return Ok((unquoted, line_number));
			} else if c == '\\' {
				todo!() // parse escape sequence
			} else if c == '\0' {
				return Err(Error::InvalidValue(location(line_number)));
			} else {
				unquoted.push(c);
			}
		}
	}
	Err(Error::UnterminatedString(start_location.clone(), delimiter))
}

impl Configuration {
	fn load_dyn(filename: &str, reader: &mut dyn Read) -> Result<Self> {
		let mut config = Configuration::default();
		let mut line = vec![];
		let mut line_number = 0;
		let mut current_section = &mut config;
		let filename: Arc<str> = filename.into();
		loop {
			line.truncate(0);
			if !reader.read_until_lf(&mut line)? {
				break;
			}
			line_number += 1;
			let location = Location {
				file: filename.clone(),
				line: line_number,
			};
			line.pop_if(|c| *c == b'\r');
			for c in &line {
				if (0..0x1f).contains(c) && *c != b'\t' {
					return Err(Error::IllegalCharacter(location, char::from(*c)));
				}
			}
			let mut line =
				str::from_utf8(&line).map_err(|_| Error::InvalidUtf8(location.clone()))?;
			line = line.trim_start_matches(['\t', ' ']);
			if line.is_empty() || line.starts_with('#') {
				// comment/blank line
				continue;
			}
			if line.starts_with('[') {
				line = line.trim_end_matches(['\t', ' ']);
				if !line.ends_with(']') {
					return Err(Error::UnmatchedLeftBrace(location));
				}
				let new_section = line[1..line.len() - 1].into();
				current_section = &mut config;
				check_valid_key(&location, new_section)?;
				if !new_section.is_empty() {
					for component in new_section.split('.') {
						current_section = Arc::get_mut(&mut current_section.children)
							.unwrap()
							.entry(component.into())
							.or_default();
					}
				}
			} else {
				let (mut relative_key, mut value) = line
					.split_once('=')
					.ok_or_else(|| Error::InvalidLine(location.clone()))?;
				check_valid_key(&location, relative_key)?;
				relative_key = relative_key.trim_end_matches(['\t', ' ']);
				value = value.trim_start_matches(['\t', ' ']);

				fn insert(
					mut section: &mut Configuration,
					location: Location,
					mut key: &str,
					value: &str,
				) {
					if let Some(last_dot) = key.rfind('.') {
						for component in key[..last_dot].split('.') {
							section = Arc::get_mut(&mut section.children)
								.unwrap()
								.entry(component.into())
								.or_default()
						}
						key = &key[last_dot + 1..];
					}
					Arc::get_mut(&mut section.values).unwrap().insert(
						key.into(),
						Value {
							value: value.into(),
							defined_at: location,
						},
					);
				}
				if value.starts_with(['`', '"', '\'']) {
					let (value, new_line_number) = read_quoted_value(value, reader, &location)?;
					insert(current_section, location, relative_key, &value);
					line_number = new_line_number;
				} else {
					value = value.trim_end_matches(['\t', ' ']);
					if value.contains('\0') {
						return Err(Error::InvalidValue(location));
					}
					insert(current_section, location, relative_key, value);
				}
			}
		}
		Ok(config)
	}
	pub fn load<R: Read>(filename: &str, mut reader: R) -> Result<Self> {
		// avoid big code size by using dyn reference.
		// the impact on performance is not really important.
		Configuration::load_dyn(filename, &mut reader)
	}
	#[cfg(feature = "std")]
	pub fn load_path<P: AsRef<std::path::Path>>(path: P) -> Result<Self> {
		let p = path.as_ref();
		let filename = p.to_string_lossy();
		let file = std::fs::File::open(p).map_err(|e| Error::IO(filename.clone().into(), e))?;
		Configuration::load(&filename, std::io::BufReader::new(file))
	}
	pub fn section(&self, key: &str) -> Configuration {
		let mut node = self;
		for component in key.split('.') {
			node = match self.children.get(component) {
				Some(x) => x,
				None => return Configuration::default(),
			};
		}
		node.clone()
	}
	fn get_val(&self, key: &str) -> Option<&Value> {
		let Some(last_dot) = key.rfind('.') else {
			return self.values.get(key);
		};
		let (path, last_component) = key.split_at(last_dot);
		let mut node = self;
		for component in path.split('.') {
			node = self.children.get(component)?;
		}
		node.values.get(last_component)
	}
	pub fn get(&self, key: &str) -> Option<&str> {
		Some(self.get_val(key)?.value.as_ref())
	}
	pub fn location(&self, key: &str) -> Option<Location> {
		Some(self.get_val(key)?.defined_at.clone())
	}
	pub fn has(&self, key: &str) -> bool {
		self.get(key).is_some()
	}
	pub fn get_or_default<'a>(&'a self, key: &str, default: &'a str) -> &'a str {
		self.get(key).unwrap_or(default)
	}
	pub fn get_int(&self, key: &str) -> Option<Result<i64>> {
		let Value { value, defined_at } = self.get_val(key)?;
		Some(parse_int(defined_at, value.as_ref()))
	}
	pub fn get_int_or_default(&self, key: &str, default: i64) -> Result<i64> {
		self.get_int(key).unwrap_or(Ok(default))
	}
	pub fn get_uint(&self, key: &str) -> Option<Result<u64>> {
		let Value { value, defined_at } = self.get_val(key)?;
		Some(parse_uint(defined_at, value.as_ref()))
	}
	pub fn get_uint_or_default(&self, key: &str, default: u64) -> Result<u64> {
		self.get_uint(key).unwrap_or(Ok(default))
	}
	pub fn get_float(&self, key: &str) -> Option<Result<f64>> {
		let Value { value, defined_at } = self.get_val(key)?;
		Some(parse_float(defined_at, value.as_ref()))
	}
	pub fn get_float_or_default(&self, key: &str, default: f64) -> Result<f64> {
		self.get_float(key).unwrap_or(Ok(default))
	}
	pub fn get_bool(&self, key: &str) -> Option<Result<bool>> {
		let Value { value, defined_at } = self.get_val(key)?;
		Some(parse_bool(defined_at, value.as_ref()))
	}
	pub fn get_bool_or_default(&self, key: &str, default: bool) -> Result<bool> {
		self.get_bool(key).unwrap_or(Ok(default))
	}
	pub fn get_list(&self, key: &str) -> Option<Vec<String>> {
		let Value { value, defined_at } = self.get_val(key)?;
		Some(parse_list(defined_at, value.as_ref()))
	}
	pub fn get_list_or_default(
		&self,
		key: &str,
		default: impl FnOnce() -> Vec<String>,
	) -> Vec<String> {
		self.get_list(key).unwrap_or_else(default)
	}
	/// Merge `conf` into `self`, preferring values in `conf`.
	pub fn merge(&mut self, conf: &Configuration) {
		let new_values = Arc::make_mut(&mut self.values);
		// merge conf.values into self.values
		for (key, val) in conf.values.iter() {
			new_values.insert(key.clone(), val.clone());
		}
		// merge conf.children into self.children
		let new_children = Arc::make_mut(&mut self.children);
		for (key, child) in conf.children.iter() {
			new_children.entry(key.clone()).or_default().merge(child)
		}
	}
}