18466861223

Committed 13 Oct 2025 01:09PM UTC coverage: 65.352% (+0.01%) from 65.341%

Build # 18466861223

Build Type

push

github

Committed by

vigna

Commit Message

Sparse partial array implementation

Run Details

43 of 65 new or added lines in 2 files covered. (66.15%)

1 existing line in 1 file now uncovered.

3746 of 5732 relevant lines covered (65.35%)

116528486.3 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

65.43

/src/array/partial_array.rs

/*
 * SPDX-FileCopyrightText: 2025 Sebastiano Vigna
 *
 * SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
 */

//! Immutable partial array implementations.

use crate::bits::{BitVec, BitVecOps, BitVecOpsMut};
use crate::dict::EliasFanoBuilder;
use crate::dict::elias_fano::EfDict;
use crate::rank_sel::Rank9;
use crate::traits::{RankUnchecked, SuccUnchecked};
use mem_dbg::*;

/// Builder for creating an immutable partial array.
///
/// The builder allows you to specify the array length and then add
/// (position, value) pairs. Positions must be added in strictly
/// increasing order.
///
/// To get a builder you can use either [new_dense] or [new_sparse].
#[derive(Debug, Clone, MemDbg, MemSize)]
pub struct PartialArrayBuilder<T, B> {
    builder: B,
    values: Vec<T>,
    len: usize,
    min_next_pos: usize,
}

/// Creates a new builder for a dense partial array of the given length.
///
/// A dense partial array stores a bit vector of the given length to mark
/// which positions contain values, and use ranking on this bit vector to
/// map positions to indices in a contiguous value array.
///
/// If your set of values is really sparse, consider using a
/// [sparse partial array](new_sparse).
///
/// # Examples
///
/// ```rust
/// use sux::array::partial_array;
///
/// let mut builder = partial_array::new_dense(10);
/// builder.set(1, "foo");
/// builder.set(2, "hello");
/// builder.set(7, "world");
///
/// let array = builder.build();
/// assert_eq!(array.get(1), Some(&"foo"));
/// assert_eq!(array.get(2), Some(&"hello"));
/// assert_eq!(array.get(3), None);
/// assert_eq!(array.get(7), Some(&"world"));
/// ```
pub fn new_dense<T>(len: usize) -> PartialArrayBuilder<T, BitVec<Box<[usize]>>> {
    PartialArrayBuilder {
        builder: BitVec::new(len).into(),
        values: vec![],
        len,
        min_next_pos: 0,
    }
}

impl<T> PartialArrayBuilder<T, BitVec<Box<[usize]>>> {
    /// Sets a value at the given position.
    ///
    /// The provided position must be greater than the last position set.
    pub fn set(&mut self, position: usize, value: T) {
        if position < self.min_next_pos {
            panic!(
                "Positions must be set in increasing order: got {} after {}",
                position,
                self.min_next_pos - 1
            );
        }
        if position >= self.len {
            panic!(
                "Position {} is out of bounds for array of len {}",
                position, self.len
            );
        }
        // SAFETY: position < len
        unsafe {
            self.builder.set_unchecked(position, true);
        }
        self.values.push(value);
        self.min_next_pos = position + 1;
    }

    /// Builds the immutable dense partial array.
    pub fn build(self) -> PartialArray<T, Rank9<BitVec<Box<[usize]>>>> {
        let (bit_vec, values) = (self.builder, self.values);
        let rank9 = Rank9::new(bit_vec);
        let values = values.into_boxed_slice();

        PartialArray {
            index: rank9,
            values,
        }
    }
}

/// Creates a new builder for a sparse partial array of the given length.
///
/// A sparse partial array stores the non-empty positions of the array in an
/// [Elias-Fano](crate::dict::EliasFano) structure.
///
/// If your set of values is really dense, consider using a [dense partial
/// array](new_dense).
///
/// # Examples
///
/// ```rust
/// use sux::array::partial_array;
///
/// let mut builder = partial_array::new_sparse(10, 3);
/// builder.set(1, "foo");
/// builder.set(2, "hello");
/// builder.set(7, "world");
///
/// let array = builder.build();
/// assert_eq!(array.get(1), Some(&"foo"));
/// assert_eq!(array.get(2), Some(&"hello"));
/// assert_eq!(array.get(3), None);
/// assert_eq!(array.get(7), Some(&"world"));
/// ```
///
/// Note that you must specify the number of values in advance.
pub fn new_sparse<T>(len: usize, num_values: usize) -> PartialArrayBuilder<T, EliasFanoBuilder> {
    dbg!(len, num_values);

    PartialArrayBuilder {
        builder: EliasFanoBuilder::new(num_values, len).into(),
        values: vec![],
        len,
        min_next_pos: 0,
    }
}

impl<T> PartialArrayBuilder<T, EliasFanoBuilder> {
    /// Sets a value at the given position.
    ///
    /// The provided position must be greater than the last position
    /// set.
    pub fn set(&mut self, position: usize, value: T) {
        dbg!(position);
        if position < self.min_next_pos {
            panic!(
                "Positions must be set in increasing order: got {} after {}",
                position,
                self.min_next_pos - 1
            );
        }
        if position >= self.len {
            panic!(
                "Position {} is out of bounds for array of len {}",
                position, self.len
            );
        }
        // SAFETY: conditions have been just checked.
        unsafe { self.builder.push_unchecked(position) };
        self.values.push(value);
        self.min_next_pos = position + 1;
    }

    /// Builds the immutable sparse partial array.
    pub fn build(self) -> PartialArray<T, (EfDict, usize)> {
        let (builder, values) = (self.builder, self.values);
        let ef_dict = builder.build_with_dict();
        let values = values.into_boxed_slice();

        PartialArray {
            index: (ef_dict, self.min_next_pos),
            values,
        }
    }
}

/// Extends the builder with an iterator of (position, value) pairs.
///
/// Position must be in strictly increasing order. The first returned
/// position must be greater than the last position set.
impl<T> Extend<(usize, T)> for PartialArrayBuilder<T, BitVec<Box<[usize]>>> {
    fn extend<I: IntoIterator<Item = (usize, T)>>(&mut self, iter: I) {
        for (pos, val) in iter {
            self.set(pos, val);
        }
    }
}

/// Extends the builder with an iterator of (position, value) pairs.
///
/// Position must be in strictly increasing order. The first returned
/// position must be greater than the last position set.
impl<T> Extend<(usize, T)> for PartialArrayBuilder<T, EliasFanoBuilder> {
    fn extend<I: IntoIterator<Item = (usize, T)>>(&mut self, iter: I) {
        for (pos, val) in iter {
            self.set(pos, val);
        }
    }
}

/// An immutable partial array that supports efficient queries
/// in compacted storage.
///
/// This structure stores a *partial array*—an array in which only
/// some positions contain values. There is a [dense](new_dense)
/// and a [sparse](new_sparse) implementation with different
/// space/time trade-offs.
///
/// See [`PartialArrayBuilder`] for details on how to create a partial array.
#[derive(Debug, Clone, MemDbg, MemSize)]
#[cfg_attr(feature = "epserde", derive(epserde::Epserde))]
#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
pub struct PartialArray<T, P> {
    index: P,
    values: Box<[T]>,
}

impl<T, P> PartialArray<T, P> {
    /// Returns the number of values stored in the array.
    #[inline(always)]
    pub fn num_values(&self) -> usize {
        self.values.len()
    }
}

impl<T> PartialArray<T, Rank9<BitVec<Box<[usize]>>>> {
    /// Returns the total length of the array.
    ///
    /// This is the length that was specified when creating the builder,
    /// not the number of values actually stored.
    #[inline(always)]
    pub fn len(&self) -> usize {
        self.index.len()
    }

    /// Returns true if the array length is 0.
    #[inline(always)]
    pub fn is_empty(&self) -> bool {
        self.len() == 0
    }

    /// Gets a reference to the value at the given position.
    ///
    /// Returns `Some(&value)` if a value is present at the position,
    /// or `None` if no value was stored there.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use sux::array::partial_array;
    /// let mut builder = partial_array::new_dense(10);
    /// builder.set(5, 42);
    ///
    /// let array = builder.build();
    /// assert_eq!(array.get(5), Some(&42));
    /// assert_eq!(array.get(6), None);
    /// ```
    pub fn get(&self, position: usize) -> Option<&T> {
        if position >= self.len() {
            panic!(
                "Position {} is out of bounds for array of len {}",
                position,
                self.len()
            );
        }
        // Check if there's a value at this position
        // SAFETY: position < len()
        if !unsafe { self.index.get_unchecked(position) } {
            return None;
        }

        // Use ranking to find the index in the values array
        // SAFETY: position < len()
        let value_index = unsafe { self.index.rank_unchecked(position) };

        // SAFETY: necessarily value_index < num_values().
        Some(unsafe { self.values.get_unchecked(value_index) })
    }
}

impl<T> PartialArray<T, (EfDict, usize)> {
    /// Returns the total length of the array.
    ///
    /// This is the length that was specified when creating the builder,
    /// not the number of values actually stored.
    #[inline(always)]
    pub fn len(&self) -> usize {
        self.index.0.upper_bound()
    }

    /// Returns true if the array len is 0.
    #[inline(always)]
    pub fn is_empty(&self) -> bool {
        self.index.0.len() == 0
    }

    /// Gets a reference to the value at the given position.
    ///
    /// Returns `Some(&value)` if a value is present at the position,
    /// or `None` if no value was stored there.
    ///
    /// # Examples
    ///
    /// ```rust
    /// use sux::array::partial_array;
    /// let mut builder = partial_array::new_sparse(10, 1);
    /// builder.set(5, 42);
    ///
    /// let array = builder.build();
    /// assert_eq!(array.get(5), Some(&42));
    /// assert_eq!(array.get(6), None);
    /// ```
    pub fn get(&self, position: usize) -> Option<&T> {
        if position >= self.index.1 {
            if position >= self.len() {
                panic!(
                    "Position {} is out of bounds for array of len {}",
                    position,
                    self.len()
                );
            }
            return None;
        }
        // Check if there's a value at this position
        // SAFETY: position <= last set position
        let (index, pos) = unsafe { self.index.0.succ_unchecked::<false>(position) };

        return if pos != position {
            None
        } else {
            // SAFETY: necessarily value_index < num values.
            Some(unsafe { self.values.get_unchecked(index) })
        };
    }
}

1	/*
2	* SPDX-FileCopyrightText: 2025 Sebastiano Vigna
3	*
4	* SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
5	*/
6
7	//! Immutable partial array implementations.
8
9	use crate::bits::{BitVec, BitVecOps, BitVecOpsMut};
10	use crate::dict::EliasFanoBuilder;
11	use crate::dict::elias_fano::EfDict;
12	use crate::rank_sel::Rank9;
13	use crate::traits::{RankUnchecked, SuccUnchecked};
14	use mem_dbg::*;
15
16	/// Builder for creating an immutable partial array.
17	///
18	/// The builder allows you to specify the array length and then add
19	/// (position, value) pairs. Positions must be added in strictly
20	/// increasing order.
21	///
22	/// To get a builder you can use either [new_dense] or [new_sparse].
23	#[derive(Debug, Clone, MemDbg, MemSize)]
24	pub struct PartialArrayBuilder<T, B> {
25	builder: B,
26	values: Vec<T>,
27	len: usize,
28	min_next_pos: usize,
29	}
30
31	/// Creates a new builder for a dense partial array of the given length.
32	///
33	/// A dense partial array stores a bit vector of the given length to mark
34	/// which positions contain values, and use ranking on this bit vector to
35	/// map positions to indices in a contiguous value array.
36	///
37	/// If your set of values is really sparse, consider using a
38	/// [sparse partial array](new_sparse).
39	///
40	/// # Examples
41	///
42	/// ```rust
43	/// use sux::array::partial_array;
44	///
45	/// let mut builder = partial_array::new_dense(10);
46	/// builder.set(1, "foo");
47	/// builder.set(2, "hello");
48	/// builder.set(7, "world");
49	///
50	/// let array = builder.build();
51	/// assert_eq!(array.get(1), Some(&"foo"));
52	/// assert_eq!(array.get(2), Some(&"hello"));
53	/// assert_eq!(array.get(3), None);
54	/// assert_eq!(array.get(7), Some(&"world"));
55	/// ```
56	pub fn new_dense<T>(len: usize) -> PartialArrayBuilder<T, BitVec<Box<[usize]>>> {	6✔
57	PartialArrayBuilder {
58	builder: BitVec::new(len).into(),	24✔
59	values: vec![],	6✔
60	len,
61	min_next_pos: 0,
62	}
63	}
64
65	impl<T> PartialArrayBuilder<T, BitVec<Box<[usize]>>> {
66	/// Sets a value at the given position.
67	///
68	/// The provided position must be greater than the last position set.
69	pub fn set(&mut self, position: usize, value: T) {	7✔
70	if position < self.min_next_pos {	7✔
71	panic!(	1✔
72	"Positions must be set in increasing order: got {} after {}",	×
73	position,	×
74	self.min_next_pos - 1	×
75	);
76	}
77	if position >= self.len {	6✔
78	panic!(	1✔
79	"Position {} is out of bounds for array of len {}",	1✔
NEW 80	position, self.len	×
81	);
82	}
83	// SAFETY: position < len
84	unsafe {
NEW 85	self.builder.set_unchecked(position, true);	×
86	}
87	self.values.push(value);	×
88	self.min_next_pos = position + 1;	×
89	}
90
91	/// Builds the immutable dense partial array.
92	pub fn build(self) -> PartialArray<T, Rank9<BitVec<Box<[usize]>>>> {	4✔
93	let (bit_vec, values) = (self.builder, self.values);	12✔
94	let rank9 = Rank9::new(bit_vec);	12✔
95	let values = values.into_boxed_slice();	12✔
96
97	PartialArray {
98	index: rank9,
99	values,
100	}
101	}
102	}
103
104	/// Creates a new builder for a sparse partial array of the given length.
105	///
106	/// A sparse partial array stores the non-empty positions of the array in an
107	/// [Elias-Fano](crate::dict::EliasFano) structure.
108	///
109	/// If your set of values is really dense, consider using a [dense partial
110	/// array](new_dense).
111	///
112	/// # Examples
113	///
114	/// ```rust
115	/// use sux::array::partial_array;
116	///
117	/// let mut builder = partial_array::new_sparse(10, 3);
118	/// builder.set(1, "foo");
119	/// builder.set(2, "hello");
120	/// builder.set(7, "world");
121	///
122	/// let array = builder.build();
123	/// assert_eq!(array.get(1), Some(&"foo"));
124	/// assert_eq!(array.get(2), Some(&"hello"));
125	/// assert_eq!(array.get(3), None);
126	/// assert_eq!(array.get(7), Some(&"world"));
127	/// ```
128	///
129	/// Note that you must specify the number of values in advance.
130	pub fn new_sparse<T>(len: usize, num_values: usize) -> PartialArrayBuilder<T, EliasFanoBuilder> {	6✔
131	dbg!(len, num_values);	18✔
132
133	PartialArrayBuilder {
134	builder: EliasFanoBuilder::new(num_values, len).into(),	30✔
135	values: vec![],	6✔
136	len,
137	min_next_pos: 0,
138	}
139	}
140
141	impl<T> PartialArrayBuilder<T, EliasFanoBuilder> {
142	/// Sets a value at the given position.
143	///
144	/// The provided position must be greater than the last position
145	/// set.
146	pub fn set(&mut self, position: usize, value: T) {	7✔
147	dbg!(position);	14✔
148	if position < self.min_next_pos {	7✔
149	panic!(	1✔
NEW 150	"Positions must be set in increasing order: got {} after {}",	×
NEW 151	position,	×
NEW 152	self.min_next_pos - 1	×
153	);
154	}
155	if position >= self.len {	6✔
156	panic!(	1✔
157	"Position {} is out of bounds for array of len {}",	1✔
NEW 158	position, self.len	×
159	);
160	}
161	// SAFETY: conditions have been just checked.
NEW 162	unsafe { self.builder.push_unchecked(position) };	×
NEW 163	self.values.push(value);	×
NEW 164	self.min_next_pos = position + 1;	×
165	}
166
167	/// Builds the immutable sparse partial array.
168	pub fn build(self) -> PartialArray<T, (EfDict, usize)> {	4✔
169	let (builder, values) = (self.builder, self.values);	12✔
170	let ef_dict = builder.build_with_dict();	12✔
171	let values = values.into_boxed_slice();	12✔
172
173	PartialArray {
174	index: (ef_dict, self.min_next_pos),	4✔
175	values,
176	}
177	}
178	}
179
180	/// Extends the builder with an iterator of (position, value) pairs.
181	///
182	/// Position must be in strictly increasing order. The first returned
183	/// position must be greater than the last position set.
184	impl<T> Extend<(usize, T)> for PartialArrayBuilder<T, BitVec<Box<[usize]>>> {
185	fn extend<I: IntoIterator<Item = (usize, T)>>(&mut self, iter: I) {	×
186	for (pos, val) in iter {	×
187	self.set(pos, val);	×
188	}
189	}
190	}
191
192	/// Extends the builder with an iterator of (position, value) pairs.
193	///
194	/// Position must be in strictly increasing order. The first returned
195	/// position must be greater than the last position set.
196	impl<T> Extend<(usize, T)> for PartialArrayBuilder<T, EliasFanoBuilder> {
NEW 197	fn extend<I: IntoIterator<Item = (usize, T)>>(&mut self, iter: I) {	×
NEW 198	for (pos, val) in iter {	×
NEW 199	self.set(pos, val);	×
200	}
201	}
202	}
203
204	/// An immutable partial array that supports efficient queries
205	/// in compacted storage.
206	///
207	/// This structure stores a partial array—an array in which only
208	/// some positions contain values. There is a [dense](new_dense)
209	/// and a [sparse](new_sparse) implementation with different
210	/// space/time trade-offs.
211	///
212	/// See [`PartialArrayBuilder`] for details on how to create a partial array.
213	#[derive(Debug, Clone, MemDbg, MemSize)]
214	#[cfg_attr(feature = "epserde", derive(epserde::Epserde))]
215	#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))]
216	pub struct PartialArray<T, P> {
217	index: P,
218	values: Box<[T]>,
219	}
220
221	impl<T, P> PartialArray<T, P> {
222	/// Returns the number of values stored in the array.
223	#[inline(always)]
224	pub fn num_values(&self) -> usize {	6✔
225	self.values.len()	12✔
226	}
227	}
228
229	impl<T> PartialArray<T, Rank9<BitVec<Box<[usize]>>>> {
230	/// Returns the total length of the array.
231	///
232	/// This is the length that was specified when creating the builder,
233	/// not the number of values actually stored.
234	#[inline(always)]
235	pub fn len(&self) -> usize {	6✔
236	self.index.len()	12✔
237	}
238
239	/// Returns true if the array length is 0.
240	#[inline(always)]
241	pub fn is_empty(&self) -> bool {	1✔
242	self.len() == 0	1✔
243	}
244
245	/// Gets a reference to the value at the given position.
246	///
247	/// Returns `Some(&value)` if a value is present at the position,
248	/// or `None` if no value was stored there.
249	///
250	/// # Examples
251	///
252	/// ```rust
253	/// use sux::array::partial_array;
254	/// let mut builder = partial_array::new_dense(10);
255	/// builder.set(5, 42);
256	///
257	/// let array = builder.build();
258	/// assert_eq!(array.get(5), Some(&42));
259	/// assert_eq!(array.get(6), None);
260	/// ```
261	pub fn get(&self, position: usize) -> Option<&T> {	1,011✔
262	if position >= self.len() {	2,022✔
263	panic!(	1✔
NEW 264	"Position {} is out of bounds for array of len {}",	×
NEW 265	position,	×
NEW 266	self.len()	×
267	);
268	}
269	// Check if there's a value at this position
270	// SAFETY: position < len()
271	if !unsafe { self.index.get_unchecked(position) } {	2,020✔
272	return None;	1,006✔
273	}
274
275	// Use ranking to find the index in the values array
276	// SAFETY: position < len()
277	let value_index = unsafe { self.index.rank_unchecked(position) };	16✔
278
279	// SAFETY: necessarily value_index < num_values().
280	Some(unsafe { self.values.get_unchecked(value_index) })	8✔
281	}
282	}
283
284	impl<T> PartialArray<T, (EfDict, usize)> {
285	/// Returns the total length of the array.
286	///
287	/// This is the length that was specified when creating the builder,
288	/// not the number of values actually stored.
289	#[inline(always)]
290	pub fn len(&self) -> usize {	5✔
291	self.index.0.upper_bound()	10✔
292	}
293
294	/// Returns true if the array len is 0.
295	#[inline(always)]
296	pub fn is_empty(&self) -> bool {	1✔
297	self.index.0.len() == 0	1✔
298	}
299
300	/// Gets a reference to the value at the given position.
301	///
302	/// Returns `Some(&value)` if a value is present at the position,
303	/// or `None` if no value was stored there.
304	///
305	/// # Examples
306	///
307	/// ```rust
308	/// use sux::array::partial_array;
309	/// let mut builder = partial_array::new_sparse(10, 1);
310	/// builder.set(5, 42);
311	///
312	/// let array = builder.build();
313	/// assert_eq!(array.get(5), Some(&42));
314	/// assert_eq!(array.get(6), None);
315	/// ```
316	pub fn get(&self, position: usize) -> Option<&T> {	1,011✔
317	if position >= self.index.1 {	1,011✔
318	if position >= self.len() {	1,004✔
319	panic!(	1✔
NEW 320	"Position {} is out of bounds for array of len {}",	×
NEW 321	position,	×
NEW 322	self.len()	×
323	);
324	}
325	return None;	501✔
326	}
327	// Check if there's a value at this position
328	// SAFETY: position <= last set position
NEW 329	let (index, pos) = unsafe { self.index.0.succ_unchecked::<false>(position) };	×
330
NEW 331	return if pos != position {	×
332	None	505✔
333	} else {
334	// SAFETY: necessarily value_index < num values.
335	Some(unsafe { self.values.get_unchecked(index) })	4✔
336	};
337	}
338	}

vigna / sux-rs / 18466861223

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous