Thursday 2:00 PM–2:40 PM in Central Park West (#6501)

pandas : Integer NA as a first class citizen

Jeff Reback

Audience level:
Intermediate

Description

pandas has recently gained the ability to extend dtypes with an interface called Extension Arrays. I will use an in depth example to show how this enables a long time requeasted feature of pandas, support for Integer NA in a first class way.

Abstract

pandas has recently gained the ability for authors to extend dtypes and arrays with an interface called Extension Arrays.

These allow pandas core structures to be extended to contain an array-like that is a novel and/or performant first class data representation. For example, pandas Categorical, Period, Interval, and Datetime w/tz are currently implemented in this way.

I will cover the why’s and the how’s of Extension Arrays. Then I will use Extension Arrays to illustrate an in depth example of how this enables a long requested feature in pandas, support for Integer NA in a first class way.

Next I will discuss other extension types external to pandas. Authors have implemented containers for geopandas, cyberpandas and the fletcher array

Finally I will show how Extension Arrays enable a bright future for pandas.

Subscribe to Receive PyData Updates

Subscribe