Skip to content

Latest commit

 

History

History
19 lines (11 loc) · 1.89 KB

README.md

File metadata and controls

19 lines (11 loc) · 1.89 KB

NDarray Ducktypes

This module provides a set of numpy ndarray ducktypes which use the new __array_function__ and __array_ufunc__ functionality in numpy. Currently, this includes a more modern MaskedArray replacement, and a structured-array-like ducktype called ArrayCollection meant to allow simple tabular data analysis.

This module also provides helper tools for defining your own new ndarray-ducktypes, including various boilerplate methods and ready-made ducktype printing functionality, in the duckprint submodule. Getting the str and repr of ndarray ducktypes to mirror that of numpy's ndarrays is not trivial, and the duckprint module's goal is to make it easy.

This project is currently in "alpha" state, and may still go through significant reorganization.

Documentation

The main documentation has an overview of this modules functionality, including installation instructions. Particular ducktypes and the ducktype-helper methods are documented at:

Contributing

PRs and issues are welcome. If you would like to help, right now a good way is to write tests for the MaskedArray numpy-api implementations in tests/test_MaskedArray.py, and fixing problems with the api implementations you find. As of writing this, over 150 api functions are implemented, but only a small fraction have a dedicated test, see Issue #1. There are already many tests ported from numpy.ma, but I want to go through each api method one-by-one and add tests for it in the TEST_API class. The easiest way to run tests, without needing to "pip install", is run "python -m pytest" from the clone root.