numr 0.5.2

High-performance numerical computing with multi-backend GPU acceleration (CPU/CUDA/WebGPU)
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

//! CSR format conversion: to_coo, to_csc

use super::CsrData;
use crate::dtype::DType;
use crate::error::Result;
use crate::runtime::Runtime;
use crate::sparse::{CooData, CscData, SparseStorage};
use crate::tensor::Tensor;

impl<R: Runtime<DType = DType>> CsrData<R> {
    /// Convert to COO format
    ///
    /// Expands the compressed row pointers into explicit row indices.
    /// The resulting COO is sorted in row-major order.
    pub fn to_coo(&self) -> Result<CooData<R>> {
        let [nrows, _ncols] = self.shape;
        let nnz = self.nnz();

        // Handle empty case
        if nnz == 0 {
            return Ok(CooData::empty(
                self.shape,
                self.dtype(),
                self.values.device(),
            ));
        }

        // Read CSR data to host
        let row_ptrs: Vec<i64> = self.row_ptrs.to_vec();
        let col_indices: Vec<i64> = self.col_indices.to_vec();

        // Expand row pointers into explicit row indices
        let mut row_indices: Vec<i64> = Vec::with_capacity(nnz);
        for row in 0..nrows {
            let start = row_ptrs[row] as usize;
            let end = row_ptrs[row + 1] as usize;
            for _ in start..end {
                row_indices.push(row as i64);
            }
        }

        // Dispatch on dtype to copy values
        let device = self.values.device();
        crate::dispatch_dtype!(self.dtype(), T => {
            let values: Vec<T> = self.values.to_vec();

            let row_indices_tensor = Tensor::from_slice(&row_indices, &[row_indices.len()], device);
            let col_indices_tensor = Tensor::from_slice(&col_indices, &[col_indices.len()], device);
            let values_tensor = Tensor::from_slice(&values, &[values.len()], device);

            let mut coo = CooData::new(row_indices_tensor, col_indices_tensor, values_tensor, self.shape)?;
            // CSR is already sorted by row, so mark COO as sorted
            unsafe { coo.set_sorted(true); }
            return Ok(coo);
        }, "CSR to COO conversion");
    }

    /// Convert to CSC format
    ///
    /// Converts via COO format, then sorts by column.
    pub fn to_csc(&self) -> Result<CscData<R>> {
        // Convert to COO first, then to CSC
        let coo = self.to_coo()?;
        coo.to_csc()
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::dtype::DType;
    use crate::runtime::cpu::CpuRuntime;

    #[test]
    fn test_csr_to_coo() {
        let device = <CpuRuntime as Runtime>::Device::default();

        // Matrix:
        // [1, 0, 2]
        // [0, 0, 3]
        // [4, 5, 0]
        //
        // CSR:
        // row_ptrs: [0, 2, 3, 5]
        // col_indices: [0, 2, 2, 0, 1]
        // values: [1, 2, 3, 4, 5]
        let row_ptrs = vec![0i64, 2, 3, 5];
        let col_indices = vec![0i64, 2, 2, 0, 1];
        let values = vec![1.0f32, 2.0, 3.0, 4.0, 5.0];

        let csr =
            CsrData::<CpuRuntime>::from_slices(&row_ptrs, &col_indices, &values, [3, 3], &device)
                .unwrap();
        let coo = csr.to_coo().unwrap();

        // Expected COO:
        // row_indices: [0, 0, 1, 2, 2]
        // col_indices: [0, 2, 2, 0, 1]
        // values: [1, 2, 3, 4, 5]
        assert_eq!(coo.nnz(), 5);
        assert_eq!(coo.shape(), [3, 3]);
        assert!(coo.is_sorted());

        let coo_rows: Vec<i64> = coo.row_indices().to_vec();
        let coo_cols: Vec<i64> = coo.col_indices().to_vec();
        let coo_values: Vec<f32> = coo.values().to_vec();

        assert_eq!(coo_rows, vec![0, 0, 1, 2, 2]);
        assert_eq!(coo_cols, vec![0, 2, 2, 0, 1]);
        assert_eq!(coo_values, vec![1.0, 2.0, 3.0, 4.0, 5.0]);
    }

    #[test]
    fn test_csr_to_coo_empty() {
        let device = <CpuRuntime as Runtime>::Device::default();
        let csr = CsrData::<CpuRuntime>::empty([3, 3], DType::F32, &device);
        let coo = csr.to_coo().unwrap();

        assert_eq!(coo.nnz(), 0);
        assert_eq!(coo.shape(), [3, 3]);
        assert!(coo.is_sorted());
    }

    #[test]
    fn test_csr_to_coo_single_element() {
        let device = <CpuRuntime as Runtime>::Device::default();

        // Single element at row 1, col 2
        let row_ptrs = vec![0i64, 0, 1, 1];
        let col_indices = vec![2i64];
        let values = vec![42.0f32];

        let csr =
            CsrData::<CpuRuntime>::from_slices(&row_ptrs, &col_indices, &values, [3, 3], &device)
                .unwrap();
        let coo = csr.to_coo().unwrap();

        assert_eq!(coo.nnz(), 1);

        let coo_rows: Vec<i64> = coo.row_indices().to_vec();
        let coo_cols: Vec<i64> = coo.col_indices().to_vec();
        let coo_values: Vec<f32> = coo.values().to_vec();

        assert_eq!(coo_rows, vec![1]);
        assert_eq!(coo_cols, vec![2]);
        assert_eq!(coo_values, vec![42.0]);
    }

    #[test]
    fn test_csr_to_csc() {
        let device = <CpuRuntime as Runtime>::Device::default();

        // Matrix:
        // [1, 0, 2]
        // [0, 0, 3]
        // [4, 5, 0]
        //
        // CSR:
        // row_ptrs: [0, 2, 3, 5]
        // col_indices: [0, 2, 2, 0, 1]
        // values: [1, 2, 3, 4, 5]
        let row_ptrs = vec![0i64, 2, 3, 5];
        let col_indices = vec![0i64, 2, 2, 0, 1];
        let values = vec![1.0f32, 2.0, 3.0, 4.0, 5.0];

        let csr =
            CsrData::<CpuRuntime>::from_slices(&row_ptrs, &col_indices, &values, [3, 3], &device)
                .unwrap();
        let csc = csr.to_csc().unwrap();

        // Expected CSC (column-major order):
        // col_ptrs: [0, 2, 3, 5]
        // row_indices: [0, 2, 2, 0, 1]
        // values: [1, 4, 5, 2, 3]
        assert_eq!(csc.nnz(), 5);
        assert_eq!(csc.shape(), [3, 3]);

        let csc_col_ptrs: Vec<i64> = csc.col_ptrs().to_vec();
        let csc_row_indices: Vec<i64> = csc.row_indices().to_vec();
        let csc_values: Vec<f32> = csc.values().to_vec();

        assert_eq!(csc_col_ptrs, vec![0, 2, 3, 5]);
        assert_eq!(csc_row_indices, vec![0, 2, 2, 0, 1]);
        assert_eq!(csc_values, vec![1.0, 4.0, 5.0, 2.0, 3.0]);
    }

    #[test]
    fn test_csr_to_csc_empty() {
        let device = <CpuRuntime as Runtime>::Device::default();
        let csr = CsrData::<CpuRuntime>::empty([3, 3], DType::F32, &device);
        let csc = csr.to_csc().unwrap();

        assert_eq!(csc.nnz(), 0);
        assert_eq!(csc.shape(), [3, 3]);
    }
}