Function dot_product_game_with_warnings 

pub fn dot_product_game_with_warnings(
    vecs: &[Vec<f64>],
    directed: bool,
    seed: u64,
) -> IgraphResult<(Graph, DotProductWarnings)>

Sample a random dot-product graph from caller-supplied latent position vectors and report whether clamp regimes were hit.

• vecs[i] is the latent position vector for vertex i. The slice length determines the resulting graph’s vertex count n. All entries must be finite (no NaN, no ±∞); every vector must share the same dimension d.
• directed — when true, both (i, j) and (j, i) are sampled independently with their own dot product (the matrix need not be symmetric — even though the dot product itself is, the draws are distinct). When false, only i < j pairs are sampled.
• seed initialises an internal SplitMix64 PRNG so the output is reproducible given the inputs.

Returns the generated graph together with DotProductWarnings flags so the caller can decide whether to log/abort/clip when the latent positions stray outside [0, 1].

Errors

• vecs[i].len() != vecs[0].len() for some i — every latent vector must share the same dimension.
• Any entry is non-finite (NaN or ±∞).
• vecs.len() > u32::MAX.

Complexity

O(n² · d) — every pair is inspected exactly once, each with a length-d dot product. The RNG-draw count is at most one per pair and is strictly less when the prob > 1 short-circuit fires.

Examples

use rust_igraph::dot_product_game_with_warnings;

// Three vertices on the unit interval (1-D latent space).
// p_{0,1} = 0.4·0.6 = 0.24; p_{0,2} = 0.4·0.5 = 0.20; p_{1,2} = 0.30.
let vecs = vec![vec![0.4], vec![0.6], vec![0.5]];
let (g, warn) = dot_product_game_with_warnings(&vecs, false, 42).unwrap();
assert_eq!(g.vcount(), 3);
assert!(!g.is_directed());
assert!(!warn.had_negative);
assert!(!warn.had_over_one);