fabioibanez / pudu

an open source Two Sigma Flint summarizers implementation in Polars

Pudu

What?

Pudu is an open source library that implements, with inspiration from Two Sigma's Flint, a summarizer class to do feature engineering on Polars dataframes. With the summarizer class primitives my goal is to implement fast time-series analysis for financial instruments.

Pudu is only a piece of the puzzle. To fully utilize the full performance of Polars, Pudu will be Quokka compatible. Read more about Quokka here.

pudu's Issues

Implement time-series indicators summarizer

Implement a summarizer for indicators; indicator_summarizer

• SMA - note that this just leverages a window summarizer.
• BBANDS
• RSI
• MACD
• FORCE INDEX
• AVERAGE TRUE RANGE

Implement window_summarizer

Time-series analysis requires an interface for windows.

window_summarizer must be able to perform the primitive operations of summarizer factory (and more) on a window of rows.

For many indicators, like SMA, all you really need is a window_summarizer() for the polars mean.