TA-Lib Python Wrapper

TA-Lib is a Python wrapper for the TA-Lib C library, implemented using Cython and NumPy instead of SWIG. It is designed for high-performance technical analysis of financial market data.

Key features include:

• Over 150 technical indicators (e.g., ADX, MACD, RSI, Stochastic, Bollinger Bands).
• Candlestick pattern recognition.
• High efficiency: Results are typically 2-4 times faster than the original SWIG-based interface.
• Support for numpy, pandas, and polars libraries.

The TA-Lib Python documentation is organized into several key areas to help you integrate technical analysis indicators into your workflow:

• Setup: Follow the Installation and Troubleshooting guide to get the library running.
• API Usage: Choose between the Function API for direct indicator calls or the Abstract API for a different interface style.
• Indicator Reference: A complete list of all available functions is provided in All Functions, categorized by indicator type (e.g., Momentum, Volatility, Pattern Recognition).

Price Transform Functions are used to calculate new price points based on existing price data (Open, High, Low, Close). These functions transform raw price inputs into single values representing a specific price metric.

# Example of using price transform functions
real = AVGPRICE(open, high, low, close)
real = MEDPRICE(high, low)

Users often expect STOCHRSI to behave like Stochastic applied to RSI, but STOCHRSI is actually STOCHF (Fast Stochastic) applied to RSI. If you want the standard Stochastic applied to RSI, use STOCH instead.

import talib
import numpy as np
c = np.random.randn(100)

# Standard library function (Fast Stochastic applied to RSI)
k, d = talib.STOCHRSI(c)

# Equivalent to STOCHRSI
rsi = talib.RSI(c)
k, d = talib.STOCHF(rsi, rsi, rsi)

# What you likely actually want (Stochastic applied to RSI)
rsi = talib.RSI(c)
k, d = talib.STOCH(rsi, rsi, rsi)

import talib
import numpy as np
c = np.random.randn(100)

# this is the library function
k, d = talib.STOCHRSI(c)

# this produces the same result, calling STOCHF
rsi = talib.RSI(c)
k, d = talib.STOCHF(rsi, rsi, rsi)

# you might want this instead, calling STOCH
rsi = talib.RSI(c)
k, d = talib.STOCH(rsi, rsi, rsi)

Functions documented as having an unstable period start with an unstable-period setting of 0. To track these periods, you must explicitly set them using set_unstable_period(). Once set, get_unstable_period() will return a non-zero value.

# Example pattern for handling unstable periods
talib.set_unstable_period(0) # Set explicitly
period = talib.get_unstable_period()

Every function in the Abstract API has an .info property that provides a dictionary containing metadata about the indicator. This is useful for discovering required input names, available parameters, and expected output names.

Metadata includes:

• name: The technical name of the indicator.
• display_name: The human-readable name.
• group: The category of the indicator (e.g., 'Momentum Indicators').
• input_names: An OrderedDict of the required input keys and their default values.
• parameters: An OrderedDict of the function's parameters and their default values.
• output_names: A list of the names of the returned arrays.

from talib import abstract

# Access metadata for the Stochastic function
print(abstract.Function('stoch').info)

The underlying TA-Lib C library handles NaN values by typically propagating them to the end of the output array. This behavior differs from libraries like Pandas, which might output NaN only until the lookback period is satisfied.

Example of TA-Lib behavior: If an input array contains a NaN, the subsequent values in the output may also become NaN depending on the indicator's calculation logic.

>>> c = np.array([1.0, 2.0, 3.0, np.nan, 4.0, 5.0, 6.0])
>>> talib.SMA(c, 3)
array([nan, nan,  2., nan, nan, nan, nan])

The version of ta-lib you use depends on your numpy version:

• For numpy < 2: Use ta-lib < 0.5.
• For numpy >= 2: Use ta-lib >= 0.5.

Because the upstream TA-Lib C library changed its library name from -lta_lib to -lta-lib in version 0.6.1, this project maintains three distinct feature branches to ensure compatibility with specific versions of the C library and NumPy: