rustebra 0.4.0

A hybrid no_std/alloc linear algebra crate for Rust, scaling from embedded targets to dynamic Krylov subspace solvers.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230

use alloc::vec;
use alloc::vec::Vec;

use crate::scalar::Scalar;

use super::{CscMatrix, CsrMatrix, SortedCscMatrix, SortedCsrMatrix};

/// Error returned by sparse arithmetic functions when operands have incompatible shapes,
/// or when the vector length does not match the matrix column count.
///
/// # Examples
///
/// ```
/// use rustebra::sparse::{CsrMatrix, DimensionMismatch, add_csr};
///
/// // 2x2 vs 3x2: row counts disagree.
/// let a = CsrMatrix::new(2, 2, vec![0, 1, 2], vec![0, 1], vec![1.0_f64, 2.0]).unwrap();
/// let b = CsrMatrix::new(3, 2, vec![0, 1, 2, 2], vec![0, 1], vec![1.0_f64, 2.0]).unwrap();
/// assert_eq!(add_csr(&a, &b), Err(DimensionMismatch::Shape));
/// ```
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum DimensionMismatch {
    /// Operands have incompatible shapes, or a vector length does not match the expected
    /// matrix dimension.
    Shape,
    /// An arithmetic overflow occurred while computing a size (e.g. total non-zero count
    /// exceeding `u32::MAX`).
    DimensionOverflow,
}

/// Adds two CSR sparse matrices element-wise, merging and deduplicating entries per row.
///
/// For each row, entries from `a` and `b` are merged; entries at the same column position
/// have their values summed. The output always has sorted, deduplicated column indices
/// within each row (same invariant as matrices produced by [`crate::sparse::coo_to_csr`]),
/// which is why the result is a [`SortedCsrMatrix`] rather than a plain `CsrMatrix`.
///
/// # Errors
///
/// Returns `Err(DimensionMismatch::Shape)` when `a` and `b` don't have the same shape.
///
/// # Examples
///
/// ```
/// use rustebra::sparse::{CsrMatrix, add_csr};
///
/// // [2  0]   [1  3]   [3  3]
/// // [0  5] + [0  4] = [0  9]
/// let a = CsrMatrix::new(2, 2, vec![0, 1, 2], vec![0, 1], vec![2.0_f64, 5.0]).unwrap();
/// let b = CsrMatrix::new(2, 2, vec![0, 2, 3], vec![0, 1, 1], vec![1.0_f64, 3.0, 4.0]).unwrap();
/// let c = add_csr(&a, &b).unwrap();
/// assert_eq!(c.row_ptr(),     &[0, 2, 3]);
/// assert_eq!(c.col_indices(), &[0, 1, 1]);
/// assert_eq!(c.values(),      &[3.0, 3.0, 9.0]);
/// ```
pub fn add_csr<T: Scalar>(
    a: &CsrMatrix<T>,
    b: &CsrMatrix<T>,
) -> Result<SortedCsrMatrix<T>, DimensionMismatch> {
    if a.rows() != b.rows() || a.cols() != b.cols() {
        return Err(DimensionMismatch::Shape);
    }
    let rows = a.rows();
    let cols = a.cols();

    let mut out_row_ptr = vec![0u32; rows + 1];
    let mut out_col: Vec<u32> = Vec::new();
    let mut out_val: Vec<T> = Vec::new();
    let mut pairs: Vec<(u32, T)> = Vec::new();

    for r in 0..rows {
        pairs.clear();
        let a_range = a.row_ptr()[r] as usize..a.row_ptr()[r + 1] as usize;
        for k in a_range {
            pairs.push((a.col_indices()[k], a.values()[k]));
        }
        let b_range = b.row_ptr()[r] as usize..b.row_ptr()[r + 1] as usize;
        for k in b_range {
            pairs.push((b.col_indices()[k], b.values()[k]));
        }
        pairs.sort_by_key(|&(c, _)| c);
        let mut k = 0;
        while k < pairs.len() {
            let (col, mut sum) = pairs[k];
            let mut j = k + 1;
            while j < pairs.len() && pairs[j].0 == col {
                sum = sum.add(pairs[j].1);
                j += 1;
            }
            if sum != T::zero() {
                out_col.push(col);
                out_val.push(sum);
            }
            k = j;
        }
        out_row_ptr[r + 1] =
            u32::try_from(out_col.len()).map_err(|_| DimensionMismatch::DimensionOverflow)?;
    }

    Ok(SortedCsrMatrix::from_sorted_unchecked(CsrMatrix::new_raw(
        rows,
        cols,
        out_row_ptr,
        out_col,
        out_val,
    )))
}

/// Adds two CSC sparse matrices element-wise, merging and deduplicating entries per column.
///
/// For each column, entries from `a` and `b` are merged; entries at the same row position
/// have their values summed. The output always has sorted, deduplicated row indices within
/// each column, which is why the result is a [`SortedCscMatrix`] rather than a plain
/// `CscMatrix`.
///
/// # Errors
///
/// Returns `Err(DimensionMismatch::Shape)` when `a` and `b` don't have the same shape.
///
/// # Examples
///
/// ```
/// use rustebra::sparse::{CscMatrix, add_csc};
///
/// // Column 0: a has row 0 → 2.0; b has row 0 → 1.0; sum = 3.0.
/// // Column 1: a has row 1 → 5.0; b has row 1 → 4.0; sum = 9.0.
/// let a = CscMatrix::new(2, 2, vec![0, 1, 2], vec![0, 1], vec![2.0_f64, 5.0]).unwrap();
/// let b = CscMatrix::new(2, 2, vec![0, 1, 2], vec![0, 1], vec![1.0_f64, 4.0]).unwrap();
/// let c = add_csc(&a, &b).unwrap();
/// assert_eq!(c.col_ptr(),     &[0, 1, 2]);
/// assert_eq!(c.row_indices(), &[0, 1]);
/// assert_eq!(c.values(),      &[3.0, 9.0]);
/// ```
pub fn add_csc<T: Scalar>(
    a: &CscMatrix<T>,
    b: &CscMatrix<T>,
) -> Result<SortedCscMatrix<T>, DimensionMismatch> {
    if a.rows() != b.rows() || a.cols() != b.cols() {
        return Err(DimensionMismatch::Shape);
    }
    let rows = a.rows();
    let cols = a.cols();

    let mut out_col_ptr = vec![0u32; cols + 1];
    let mut out_row: Vec<u32> = Vec::new();
    let mut out_val: Vec<T> = Vec::new();
    let mut pairs: Vec<(u32, T)> = Vec::new();

    for c in 0..cols {
        pairs.clear();
        let a_range = a.col_ptr()[c] as usize..a.col_ptr()[c + 1] as usize;
        for k in a_range {
            pairs.push((a.row_indices()[k], a.values()[k]));
        }
        let b_range = b.col_ptr()[c] as usize..b.col_ptr()[c + 1] as usize;
        for k in b_range {
            pairs.push((b.row_indices()[k], b.values()[k]));
        }
        pairs.sort_by_key(|&(r, _)| r);
        let mut k = 0;
        while k < pairs.len() {
            let (row, mut sum) = pairs[k];
            let mut j = k + 1;
            while j < pairs.len() && pairs[j].0 == row {
                sum = sum.add(pairs[j].1);
                j += 1;
            }
            if sum != T::zero() {
                out_row.push(row);
                out_val.push(sum);
            }
            k = j;
        }
        out_col_ptr[c + 1] =
            u32::try_from(out_row.len()).map_err(|_| DimensionMismatch::DimensionOverflow)?;
    }

    Ok(SortedCscMatrix::from_sorted_unchecked(CscMatrix::new_raw(
        rows,
        cols,
        out_col_ptr,
        out_row,
        out_val,
    )))
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn s1_add_csr_cancellation_zeros() {
        let a = CsrMatrix::new(1, 2, vec![0, 1], vec![0], vec![3.0_f64]).unwrap();
        let b = CsrMatrix::new(1, 2, vec![0, 1], vec![0], vec![-3.0_f64]).unwrap();
        let c = add_csr(&a, &b).unwrap();
        assert_eq!(c.nnz(), 0);
    }

    #[test]
    fn s1_add_csc_cancellation_zeros() {
        let a = CscMatrix::new(2, 1, vec![0, 1], vec![0], vec![5.0_f64]).unwrap();
        let b = CscMatrix::new(2, 1, vec![0, 1], vec![0], vec![-5.0_f64]).unwrap();
        let c = add_csc(&a, &b).unwrap();
        assert_eq!(c.nnz(), 0);
    }

    #[test]
    fn add_csr_result_has_sorted_col_indices_per_row() {
        // Row 0: a contributes col 2, b contributes cols 0 and 1 -> merged/sorted [0, 1, 2].
        let a = CsrMatrix::new(1, 3, vec![0, 1], vec![2], vec![9.0_f64]).unwrap();
        let b = CsrMatrix::new(1, 3, vec![0, 2], vec![0, 1], vec![1.0_f64, 2.0]).unwrap();
        let c = add_csr(&a, &b).unwrap();
        assert_eq!(c.col_indices(), &[0, 1, 2]);
        for w in c.col_indices().windows(2) {
            assert!(w[0] < w[1]);
        }
    }

    #[test]
    fn add_csc_result_has_sorted_row_indices_per_column() {
        // Column 0: a contributes row 2, b contributes rows 0 and 1 -> merged/sorted [0, 1, 2].
        let a = CscMatrix::new(3, 1, vec![0, 1], vec![2], vec![9.0_f64]).unwrap();
        let b = CscMatrix::new(3, 1, vec![0, 2], vec![0, 1], vec![1.0_f64, 2.0]).unwrap();
        let c = add_csc(&a, &b).unwrap();
        assert_eq!(c.row_indices(), &[0, 1, 2]);
        for w in c.row_indices().windows(2) {
            assert!(w[0] < w[1]);
        }
    }
}