salib 0.1.1

Global sensitivity analysis in Rust. Variance-based (Sobol'), Morris, FAST/eFAST/RBD-FAST, Borgonovo δ, PAWN, DGSM, regression (SRC/SRRC/PCC/PRCC), PCE surrogate, Shapley effects, and more. Implemented from the primary literature.
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

//! Arrow `RecordBatch` conversions for SALib result types.

use arrow::array::{Float64Array, StringArray, UInt32Array};
use arrow::datatypes::{DataType, Field, Schema};
use arrow::record_batch::RecordBatch;
use std::sync::Arc;

fn factor_column(d: usize, names: Option<&[&str]>) -> (Field, Arc<dyn arrow::array::Array>) {
    match names {
        Some(n) => {
            let arr = StringArray::from(n[..d].to_vec());
            (Field::new("factor", DataType::Utf8, false), Arc::new(arr))
        }
        None => {
            let arr = UInt32Array::from((0..d as u32).collect::<Vec<_>>());
            (Field::new("factor", DataType::UInt32, false), Arc::new(arr))
        }
    }
}

fn f64_col(name: &str, data: &[f64]) -> (Field, Arc<dyn arrow::array::Array>) {
    (
        Field::new(name, DataType::Float64, false),
        Arc::new(Float64Array::from(data.to_vec())),
    )
}

fn build_batch(cols: Vec<(Field, Arc<dyn arrow::array::Array>)>) -> RecordBatch {
    let (fields, arrays): (Vec<_>, Vec<_>) = cols.into_iter().unzip();
    let schema = Arc::new(Schema::new(fields));
    RecordBatch::try_new(schema, arrays).unwrap_or_else(|e| panic!("arrow schema error: {e}"))
}

// ---------------------------------------------------------------------------
// Estimator types
// ---------------------------------------------------------------------------

/// Convert `SobolIndices` to an Arrow `RecordBatch` with columns
/// `[factor, S1, ST]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn sobol_to_batch(
    idx: &crate::estimators::SobolIndices,
    names: Option<&[&str]>,
) -> RecordBatch {
    let d = idx.dim;
    build_batch(vec![
        factor_column(d, names),
        f64_col("S1", &idx.first_order),
        f64_col("ST", &idx.total_order),
    ])
}

/// Convert `MorrisEffects` to an Arrow `RecordBatch` with columns
/// `[factor, mu, mu_star, sigma]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn morris_to_batch(
    eff: &crate::estimators::MorrisEffects,
    names: Option<&[&str]>,
) -> RecordBatch {
    let d = eff.d;
    build_batch(vec![
        factor_column(d, names),
        f64_col("mu", &eff.mu),
        f64_col("mu_star", &eff.mu_star),
        f64_col("sigma", &eff.sigma),
    ])
}

/// Convert `FastIndices` to an Arrow `RecordBatch` with columns
/// `[factor, S, ST]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn fast_to_batch(idx: &crate::estimators::FastIndices, names: Option<&[&str]>) -> RecordBatch {
    let d = idx.d();
    build_batch(vec![
        factor_column(d, names),
        f64_col("S", &idx.s),
        f64_col("ST", &idx.st),
    ])
}

/// Convert `RegressionIndices` to an Arrow `RecordBatch` with columns
/// `[factor, SRC, SRRC, PCC, PRCC]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn regression_to_batch(
    idx: &crate::estimators::RegressionIndices,
    names: Option<&[&str]>,
) -> RecordBatch {
    let d = idx.d();
    build_batch(vec![
        factor_column(d, names),
        f64_col("SRC", &idx.src),
        f64_col("SRRC", &idx.srrc),
        f64_col("PCC", &idx.pcc),
        f64_col("PRCC", &idx.prcc),
    ])
}

/// Convert `BorgonovoIndices` to an Arrow `RecordBatch` with columns
/// `[factor, delta]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn borgonovo_to_batch(
    idx: &crate::estimators::BorgonovoIndices,
    names: Option<&[&str]>,
) -> RecordBatch {
    let d = idx.d();
    build_batch(vec![factor_column(d, names), f64_col("delta", &idx.delta)])
}

/// Convert `PawnIndices` to an Arrow `RecordBatch` with columns
/// `[factor, median, mean, maximum, minimum, cv]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn pawn_to_batch(idx: &crate::estimators::PawnIndices, names: Option<&[&str]>) -> RecordBatch {
    let d = idx.d();
    build_batch(vec![
        factor_column(d, names),
        f64_col("median", &idx.median),
        f64_col("mean", &idx.mean),
        f64_col("maximum", &idx.maximum),
        f64_col("minimum", &idx.minimum),
        f64_col("cv", &idx.cv),
    ])
}

/// Convert `DgsmIndices` to an Arrow `RecordBatch` with columns
/// `[factor, vi, st_upper]`.
#[cfg(feature = "estimators")]
#[must_use]
pub fn dgsm_to_batch(idx: &crate::estimators::DgsmIndices, names: Option<&[&str]>) -> RecordBatch {
    let d = idx.d();
    build_batch(vec![
        factor_column(d, names),
        f64_col("vi", &idx.vi),
        f64_col("st_upper", &idx.st_upper),
    ])
}

// ---------------------------------------------------------------------------
// Shapley
// ---------------------------------------------------------------------------

/// Convert `ShapleyIndices` to an Arrow `RecordBatch` with columns
/// `[factor, Sh]`.
#[cfg(feature = "shapley")]
#[must_use]
pub fn shapley_to_batch(
    idx: &crate::shapley::ShapleyIndices,
    names: Option<&[&str]>,
) -> RecordBatch {
    let d = idx.k();
    build_batch(vec![factor_column(d, names), f64_col("Sh", &idx.sh)])
}