orchid/orchid-base/src/name.rs

//! Various datatypes that all represent namespaced names.

use std::borrow::Borrow;
use std::hash::Hash;
use std::num::{NonZeroU64, NonZeroUsize};
use std::ops::{Deref, Index};
use std::path::Path;
use std::{fmt, vec};

use futures::future::{OptionFuture, join_all};
use itertools::Itertools;
use trait_set::trait_set;

use crate::api;
use crate::interner::{IStr, IStrv, es, ev, is, iv};

trait_set! {
	/// Traits that all name iterators should implement
	pub trait NameIter = Iterator<Item = IStr> + DoubleEndedIterator + ExactSizeIterator;
}

/// A token path which may be empty. [VName] is the non-empty version
#[derive(Clone, Default, Hash, PartialEq, Eq)]
pub struct VPath(Vec<IStr>);
impl VPath {
	/// Collect segments into a vector
	pub fn new(items: impl IntoIterator<Item = IStr>) -> Self { Self(items.into_iter().collect()) }
	/// Number of path segments
	pub fn len(&self) -> usize { self.0.len() }
	/// Whether there are any path segments. In other words, whether this is a
	/// valid name
	pub fn is_empty(&self) -> bool { self.len() == 0 }
	/// Prepend some tokens to the path
	pub fn prefix(self, items: impl IntoIterator<Item = IStr>) -> Self {
		Self(items.into_iter().chain(self.0).collect())
	}
	/// Append some tokens to the path
	pub fn suffix(self, items: impl IntoIterator<Item = IStr>) -> Self {
		Self(self.0.into_iter().chain(items).collect())
	}
	/// Partition the string by `::` namespace separators
	pub async fn parse(s: &str) -> Self {
		Self(if s.is_empty() { vec![] } else { join_all(s.split("::").map(is)).await })
	}
	/// Walk over the segments
	pub fn str_iter(&self) -> impl Iterator<Item = &'_ str> { Box::new(self.0.iter().map(|s| &**s)) }
	/// Try to convert into non-empty version
	pub fn into_name(self) -> Result<VName, EmptyNameError> { VName::new(self.0) }
	/// Add a token to the path. Since now we know that it can't be empty, turn it
	/// into a name.
	pub fn name_with_suffix(self, name: IStr) -> VName {
		VName(self.into_iter().chain([name]).collect())
	}
	/// Add a token to the beginning of the. Since now we know that it can't be
	/// empty, turn it into a name.
	pub fn name_with_prefix(self, name: IStr) -> VName {
		VName([name].into_iter().chain(self).collect())
	}

	/// Convert a fs path to a vpath
	pub async fn from_path(path: &Path, ext: &str) -> Option<(Self, bool)> {
		async fn to_vpath(p: &Path) -> Option<VPath> {
			let tok_opt_v = join_all(p.iter().map(|c| OptionFuture::from(c.to_str().map(is)))).await;
			tok_opt_v.into_iter().collect::<Option<_>>().map(VPath)
		}
		match path.extension().map(|s| s.to_str()) {
			Some(Some(s)) if s == ext => Some((to_vpath(&path.with_extension("")).await?, true)),
			None => Some((to_vpath(path).await?, false)),
			Some(_) => None,
		}
	}
}
impl fmt::Debug for VPath {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "VName({self})") }
}
impl fmt::Display for VPath {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
		write!(f, "{}", self.str_iter().join("::"))
	}
}
impl FromIterator<IStr> for VPath {
	fn from_iter<T: IntoIterator<Item = IStr>>(iter: T) -> Self { Self(iter.into_iter().collect()) }
}
impl IntoIterator for VPath {
	type Item = IStr;
	type IntoIter = vec::IntoIter<Self::Item>;
	fn into_iter(self) -> Self::IntoIter { self.0.into_iter() }
}
impl Borrow<[IStr]> for VPath {
	fn borrow(&self) -> &[IStr] { &self.0[..] }
}
impl Deref for VPath {
	type Target = [IStr];
	fn deref(&self) -> &Self::Target { self.borrow() }
}

impl<T> Index<T> for VPath
where [IStr]: Index<T>
{
	type Output = <[IStr] as Index<T>>::Output;

	fn index(&self, index: T) -> &Self::Output { &Borrow::<[IStr]>::borrow(self)[index] }
}

/// A mutable representation of a namespaced identifier of at least one segment.
///
/// These names may be relative or otherwise partially processed.
///
/// See also [Sym] for the immutable representation, and [VPath] for possibly
/// empty values
#[derive(Clone, Hash, PartialEq, Eq)]
pub struct VName(Vec<IStr>);
impl VName {
	/// Assert that the sequence isn't empty and wrap it in [VName] to represent
	/// this invariant
	pub fn new(items: impl IntoIterator<Item = IStr>) -> Result<Self, EmptyNameError> {
		let data: Vec<_> = items.into_iter().collect();
		if data.is_empty() { Err(EmptyNameError) } else { Ok(Self(data)) }
	}
	pub async fn deintern(name: impl IntoIterator<Item = api::TStr>) -> Result<Self, EmptyNameError> {
		Self::new(join_all(name.into_iter().map(es)).await)
	}
	/// Unwrap the enclosed vector
	pub fn into_vec(self) -> Vec<IStr> { self.0 }
	/// Get a reference to the enclosed vector
	pub fn vec(&self) -> &Vec<IStr> { &self.0 }
	/// Mutable access to the underlying vector. To ensure correct results, this
	/// must never be empty.
	pub fn vec_mut(&mut self) -> &mut Vec<IStr> { &mut self.0 }
	/// Intern the name and return a [Sym]
	pub async fn to_sym(&self) -> Sym { Sym(iv(&self.0[..]).await) }
	/// If this name has only one segment, return it
	pub fn as_root(&self) -> Option<IStr> { self.0.iter().exactly_one().ok().cloned() }
	/// Prepend the segments to this name
	#[must_use = "This is a pure function"]
	pub fn prefix(self, items: impl IntoIterator<Item = IStr>) -> Self {
		Self(items.into_iter().chain(self.0).collect())
	}
	/// Append the segments to this name
	#[must_use = "This is a pure function"]
	pub fn suffix(self, items: impl IntoIterator<Item = IStr>) -> Self {
		Self(self.0.into_iter().chain(items).collect())
	}
	/// Read a `::` separated namespaced name
	pub async fn parse(s: &str) -> Result<Self, EmptyNameError> { Self::new(VPath::parse(s).await) }
	pub async fn literal(s: &'static str) -> Self { Self::parse(s).await.expect("empty literal !?") }
	/// Obtain an iterator over the segments of the name
	pub fn iter(&self) -> impl Iterator<Item = IStr> + '_ { self.0.iter().cloned() }
}
impl fmt::Debug for VName {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "VName({self})") }
}
impl fmt::Display for VName {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
		write!(f, "{}", self.str_iter().join("::"))
	}
}
impl IntoIterator for VName {
	type Item = IStr;
	type IntoIter = vec::IntoIter<Self::Item>;
	fn into_iter(self) -> Self::IntoIter { self.0.into_iter() }
}
impl<T> Index<T> for VName
where [IStr]: Index<T>
{
	type Output = <[IStr] as Index<T>>::Output;

	fn index(&self, index: T) -> &Self::Output { &self.deref()[index] }
}
impl Borrow<[IStr]> for VName {
	fn borrow(&self) -> &[IStr] { self.0.borrow() }
}
impl Deref for VName {
	type Target = [IStr];
	fn deref(&self) -> &Self::Target { self.borrow() }
}

/// Error produced when a non-empty name [VName] or [Sym] is constructed with an
/// empty sequence
#[derive(Debug, Copy, Clone, Default, Hash, PartialEq, Eq, PartialOrd, Ord)]
pub struct EmptyNameError;
impl TryFrom<&[IStr]> for VName {
	type Error = EmptyNameError;
	fn try_from(value: &[IStr]) -> Result<Self, Self::Error> { Self::new(value.iter().cloned()) }
}

/// An interned representation of a namespaced identifier.
///
/// These names are always absolute.
///
/// See also [VName]
#[derive(Clone, Hash, PartialEq, Eq)]
pub struct Sym(IStrv);
impl Sym {
	/// Assert that the sequence isn't empty, intern it and wrap it in a [Sym] to
	/// represent this invariant
	pub async fn new(v: impl IntoIterator<Item = IStr>) -> Result<Self, EmptyNameError> {
		let items = v.into_iter().collect_vec();
		Self::from_tok(iv(&items).await)
	}
	/// Read a `::` separated namespaced name. Do not use this for statically
	/// known names, use the [sym] macro instead which is cached.
	pub async fn parse(s: &str) -> Result<Self, EmptyNameError> {
		Ok(Sym(iv(&VName::parse(s).await?.into_vec()).await))
	}
	/// Assert that a token isn't empty, and wrap it in a [Sym]
	pub fn from_tok(t: IStrv) -> Result<Self, EmptyNameError> {
		if t.is_empty() { Err(EmptyNameError) } else { Ok(Self(t)) }
	}
	/// Grab the interner token
	pub fn tok(&self) -> IStrv { self.0.clone() }
	/// Get a number unique to this name suitable for arbitrary ordering.
	pub fn id(&self) -> NonZeroU64 { self.0.to_api().0 }
	/// Extern the sym for editing
	pub fn to_vname(&self) -> VName { VName(self[..].to_vec()) }
	pub async fn from_api(marker: api::TStrv) -> Sym {
		Self::from_tok(ev(marker).await).expect("Empty sequence found for serialized Sym")
	}
	pub fn to_api(&self) -> api::TStrv { self.tok().to_api() }
	pub async fn suffix(&self, tokv: impl IntoIterator<Item = IStr>) -> Sym {
		Self::new(self.0.iter().cloned().chain(tokv)).await.unwrap()
	}
}
impl fmt::Debug for Sym {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "Sym({self})") }
}
impl fmt::Display for Sym {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
		write!(f, "{}", self.str_iter().join("::"))
	}
}
impl<T> Index<T> for Sym
where [IStr]: Index<T>
{
	type Output = <[IStr] as Index<T>>::Output;

	fn index(&self, index: T) -> &Self::Output { &self.deref()[index] }
}
impl Borrow<[IStr]> for Sym {
	fn borrow(&self) -> &[IStr] { &self.0[..] }
}
impl Deref for Sym {
	type Target = [IStr];
	fn deref(&self) -> &Self::Target { self.borrow() }
}

/// An abstraction over tokenized vs non-tokenized names so that they can be
/// handled together in datastructures. The names can never be empty
#[allow(clippy::len_without_is_empty)] // never empty
pub trait NameLike:
	'static + Clone + Eq + Hash + fmt::Debug + fmt::Display + Borrow<[IStr]>
{
	/// Convert into held slice
	fn as_slice(&self) -> &[IStr] { Borrow::<[IStr]>::borrow(self) }
	/// Get iterator over tokens
	fn segs(&self) -> impl NameIter + '_ { self.as_slice().iter().cloned() }
	/// Get iterator over string segments
	fn str_iter(&self) -> impl Iterator<Item = &'_ str> + '_ { self.as_slice().iter().map(|t| &**t) }
	/// Fully resolve the name for printing
	#[must_use]
	fn to_strv(&self) -> Vec<String> { self.segs().map(|s| s.to_string()).collect() }
	/// Format the name as an approximate filename
	fn as_src_path(&self) -> String { format!("{}.orc", self.segs().join("/")) }
	/// Return the number of segments in the name
	fn len_nz(&self) -> NonZeroUsize {
		NonZeroUsize::try_from(self.segs().count()).expect("NameLike never empty")
	}
	/// Like slice's `split_first` except we know that it always returns Some
	fn split_first_seg(&self) -> (IStr, &[IStr]) {
		let (foot, torso) = self.as_slice().split_last().expect("NameLike never empty");
		(foot.clone(), torso)
	}
	/// Like slice's `split_last` except we know that it always returns Some
	fn split_last_seg(&self) -> (IStr, &[IStr]) {
		let (foot, torso) = self.as_slice().split_last().expect("NameLike never empty");
		(foot.clone(), torso)
	}
	/// Get the first element
	fn first_seg(&self) -> IStr { self.split_first_seg().0 }
	/// Get the last element
	fn last_seg(&self) -> IStr { self.split_last_seg().0 }
}

impl NameLike for Sym {}
impl NameLike for VName {}

/// Create a [Sym] literal.
///
/// The name and its components will be cached in a thread-local static so
/// that subsequent executions of the expression only incur an Arc-clone for
/// cloning the token.
#[macro_export]
macro_rules! sym {
  ($seg1:tt $( :: $seg:tt)*) => {
		$crate::tl_cache!(async $crate::name::Sym : {
			$crate::name::Sym::from_tok(
				$crate::interner::iv(&[
					$crate::interner::is($crate::sym!(@SEG $seg1)).await
					$( , $crate::interner::is($crate::sym!(@SEG $seg)).await )*
				])
				.await
			).unwrap()
		})
  };
	(@SEG [ $($data:tt)* ]) => {
		stringify!($($data)*)
	};
	(@SEG $data:tt) => {
		stringify!($data)
	};
}

/// Create a [VName] literal.
///
/// The components are interned much like in [sym].
#[macro_export]
macro_rules! vname {
  ($seg1:tt $( :: $seg:tt)*) => {
		$crate::tl_cache!(async $crate::name::VName : {
			$crate::name::VName::new([
				$crate::interner::is(stringify!($seg1)).await
				$( , $crate::interner::is(stringify!($seg)).await )*
			]).unwrap()
		})
	};
}

/// Create a [VPath] literal.
///
/// The components are interned much like in [sym].
#[macro_export]
macro_rules! vpath {
  ($seg1:tt $( :: $seg:tt)*) => {
		$crate::tl_cache!(async $crate::name::VPath : {
			$crate::name::VPath(vec![
				$crate::interner::is(stringify!($seg1)).await
				$( , $crate::interner::is(stringify!($seg)).await )*
			])
		})
	};
  () => {
    $crate::name::VPath(vec![])
  }
}

#[cfg(test)]
pub mod test {
	use std::borrow::Borrow;

	use orchid_api_traits::spin_on;

	use super::{NameLike, Sym, VName};
	use crate::interner::local_interner::local_interner;
	use crate::interner::{IStr, is, with_interner};
	use crate::name::VPath;

	#[test]
	pub fn recur() {
		spin_on(with_interner(local_interner(), async {
			let myname = vname!(foo::bar);
			let _borrowed_slice: &[IStr] = myname.borrow();
			let _deref_pathslice: &[IStr] = &myname;
			let _as_slice_out: &[IStr] = myname.as_slice();
		}))
	}

	/// Tests that literals are correctly interned as equal
	#[test]
	pub fn literals() {
		spin_on(with_interner(local_interner(), async {
			assert_eq!(
				sym!(foo::bar::baz),
				Sym::new([is("foo").await, is("bar").await, is("baz").await]).await.unwrap()
			);
			assert_eq!(
				sym!(foo::bar::[|>]),
				Sym::new([is("foo").await, is("bar").await, is("|>").await]).await.unwrap()
			);
			assert_eq!(
				vname!(foo::bar::baz),
				VName::new([is("foo").await, is("bar").await, is("baz").await]).unwrap()
			);
			assert_eq!(
				{ vpath!(foo::bar::baz) },
				VPath::new([is("foo").await, is("bar").await, is("baz").await])
			);
		}))
	}
}