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

//! Random sparse matrix generation

use crate::rand::distributions::Distribution;
use crate::rand::Rng;
use crate::rand::SeedableRng;
use sprs::indexing::SpIndex;
use sprs::{CsMat, CsMatI};

/// Re-export [`rand`](https://docs.rs/rand/0.7.3/rand/)
/// for version compatibility
pub mod rand {
    pub use rand::*;
}

/// Re-export [`rand_distr`](https://docs.rs/rand_distr/0.2.2/rand_distr)
/// for version compatibility
pub mod rand_distr {
    pub use rand_distr::*;
}

/// Generate a random sparse matrix matching the given density and sampling
/// the values of its non-zero elements from the provided distribution.
pub fn rand_csr<R, N, D, I>(
    rng: &mut R,
    dist: D,
    shape: (usize, usize),
    density: f64,
) -> CsMatI<N, I>
where
    R: Rng + ?Sized,
    D: Distribution<N>,
    N: Copy,
    I: SpIndex,
{
    assert!((0.0..=1.0).contains(&density));
    let exp_nnz =
        (density * (shape.0 as f64) * (shape.1 as f64)).ceil() as usize;
    let mut indptr = Vec::with_capacity(shape.0 + 1);
    let mut indices = Vec::with_capacity(exp_nnz);
    let mut data = Vec::with_capacity(exp_nnz);
    // sample row indices
    for _ in 0..exp_nnz {
        indices.push(I::from_usize(rng.gen_range(0, shape.0)));
        // Note: there won't be any correspondence between the data
        // sampled here and the row sampled before, but this does not matter
        // as we are sampling.
        data.push(dist.sample(rng));
    }
    indices.sort_unstable();
    indptr.push(I::from_usize(0));
    let mut count = 0;
    for &row in &indices {
        while indptr.len() != row.index() + 1 {
            indptr.push(I::from_usize(count));
        }
        count += 1;
    }
    while indptr.len() != shape.0 + 1 {
        indptr.push(I::from_usize(count));
    }
    assert_eq!(indptr.last().unwrap().index(), exp_nnz);
    indices.clear();
    for row in 0..shape.0 {
        let start = indptr[row].index();
        let end = indptr[row + 1].index();
        for _ in start..end {
            loop {
                let col = I::from_usize(rng.gen_range(0, shape.1));
                let loc = indices[start..].binary_search(&col);
                match loc {
                    Ok(_) => {
                        continue;
                    }
                    Err(loc) => {
                        indices.insert(start + loc, col);
                        break;
                    }
                }
            }
        }
        indices[start..end].sort_unstable();
    }

    CsMatI::new(shape, indptr, indices, data)
}

/// Convenient wrapper for the common case of sampling a matrix with standard
/// normal distribution of the nnz values, using a lightweight rng.
pub fn rand_csr_std(shape: (usize, usize), density: f64) -> CsMat<f64> {
    let mut rng = rand_pcg::Pcg64Mcg::from_entropy();
    rand_csr(&mut rng, crate::rand_distr::StandardNormal, shape, density)
}

#[cfg(test)]
mod tests {
    use rand::distributions::Standard;
    use rand::SeedableRng;
    use sprs::CsMat;

    #[test]
    fn empty_random_mat() {
        let mut rng = rand::thread_rng();
        let empty: CsMat<f64> =
            super::rand_csr(&mut rng, Standard, (0, 0), 0.3);
        assert_eq!(empty.nnz(), 0);
    }

    #[test]
    fn random_csr() {
        let mut rng = rand::rngs::StdRng::seed_from_u64(1234);
        let mat: CsMat<f64> =
            super::rand_csr(&mut rng, Standard, (100, 70), 0.3);
        assert!(mat.density() > 0.25);
        assert!(mat.density() < 0.35);

        let mat: CsMat<f64> =
            super::rand_csr(&mut rng, Standard, (1, 10000), 0.3);
        assert!(mat.density() > 0.28);
        assert!(mat.density() < 0.32);
    }

    #[test]
    fn random_csr_std() {
        let mat = super::rand_csr_std((100, 1000), 0.2);
        assert_eq!(mat.shape(), (100, 1000));
        // Not checking the density as I have no control over the seed
        // Checking the mean nnz value should be safe though
        assert!(
            mat.data().iter().sum::<f64>().abs() / (mat.data().len() as f64)
                < 0.05
        );
    }
}