Add BitSet

doc/set.md

@@ -0,0 +1,12 @@
+# Automatically-Managed Index Set
+
+This module defines the [`BitSet`] collection as a useful wrapper over a
+[`BitVec`].
+
+A `BitVec` is a very efficient way of storing a set of [`usize`] values since
+the various set operations can be easily represented using bit operations.
+However, a `BitVec` is less ergonomic than a `BitSet` because of the need to
+resize when inserting elements larger than any already in the set.
+
+[`BitSet`]: crate::set::BitSet
+[`BitVec`]: crate::vec::BitVec

doc/set/BitSet.md

@@ -0,0 +1,70 @@
+# Packed-Bits Set
+
+This is a data structure that consists of an automatically managed [`BitVec`]
+which stores a set of `usize` values as `true` bits in the `BitVec`.
+
+The main benefit of this structure is the automatic handling of the memory
+backing the [`BitVec`], which must be resized to account for the sizes of data
+inside it. If you know the bounds of your data ahead of time, you may prefer to
+use a regular [`BitVec`] or even a [`BitArray`] instead, the latter of which
+will be allocated on the stack instead of the heap.
+
+## Documentation Practices
+
+`BitSet` attempts to replicate the API of the standard-library `BTreeSet` type,
+including inherent methods, trait implementations, and relationships with the
+[`BitSet`] analogue.
+
+Items that are either direct ports, or renamed variants, of standard-library
+APIs will have a `## Original` section that links to their standard-library
+documentation. Items that map to standard-library APIs but have a different API
+signature will also have an `## API Differences` section that describes what
+the difference is, why it exists, and how to transform your code to fit it. For
+example:
+
+## Original
+
+[`BTreeSet<T>`](alloc::collections::BTreeSet)
+
+## API Differences
+
+As with all `bitvec` data structures, this takes two type parameters `<T, O>`
+that govern the bit-vector’s storage representation in the underlying memory,
+and does *not* take a type parameter to govern what data type it stores (always
+`usize`)
+
+### Accessing the internal [`BitVec`]
+
+Since `BitSet` is merely an API over the internal `BitVec`, you can freely
+take ownership of the internal buffer or borrow the buffer as a `BitSlice`.
+
+However, since would be inconsistent with the set-style API, these require
+dedicated methods instead of simple deref:
+
+```rust
+use bitvec::prelude::*;
+use bitvec::set::BitSet;
+
+fn mutate_bitvec(vec: &mut BitVec) {
+  // …
+}
+
+fn read_bitslice(bits: &BitSlice) {
+  // …
+}
+
+let mut bs: BitSet = BitSet::new();
+bs.insert(10);
+bs.insert(20);
+bs.insert(30);
+read_bitslice(bs.as_bitslice());
+mutate_bitvec(bs.as_mut_bitvec());
+```
+
+Since a `BitSet` requires no additional invariants over `BitVec`, any mutations
+to the internal vec are allowed without restrictions. For more details on the
+safety guarantees of [`BitVec`], see its specific documentation.
+
+[`BitArray`]: crate::array::BitArray
+[`BitSet`]: crate::set::BitSet
+[`BitVec`]: crate::vec::BitVec

doc/set/iter.md

@@ -0,0 +1,14 @@
+# Bit-Set Iteration
+
+This module provides iteration protocols for `BitSet`, including:
+
+- extension of existing bit-sets with new data
+- collection of data into new bit-sets
+- iteration over the contents of a bit-sets
+
+`BitSet` implements `Extend` and `FromIterator` for sources of `usize`.
+
+Since the implementation is the same for sets, the [`IterOnes`] iterator from
+the `slice` module is used for the set iterator instead of a wrapper.
+
+[`IterOnes`]: crate::slice::IterOnes

doc/set/iter/Range.md

@@ -0,0 +1,33 @@
+# Bit-Set Range Iteration
+
+This view iterates over the elements in a bit-set within a given range. It is
+created by the [`BitSet::range`] method.
+
+## Original
+
+[`btree_map::Range`](alloc::collections::btree_map::Range)
+
+## API Differences
+
+Since the `usize` are not physically stored in the set, this yields `usize`
+values instead of references.
+
+## Examples
+
+```rust
+use bitvec::prelude::*;
+use bitvec::set::BitSet;
+
+let mut bs: BitSet = BitSet::new();
+bs.insert(1);
+bs.insert(2);
+bs.insert(3);
+bs.insert(4);
+for val in bs.range(2..6) {
+  # #[cfg(feature = "std")] {
+  println!("{val}");
+  # }
+}
+```
+
+[`BitSet::range`]: crate::set::BitSet::range

rustfmt-stable.toml

@@ -13,7 +13,7 @@
 # attr_fn_like_width = 70            # Leave implicit
 # chain_width = 60                   # Leave implicit
 edition = "2018"
-fn_args_layout = "Tall"
+fn_params_layout = "Tall"
 # fn_call_width = 60                 # Leave implicit
 force_explicit_abi = true
 hard_tabs = true

rustfmt.toml

@@ -14,7 +14,7 @@
 # attr_fn_like_width = 70            # Leave implicit
 # chain_width = 60                   # Leave implicit
 edition = "2018"
-fn_args_layout = "Tall"
+fn_params_layout = "Tall"
 # fn_call_width = 60                 # Leave implicit
 force_explicit_abi = true
 hard_tabs = true
@@ -69,7 +69,7 @@ normalize_comments = false
 normalize_doc_attributes = false
 overflow_delimited_expr = true
 reorder_impl_items = true
-required_version = "1.5.1"
+required_version = "1.6.0"
 skip_children = false
 space_after_colon = true
 space_before_colon = false

src/lib.rs

@@ -34,6 +34,7 @@ pub mod mem;
 pub mod order;
 pub mod ptr;
 mod serdes;
+pub mod set;
 pub mod slice;
 pub mod store;
 pub mod vec;

src/set.rs

@@ -0,0 +1,188 @@
+#![doc = include_str!("../doc/set.md")]
+#![cfg(feature = "alloc")]
+
+#[cfg(not(feature = "std"))]
+use alloc::vec;
+use core::ops;
+
+use wyz::comu::{
+	Const,
+	Mut,
+};
+
+use crate::{
+	boxed::BitBox,
+	order::{
+		BitOrder,
+		Lsb0,
+	},
+	ptr::BitPtr,
+	slice::BitSlice,
+	store::BitStore,
+	vec::BitVec,
+};
+
+mod api;
+mod iter;
+mod traits;
+
+pub use iter::Range;
+
+#[repr(transparent)]
+#[doc = include_str!("../doc/set/BitSet.md")]
+pub struct BitSet<T = usize, O = Lsb0>
+where
+	T: BitStore,
+	O: BitOrder,
+{
+	inner: BitVec<T, O>,
+}
+
+/// Constructors.
+impl<T, O> BitSet<T, O>
+where
+	T: BitStore,
+	O: BitOrder,
+{
+	/// An empty bit-set with no backing allocation.
+	pub const EMPTY: Self = Self {
+		inner: BitVec::EMPTY,
+	};
+
+	/// Creates a new bit-set for a range of indices.
+	#[inline]
+	pub fn from_range(range: ops::Range<usize>) -> Self {
+		let mut inner = BitVec::with_capacity(range.end);
+		unsafe {
+			inner.set_len(range.end);
+			inner[.. range.start].fill(false);
+			inner[range.start ..].fill(true);
+		}
+		Self { inner }
+	}
+
+	/// Constructs a new bit-set from an existing bit-vec.
+	#[inline]
+	pub fn from_bitvec(inner: BitVec<T, O>) -> Self {
+		Self { inner }
+	}
+}
+
+/// Converters.
+impl<T, O> BitSet<T, O>
+where
+	T: BitStore,
+	O: BitOrder,
+{
+	/// Explicitly views the bit-set as a bit-slice.
+	#[inline]
+	pub fn as_bitslice(&self) -> &BitSlice<T, O> {
+		self.inner.as_bitslice()
+	}
+
+	/// Explicitly views the bit-set as a mutable bit-slice.
+	#[inline]
+	pub fn as_mut_bitslice(&mut self) -> &mut BitSlice<T, O> {
+		self.inner.as_mut_bitslice()
+	}
+
+	/// Explicitly views the bit-set as a bit-vec.
+	#[inline]
+	pub fn as_bitvec(&self) -> &BitVec<T, O> {
+		&self.inner
+	}
+
+	/// Explicitly views the bit-set as a mutable bit-vec.
+	#[inline]
+	pub fn as_mut_bitvec(&mut self) -> &mut BitVec<T, O> {
+		&mut self.inner
+	}
+
+	/// Views the bit-set as a slice of its underlying memory elements.
+	#[inline]
+	pub fn as_raw_slice(&self) -> &[T] {
+		self.inner.as_raw_slice()
+	}
+
+	/// Views the bit-set as a mutable slice of its underlying memory
+	/// elements.
+	#[inline]
+	pub fn as_raw_mut_slice(&mut self) -> &mut [T] {
+		self.inner.as_raw_mut_slice()
+	}
+
+	/// Creates an unsafe shared bit-pointer to the start of the buffer.
+	///
+	/// ## Original
+	///
+	/// [`Vec::as_ptr`](alloc::vec::Vec::as_ptr)
+	///
+	/// ## Safety
+	///
+	/// You must initialize the contents of the underlying buffer before
+	/// accessing memory through this pointer. See the `BitPtr` documentation
+	/// for more details.
+	#[inline]
+	pub fn as_bitptr(&self) -> BitPtr<Const, T, O> {
+		self.inner.as_bitptr()
+	}
+
+	/// Creates an unsafe writable bit-pointer to the start of the buffer.
+	///
+	/// ## Original
+	///
+	/// [`Vec::as_mut_ptr`](alloc::vec::Vec::as_mut_ptr)
+	///
+	/// ## Safety
+	///
+	/// You must initialize the contents of the underlying buffer before
+	/// accessing memory through this pointer. See the `BitPtr` documentation
+	/// for more details.
+	#[inline]
+	pub fn as_mut_bitptr(&mut self) -> BitPtr<Mut, T, O> {
+		self.inner.as_mut_bitptr()
+	}
+
+	/// Converts a bit-set into a boxed bit-slice.
+	///
+	/// This may cause a reällocation to drop any excess capacity.
+	///
+	/// ## Original
+	///
+	/// [`Vec::into_boxed_slice`](alloc::vec::Vec::into_boxed_slice)
+	#[inline]
+	pub fn into_boxed_bitslice(self) -> BitBox<T, O> {
+		self.inner.into_boxed_bitslice()
+	}
+
+	/// Converts a bit-set into a bit-vec.
+	#[inline]
+	pub fn into_bitvec(self) -> BitVec<T, O> {
+		self.inner
+	}
+}
+
+/// Utilities.
+impl<T, O> BitSet<T, O>
+where
+	T: BitStore,
+	O: BitOrder,
+{
+	/// Shrinks the inner vector to the minimum size, without changing capacity.
+	#[inline]
+	fn shrink_inner(&mut self) {
+		match self.inner.last_one() {
+			Some(idx) => self.inner.truncate(idx + 1),
+			None => self.inner.clear(),
+		}
+	}
+
+	/// Immutable shrink as a bitslice.
+	#[inline]
+	fn shrunken(&self) -> &BitSlice<T, O> {
+		match self.inner.last_one() {
+			Some(idx) => &self.inner[.. idx + 1],
+			None => Default::default(),
+		}
+	}
+}