In a Python data analysis environment, the numpy library is of central importance. It has a native persistence form for numeric arrays: .npy files (compressed .npz). It would be useful to have the ability to read, and perhaps write, files in this format.

https://docs.scipy.org/doc/numpy/neps/npy-format.html

This could be done in Java code, as the file format is documented (though some of it may be difficult to reproduce exactly in Java), or via a Python bridge as we will do for the native LSST datatypes.

This ticket at the moment is only intended to stick a pin in the idea, not to mandate its implementation. It may be that we should focus our attention on the cross-language Apache Arrow project instead.