Crate dace_rs 

Pure Rust implementation of DACE, the Differential Algebra Computational Toolbox.

DACE computes with truncated multivariate Taylor polynomials (“differential algebra”): it propagates high-order expansions of arbitrary multivariate functions through arithmetic and elementary-function composition, with a configurable cut-off epsilon and truncation order. This crate is a full rewrite of DACE 2.1 in pure Rust.

Quickstart

dace_rs::init(20, 2).unwrap();
let x = dace_rs::Da::variable(1);
let y = dace_rs::Da::variable(2);
let f = (x.clone() * x.clone() + y).sin();
// coefficient of x^2 in sin(x^2+y) at the origin is cos(0) = 1
assert!((f.get_coefficient(&[2, 0]) - 1.0).abs() < 1e-14);

Multithreading

Da and CompiledDa are Send + Sync: sharing them across threads (e.g. Rayon, or PyO3’s allow_threads) is supported. New threads need no explicit initialization — computation settings (epsilon cutoff, truncation order) and scratch buffers are lazily derived from the active context on each thread’s first use.

After init is called again, every thread re-derives its settings on next use, as if the thread had never been used; user-set epsilon and truncation order therefore do not survive re-initialization on any thread. Values keep operating on their original context, and mixing values from different contexts in one operation panics (existing semantics, enforced by Da::assert_same_context).

Da::random uses a per-thread LCG seeded with a fixed identical value on every thread: parallel draws produce the same sequence on each thread, not distinct streams.

dace_rs::init(20, 2).unwrap();
let x = dace_rs::Da::variable(1);
let c = std::thread::spawn(move || {
    let y = dace_rs::Da::variable(2);
    (x * y).get_coefficient(&[1, 1])
})
.join()
.unwrap();
assert_eq!(c, 1.0);

Re-exports

pub use ads::AdsConfig;

pub use ads::AdsLeaf;

pub use ads::AdsResult;

pub use ads::ToleranceKind;

pub use context::epsilon;

pub use context::init;

pub use context::initialized;

pub use context::machine_epsilon;

pub use context::max_monomials;

pub use context::max_order;

pub use context::max_variables;

pub use context::pop_truncation_order;

pub use context::push_truncation_order;

pub use context::set_epsilon;

pub use context::set_truncation_order;

pub use context::truncation_order;

pub use context::version;

pub use elementary::acos;

pub use elementary::acosh;

pub use elementary::asin;

pub use elementary::asinh;

pub use elementary::atan;

pub use elementary::atan2;

pub use elementary::atanh;

pub use elementary::cbrt;

pub use elementary::cos;

pub use elementary::cosh;

pub use elementary::erf;

pub use elementary::erfc;

pub use elementary::exp;

pub use elementary::hypot;

pub use elementary::icrt;

pub use elementary::isrt;

pub use elementary::log;

pub use elementary::log_base;

pub use elementary::log2;

pub use elementary::log10;

pub use elementary::modulo;

pub use elementary::powf;

pub use elementary::powi;

pub use elementary::root;

pub use elementary::round;

pub use elementary::sin;

pub use elementary::sinh;

pub use elementary::sqrt;

pub use elementary::tan;

pub use elementary::tanh;

pub use elementary::trunc;

pub use error::DaceError;

pub use eval::CompiledDa;

pub use norm::Interval;

pub use norm::NormType;

pub use special::bessel_i;

pub use special::bessel_j;

pub use special::bessel_k;

pub use special::bessel_y;

pub use special::gamma;

pub use special::log_gamma;

pub use special::psi;

Modules

ads

Automatic domain splitting (ADS): recursive bisection and re-expansion of a DA map over a large uncertainty box.

context

DACE computation context: initialization, monomial index encoding, and thread-local computation settings.

elementary

Elementary functions of DA values.

error

Error type for DACE operations.

eval

Polynomial evaluation: compiled evaluation trees, partial evaluation, and variable substitution.

io

Text and binary I/O of Da values: the daceWrite display format and its parser (daceRead), the binary blob format (daceExportBlob / daceImportBlob), and the code-generation formatter (DASimpleFormatter).

monomial

A decoded monomial: exponent vector plus coefficient.

norm

Norms, order-sorted norms, estimates, and interval bounds of DA values.

special

Special functions of DA values: Bessel functions, gamma, and polygamma.

vector

Vector-of-DA operations: the DaVector extension trait, inverse maps, and scalar-vector helpers.

Structs

Da

A truncated multivariate Taylor polynomial (“differential algebra” value).

Functions

fma

The weighted sum afac * a + bfac * b (daceWeightedSum).