A pure fast JavaScript/TypeScript library of technical‑analysis indicators, trading performance/risk metrics, array utilities, and numerical helpers.

This package implements several TA indicators (EMA, TEMA, T3, MFI, KAMA, etc.), common trading performance metrics/utilities, vectorized math functions, and statistical helpers.

By default the main (typed) build returns typed arrays (e.g. Float64Array) for better numeric performance and predictable memory layout. A companion "untyped" build exposes the same API but returns plain number[] values for easier interoperability with plain JavaScript code.

The library has no runtime dependencies. It can be used in browser web applications or in Node.js environments that support ESM imports.

Quick Start

Install:

``bash
npm install qntjs-lib
`

Basic usage (default — typed output):

`js
import { ta } from 'qntjs-lib';

const prices = [1,2,3,4,5,6,7];
// returns Float64Array by default (typed numeric output)
const out = ta.ema(prices, 3);
`

Basic usage (untyped — plain arrays):

`js
import { ta } from 'qntjs-lib/untyped';

const prices = [1,2,3,4,5,6,7];
// returns plain number[] (easier to inspect/serialize)
const out = ta.ema(prices, 3);
`

When to use each:
- Use the default import (qntjs-lib) when you want outputs as Float64Array for numeric performance and predictable memory layout.
- Use qntjs-lib/untyped when you prefer plain number[] outputs for easier inspection or serialization.

Modules

Overview of top-level modules and minimal examples showing common usage patterns.

$3

Example: compute an exponential moving average (EMA)

`js
import { ta } from 'qntjs-lib';
const prices = [1,2,3,4,5,6,7];
const ema3 = ta.ema(prices, 3); // Float64Array
`

$3

Example: elementwise subtract and scale

`js
import { math } from 'qntjs-lib';
const a = [1,2,3];
const b = [0.1,0.1,0.1];
const diff = math.sub(a, b); // Float64Array of a-b
const scaled = math.scale(diff, 100);
`

$3

Example: compute daily returns, Sharpe, and parametric VaR

`js
import { perf } from 'qntjs-lib';
const prices = [100, 110, 105, 120];
const rets = perf.returns(prices); // simple returns (Float32Array)
const daily = perf.dailyReturns([Date.now(), Date.now() + 86400000], [0.01, 0.02]);
const sr = perf.sharpe([0.01, -0.02, 0.03]);
const varP = perf.valueAtRisk([0.01, -0.02, 0.03], 0.05, 'parametric');
`

$3

Example: quantile and sample

`js
import { stats } from 'qntjs-lib';
const v = stats.quantile([1,2,3,4,5], 0.1);
const sample = stats.sample([1,2,3,4,5], 3);
`

$3

Example: drop NaNs and forward-fill

`js
import { arr } from 'qntjs-lib';
const a = [NaN, 1, NaN, 2];
const clean = arr.dropna(a);
const filled = arr.ffill(a);
`

List of available API

- arr.* : isna, notna, fillna, ffill, bfill, replace, dropna, allna, equals, countna, havena, lag

- math.* : add, sub, avg, mul, div, scale, abs, sign, round, floor, ceil, eq, neq, gt, gte, lt, lte, and, or, not, clamp, sum, prod, min, max, argmin, argmax, cumsum, cumprod, cummax, cummin, rollsum, rollmin, rollmax, rollminmax, rollprod, rollargmin, rollargmax, diff, randuniform, randnormal, dot, norm, ols, olsMulti

- stats.* : mean, hmean, gmean, mad, skew, kurtosis, median, quantile, percentiles, var, covar, stdev, corr, zscore, norminmax, winsorize, sample, shuffle, bootstrap, rollmean, rollmedian, rollquantile, rollvar, rollcovar, rollstdev, rollcorr, rollskew, rollkurtosis

- ta.* : dema, ema, hma, kama, sma, wma, vwma, trima, t3, tema, rma, ao, apo, aroon, change, cmo, kst, macd, mom, ppo, roc, rsi, stoch, stochrsi, ultosc, wpr, supertrend, adx, adxr, dx, cci, di, dpo, ichimoku, psar, atr, tr, natr, bb, bbw, donchian, keltner, adosc, obv, pnvi, wad, ad, mfi, cross, crossover, crossunder, rising, falling

- perf.* : returns, logreturns, cumreturns, cagr, dailyreturns, dd, maxdd, maxddDetails, dduration, rollmaxdd, recoveryfactor, calmar, ulcer, rollulcer, sharpe, sortino, rollsharpe, rollsortino, vol, rollvol, var (valueAtRisk), expshortfall, tail, omega

skipna and dense fast-path

Where applicable, implementations are optimized for two common usage patterns:

- NaN‑aware workflows (default): functions are NaN‑aware and will skip NaN values where appropriate.
- Dense fast‑path: when you know inputs contain no NaNs you can opt into a dense, faster implementation by passing skipna=false to supported functions.

`js
// NaN-aware (default)
ta.sma(pricesWithGaps, 5);

// Dense fast-path (assume no NaNs)
ta.sma(densePrices, 5, false);
`

Tests & development

Many functions, especially TA indicators are tested for correctness against Tulind library.

Run tests:

`bash
npm test
`

Build

`bash
npm run build
``

License

This project is licensed under the terms of the MIT license