gsva/
matrix.rs

1//! Minimal expression-matrix container: features (genes) in rows, samples in
2//! columns. Data is stored row-major, so a whole gene (row) is contiguous.
3
4use std::collections::HashMap;
5
6/// An expression matrix with named rows (features/genes) and columns (samples).
7///
8/// Element `(row i, col j)` is stored at `data[i * ncol + j]`. Missing values
9/// are represented as [`f64::NAN`] (the analog of R's `NA`).
10#[derive(Clone, Debug)]
11pub struct ExprMatrix {
12    row_names: Vec<String>,
13    col_names: Vec<String>,
14    data: Vec<f64>,
15    /// First-occurrence index for each row name, mirroring R's by-name matrix
16    /// indexing (`m[name, ]` selects the first row with that name).
17    row_index: HashMap<String, usize>,
18}
19
20impl ExprMatrix {
21    /// Build a matrix from row names, column names, and row-major data.
22    ///
23    /// Panics if `data.len() != row_names.len() * col_names.len()`.
24    pub fn new(row_names: Vec<String>, col_names: Vec<String>, data: Vec<f64>) -> Self {
25        assert_eq!(
26            data.len(),
27            row_names.len() * col_names.len(),
28            "data length must equal nrow * ncol"
29        );
30        let mut row_index = HashMap::with_capacity(row_names.len());
31        for (i, name) in row_names.iter().enumerate() {
32            // First occurrence wins, matching R's `m[name, ]` indexing.
33            row_index.entry(name.clone()).or_insert(i);
34        }
35        ExprMatrix {
36            row_names,
37            col_names,
38            data,
39            row_index,
40        }
41    }
42
43    /// Number of rows (features).
44    pub fn nrow(&self) -> usize {
45        self.row_names.len()
46    }
47
48    /// Number of columns (samples).
49    pub fn ncol(&self) -> usize {
50        self.col_names.len()
51    }
52
53    /// Row (feature) names.
54    pub fn row_names(&self) -> &[String] {
55        &self.row_names
56    }
57
58    /// Column (sample) names.
59    pub fn col_names(&self) -> &[String] {
60        &self.col_names
61    }
62
63    /// First row index for a feature name, if present.
64    pub fn row_of(&self, name: &str) -> Option<usize> {
65        self.row_index.get(name).copied()
66    }
67
68    /// Value at `(row, col)`.
69    #[inline]
70    pub fn get(&self, row: usize, col: usize) -> f64 {
71        self.data[row * self.col_names.len() + col]
72    }
73
74    /// A whole row (feature) as a contiguous slice across all samples.
75    #[inline]
76    pub fn row(&self, row: usize) -> &[f64] {
77        let ncol = self.col_names.len();
78        &self.data[row * ncol..(row + 1) * ncol]
79    }
80
81    /// Collect a whole column (sample) as a freshly allocated vector.
82    pub fn column(&self, col: usize) -> Vec<f64> {
83        (0..self.nrow()).map(|r| self.get(r, col)).collect()
84    }
85}
gsva/matrix.rs

gsva/
matrix.rs
