For when your data won't fit in your dataframe
Akimbo provides fast vectorized processing of nested, ragged data
in dataframes, using the .ak accessor.
- numpy-like API for slicing and accessing data deep in nested structures
- apply any function, ufunc or aggregation at any level
- use with different backends: pandas, polars, dask-dataframe and cuDF exactly the same way (more backends may come in the future)
- CPU and GPU processing and support for
numba-jit - attach object-like behaviours to your record (sub)structures
See the quick start in the documentation for an introduction to akimbo.
Support for this work was provided by NSF grant OAC-2103945.