mantis-ta

Composable technical analysis and strategy engine for Rust.

---

Pure Rust technical indicators with a type-safe strategy composition API. No C dependencies. No FFI. No unsafe in the default build.

Every indicator is verified against TA-Lib reference outputs.

[dependencies]
mantis-ta = "0.5.3"

Quick Start

Indicators — Streaming

Feed candles one at a time. Get values out. O(1) per update, zero heap allocations in the hot path.

use mantis_ta::prelude::*;

let mut ema = EMA::new(20);
let mut rsi = RSI::new(14);

for candle in candles.iter() {
    if let Some(ema_val) = ema.next(candle) {
        println!("EMA(20) = {:.2}", ema_val);
    }
    if let Some(rsi_val) = rsi.next(candle) {
        println!("RSI(14) = {:.2}", rsi_val);
    }
}

Indicators — Batch

Compute over a full series at once. Returns Vec<Option<f64>> aligned with input candles (None during warmup).

use mantis_ta::prelude::*;

let sma_values = SMA::new(50).calculate(&candles);
let bb_values = BollingerBands::new(20, 2.0).calculate(&candles);

for (i, bb) in bb_values.iter().enumerate() {
    if let Some(bb) = bb {
        println!("Bar {}: Upper={:.2} Mid={:.2} Lower={:.2}", i, bb.upper, bb.middle, bb.lower);
    }
}

Strategy Composition

Define complete trading strategies as composable, type-checked rules. Invalid strategies don't compile.

use mantis_ta::prelude::*;
use mantis_ta::strategy::*;

let strategy = Strategy::builder("Golden Cross Momentum")
    .timeframe(Timeframe::D1)
    .entry(
        all_of([
            ema(20).crosses_above(ema(50)),
            rsi(14).is_between(40.0, 65.0),
            volume().is_above(volume_sma(20).scaled(1.5)),
        ])
    )
    .exit(
        any_of([
            ema(20).crosses_below(ema(50)),
            rsi(14).is_above(80.0),
        ])
    )
    .stop_loss(StopLoss::atr_multiple(14, 2.0))
    .take_profit(TakeProfit::atr_multiple(14, 3.0))
    .max_position_size_pct(5.0)
    .build()?;

// Evaluate against historical data
let signals: Vec<Signal> = strategy.evaluate(&candles)?;

// Or stream live — same strategy, bar by bar
let mut engine = strategy.into_engine();
for candle in live_feed {
    match engine.next(&candle) {
        Signal::Entry(Side::Long)  => { /* open long */ },
        Signal::Exit(reason)       => { /* close position */ },
        Signal::Hold               => { /* wait */ },
        _ => {}
    }
}

Backtesting

Honest simulation with realistic slippage, commissions, and next-bar execution.

use mantis_ta::backtest::*;

let result = backtest(&strategy, &candles, &BacktestConfig::default())?;

println!("Return:       {:.2}%", result.metrics.total_return_pct);
println!("Sharpe Ratio: {:.2}",  result.metrics.sharpe_ratio);
println!("Max Drawdown: {:.2}%", result.metrics.max_drawdown_pct);
println!("Win Rate:     {:.2}%", result.metrics.win_rate_pct);
println!("Trades:       {}",     result.metrics.total_trades);

Available Indicators

Trend

v0.5.0 Batch A: SMA · EMA · WMA · DEMA · TEMA · MACD · ADX Future: Ichimoku · Parabolic SAR · Supertrend

Momentum

v0.5.0 Batch A: RSI · Stochastic · CCI · Williams %R · ROC Future: MFI

Volatility

v0.5.0 Batch A: Bollinger Bands · ATR · Standard Deviation Future: Keltner Channels

Volume

OBV · Volume SMA · VWAP · Accumulation/Distribution

Support/Resistance

Pivot Points · Donchian Channels · Fibonacci Retracement

See the full indicator list in the API docs.

Features

[dependencies]
mantis-ta = { version = "0.5", features = ["strategy", "backtest"] }

Feature	Default	Description
serde	✓	Serialize strategies, indicators, and results to JSON
strategy	✓	Strategy composition engine (v0.2.0+)
backtest	✓	Backtesting engine with metrics (v0.4.0+)
ndarray		Interop with the ndarray ecosystem
full-indicators		All 50+ indicators (default includes 30 most common)
simd		SIMD-accelerated batch computation (uses unsafe)
all		Everything

Design Principles

• Correctness first. Every indicator verified against TA-Lib (< 1e-10 relative error).
• Streaming-first. O(1) incremental updates for live data. Batch is also first-class.
• Zero allocation in the hot path. next() never heap-allocates.
• No unsafe by default. Safe Rust is fast enough.
• Type system enforces validity. A strategy without a stop-loss is a compile error, not a runtime surprise.
• Honest backtesting. No lookahead bias. Slippage and commissions are mandatory, not optional.

Performance

Benchmarked on Apple M-series, single core:

Operation	Time
EMA(20) per bar (streaming)	< 100 ns
RSI(14) batch, 2000 bars	< 15 µs
Strategy eval (5 conditions), 2000 bars	< 200 µs
Full backtest, 2 years daily	< 5 ms

Run benchmarks yourself: cargo bench

Custom Indicators

Implement the Indicator trait to create your own:

use mantis_ta::prelude::*;

pub struct MyIndicator {
    period: usize,
    buffer: Vec<f64>,
}

impl Indicator for MyIndicator {
    type Output = f64;

    fn next(&mut self, candle: &Candle) -> Option<Self::Output> {
        self.buffer.push(candle.close);
        if self.buffer.len() < self.period {
            return None;
        }
        // Your calculation here
        Some(self.buffer.iter().sum::<f64>() / self.period as f64)
    }

    fn warmup_period(&self) -> usize { self.period }
    fn reset(&mut self) { self.buffer.clear(); }
    fn clone_boxed(&self) -> Box<dyn Indicator<Output = Self::Output>> {
        Box::new(self.clone())
    }
}

Contributing

Contributions welcome! Please read CONTRIBUTING.md before opening a PR.

Adding a new indicator? See the Contributor Guide for the full checklist: implement the trait, add TA-Lib verification, write benchmarks, document it.

License

Licensed under either of:

• MIT License
• Apache License, Version 2.0

at your option.

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this crate shall be dual-licensed as above, without any additional terms or conditions.