TulipRS

High-performance technical analysis in Rust.

TulipRS is a production-ready library implementing 100+ technical indicators and 77+ candlestick patterns, with first-class SIMD acceleration, stateful streaming, and native bindings for C, Python, Node.js, Go, and the browser (WASM).

---

Why TulipRS?

Most technical analysis libraries are wrappers around the same scalar C code written decades ago. TulipRS is built differently, with several capabilities that compound into meaningfully faster and more practical pipelines:

SIMD acceleration — rather than looping over one asset or one parameter set at a time, TulipRS can process N assets or N option sets simultaneously in a single CPU pass using portable SIMD intrinsics. On AVX2 hardware that is 4× the throughput for the same wall-clock time. See SIMD.

Stateful streaming — built for live systems — every indicator returns an IndicatorState alongside its outputs. Feed it new bars as they arrive and computation resumes from where it left off — no reprocessing of historical data, no O(n) cost per tick. State is fully serialisable to JSON and other Serde formats for persistence across restarts. See State Management.

Optional outputs at no extra cost — many indicators compute intermediate series (sub-EMAs, TR, AD line, etc.) as a natural part of their calculation. TulipRS can return those alongside the primary output in the same pass. C Tulip and TA-Lib require a separate function call — and a full extra data scan — for each one. TulipRS is 1.3× – 8.7× faster when you need those intermediate values. See Indicator API.

---

Browser-native via WebAssembly — the full indicator library compiles to a WebAssembly module via wasm-pack, bringing the same computation to any modern browser with zero server round-trips and zero native dependencies. Pair it with tulip-rs-lwc for drop-in TradingView Lightweight Charts overlays and oscillators, or use tulip-rs-wasm directly for headless browser-side computation. See Live Demo.

---

Performance

TulipRS isn't just SIMD-accelerated in theory — it's measurably faster in practice, benchmarked against C Tulip, TA-Lib, pure-Rust libraries (kand, RustTa), pandas-based Python libraries, pure-Go libraries, and popular Node.js TA packages:

• ~22× faster than Python's ta library, up to 20,802× on select indicators
• ~46× faster than Node's technicalindicators, up to 403× on select indicators
• ~2,283× faster than Go's cinar/indicator, up to 7,491× on select indicators
• Up to 11.93× faster than TA-Lib, the industry-standard C library
• Beats every pure-Rust TA-Lib-inspired competitor on 100% of compared indicators — up to 36.10× faster than kand (mfi) and up to 6.78× faster than RustTa (ef)
• Up to 3.85× additional throughput from SIMD batch processing

📊 Full benchmark methodology and results

---

Features at a Glance

Capability	Detail
Indicators	100+ (moving averages, oscillators, trend, volatility, volume, price/statistical)
Candlestick Patterns	60+ patterns with bullish/bearish forecasting
SIMD — by assets	Same options applied to N assets in one CPU pass (indicator_by_assets::<N>)
SIMD — by options	N option sets applied to one asset in one CPU pass (indicator_by_options::<N>)
State management	Every indicator returns a serialisable IndicatorState for streaming / incremental use
Browser / WASM	Full indicator set compiled to WebAssembly — runs in any modern browser, no server needed (tulip-rs-wasm)
Languages	Rust (native), C (tulip_rs_ffi — hand-rolled extern "C" FFI), Python (tulip_rs_python via PyO3), Node.js (tulip-rs-node via napi-rs), Go (tulip_rs_go via cgo), Browser (WASM)

---

Quick Example

RustGoJavaPythonNode.jsWASM

use tulip_rs::indicators::sma::{Sma, 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];

let (outputs, state) = Sma::indicator(&[close.as_slice()], &[5.0], None).unwrap();

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

import "github.com/me60732/tulip_rs_go/indicators"

close := []float64{81.59, 81.06, 82.87, 83.00, 83.61,
                   83.15, 82.84, 83.99, 84.55, 84.36}

res, st, _ := indicators.Sma.Indicator(close, []float64{5.0}, nil)
defer res.Close()
defer st.Close()

fmt.Println(res.Rows[0]) // SMA(5) values — zero-copy views

import org.tuliprs.*;
import org.tuliprs.indicators.Sma;

double[] close = {81.59, 81.06, 82.87, 83.00, 83.61,
                  83.15, 82.84, 83.99, 84.55, 84.36};

Outcome oc = Sma.indicator(new double[][] {close}, new double[] {5.0});
try (Result res = oc.result(); State st = oc.state()) {
    System.out.println(java.util.Arrays.toString(res.toDoubleArray(0))); // SMA(5) values — zero-copy views
}

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.sma.indicator([close], [5.0])

print(outputs[0])  # SMA(5) values

import * as ti from 'tulip-rs-node';

const close = Float64Array.from([81.59, 81.06, 82.87, 83.00, 83.61,
                                 83.15, 82.84, 83.99, 84.55, 84.36]);

const [outputs, state] = ti.sma.indicator([close], [5]);

console.log(outputs[0]); // SMA(5) values

import { init } from 'tulip-rs-wasm';
import * as ti from 'tulip-rs-wasm';

await init(); // bundler resolves the WASM asset automatically

const close = [81.59, 81.06, 82.87, 83.00, 83.61,
               83.15, 82.84, 83.99, 84.55, 84.36];

const [outputs, state] = ti.sma.indicator([close], [5]);

console.log(outputs[0]); // SMA(5) values

---

Documentation Pages

Page	Description
Getting Started	Installation, feature flags, calling convention, and first examples
Indicators — Overview	Full indicator index with inputs, options, and output counts
Moving Averages	SMA, EMA, WMA, DEMA, TEMA, TRIMA, HMA, ZLEMA, KAMA, VIDYA, VWMA, Wilders
Oscillators	RSI, MACD, Stoch, StochRSI, Williams %R, CCI, CMO, UltOsc, AO, Fisher, FOSC, MSW
Trend	PPO, APO, ADX, ADXR, DM, DI, DX, Aroon, AroonOsc, PSAR
Volatility	BBands, ATR, NATR, TR, StdDev, Volatility, VHF, CVI
Volume	AD, ADOSC, OBV, MFI, NVI, PVI, VOSC, KVO, EMV, WAD
Price & Statistical	AvgPrice, MedPrice, TypPrice, WCPrice, Max, Min, MOM, ROC, ROCR, BOP, LinReg, TSF, TRIX, DPO, Mass, MD, MarketFi, QStick, PivotPoint
Candlestick Patterns	60+ patterns, forecast types, Rust, C, Python, and Node.js usage
SIMD	Conceptual overview: by-assets and by-options modes, lane counts, when to use each
State Management	Streaming computation, chunked processing, JSON serialisation
Language Bindings	Per-binding subpages: C FFI, Python (PyO3/maturin), Node.js (napi-rs), Go (cgo), Browser (WASM)
Benchmarks	Comparison against Tulip Indicators (C) and TA-Lib, methodology, how to run

---

Language Support

Language	Status	Package
Rust	✅ Native	tulip_rs (this crate)
C	✅ Supported	tulip_rs_ffi — hand-rolled extern "C" FFI, usable from C, C++/any C ABI
Go	✅ Supported	tulip_rs_go — cgo bindings over the extern "C" FFI, zero-copy outputs
Java	✅ Supported	tulip-rs-java — JDK 22+ Foreign Function API
Python	✅ Supported	tulip_rs_python
Node.js	✅ Supported	tulip-rs-node
Browser (WASM)	✅ Supported	tulip-rs-wasm · tulip-rs-lwc