quantalib

447 technical analysis indicators compiled to native code via .NET NativeAOT, called from Python through ctypes. Same SIMD-accelerated engine as the QuanTAlib .NET package. Zero Python math reimplementation.

pip install quantalib

Quick Start

import numpy as np
import quantalib as qtl

close = np.random.default_rng(42).normal(100, 2, size=500)

sma = qtl.sma(close, period=20)
rsi = qtl.rsi(close, period=14)
upper, mid, lower = qtl.bbands(close, period=20, std=2.0)

Works with pandas, polars, and pyarrow — same-type-in, same-type-out:

# pandas — preserves index
import pandas as pd
s = pd.Series(close, name="close")
rsi = qtl.rsi(s, period=14)        # → pd.Series

# polars — zero-copy-friendly
import polars as pl
s = pl.Series("close", close)
rsi = qtl.rsi(s, period=14)        # → pl.Series
bb  = qtl.bbands(s, period=20)     # → pl.DataFrame (upper, mid, lower)

# pyarrow — for Arrow-native pipelines
import pyarrow as pa
a = pa.array(close, type=pa.float64())
rsi = qtl.rsi(a, period=14)        # → pa.Array

pandas-ta users: length= is accepted everywhere as an alias for period=.

Install optional backends:

pip install quantalib[pandas]       # pandas / pd.Series support
pip install quantalib[polars]       # polars / pl.Series support
pip install quantalib[pyarrow]      # pyarrow / pa.Array support
pip install quantalib[all]          # all three

Performance (500,000 bars)

quantalib calls the same SIMD-optimized NativeAOT core as the .NET package through a single ctypes call (5-15 μs overhead). wickra (Rust/PyO3) returns stdlib array.array; pandas-ta stays in pure Python/numpy. For arrays above a few hundred bars, NativeAOT wins by an order of magnitude or more.

Categories

Requirements

• Python 3.10+
• NumPy >= 1.24

Pre-built wheels — the NativeAOT core is compiled per platform in CI and ships inside the wheel, so there is no source build and no C++ toolchain:

Optional dependencies

License

Apache License 2.0