TulipRS

High-performance technical analysis in Rust.

TulipRS implements 100+ technical indicators and 77+ candlestick patterns with first-class SIMD acceleration. Process multiple assets or multiple parameter sets in a single CPU pass, stream live bars into stateful indicators without reprocessing history, and call everything from Rust or Python with the same universal API.

📖 Full documentation

Why TulipRS?

When optional intermediate outputs are needed (sub-EMAs, TR, AD line, etc.) TulipRS is 1.3× – 8.7× faster than running the equivalent TA-Lib calls.

Installation

Rust

Add TulipRS to your Cargo.toml. The crate is distributed via Git:

[dependencies]
tulip_rs = { git = "https://github.com/me60732/tulip_rs" }

Nightly required. TulipRS uses portable_simd. The correct nightly version is pinned automatically by rust-toolchain.toml — no manual toolchain management needed.

To disable the SIMD multi-asset / multi-option variants (reduces compile time):

tulip_rs = { git = "https://github.com/me60732/tulip_rs", default-features = false }

Python

pip install tulip-rs

Build from source with native CPU optimisations:

git clone https://github.com/me60732/tulip_rs_python
cd tulip_rs_python
RUSTFLAGS="-C target-cpu=native" maturin develop --release

Quick Start

Every indicator follows the same signature — learn it once, use it everywhere.

Rust

use tulip_rs::indicators::ema::indicator;

let close = vec![81.59, 81.06, 82.87, 83.00, 83.61,
                 83.15, 82.84, 83.99, 84.55, 84.36_f64];

// inputs: &[&[f64]]  |  options: &[f64]  |  optional_outputs: Option<&[bool]>
let (outputs, state) = indicator(&[close.as_slice()], &[5.0], None).unwrap();

println!("{:?}", outputs[0]); // EMA(5) values

// Streaming: feed new bars without reprocessing history
let new_bar = [85.10_f64];
let (next_outputs, next_state) = state.batch_indicator(&[&new_bar], None).unwrap();

Python

import numpy as np
import tulip_rs

close = np.array([81.59, 81.06, 82.87, 83.00, 83.61,
                  83.15, 82.84, 83.99, 84.55, 84.36], dtype=np.float64)

outputs, state = tulip_rs.indicators.ema.indicator([close], [5.0])
print(outputs[0])  # EMA(5) values

# Streaming
next_outputs, next_state = state.batch_indicator([np.array([85.10])], None)

SIMD — same indicator, N assets at once (Rust)

use tulip_rs::indicators::ema::indicator_by_assets;

// 4 assets processed simultaneously in one CPU pass
let inputs = [asset1.as_slice(), asset2.as_slice(),
              asset3.as_slice(), asset4.as_slice()];

let results = indicator_by_assets::<4>(&inputs, &[14.0], None).unwrap();

Benchmarks

Benchmarks compare TulipRS (Rust scalar, Rust SIMD) against the reference C implementation (Tulip Indicators) and TA-Lib across 8 real market symbols.

→ Benchmark results → How to run the benchmarks

Documentation

Language Support

License

MIT