ahaldane/ndarray_ducktypes

A set of ndarray ducktypes testing the new __array_function__ and __array_ufunc__ functionality in numpy.

ndarray_ducktypes's Issues

MaskedArray TODO list

TODO list:

• Complete writing tests
  • tests for parsing input with X values (mixtures of ndarrays, X, scalars)
  • tests for each npy-api method
  • tests of array string representations
  • test ufuncs
  • test constructor
• Finish npy-api functions still planned for implementation
  • np.cov, np.correlate
  • np.einsum_path
  • np.apply_along_axis, np.apply_over_axes
  • np.rot90
  • np.fliplr, np.flipud
  • np.pad (? may be too much effort to implement)
  • np.piecewise
  • np.ediff1d, np.gradient
  • np.unwrap
• Finish writing docs about how to create composite MaskedArray ducktypes
• Add note about new method '.count` in docs
• Planning for future configurability: How to provide optional NA mask style? How to give optional ufunc auto-mask-invalid behavior. Should it be by MaskedArray factory? submodule? configuration method?

Possible further npy-api methods to implement: np.is_busday, np.busday_offset, np.busday_count, np.datetime_as_string, np.asfarray, np.vander, np.tril_indices_from, np.triu_indices_from, np.sort_complex, np.trim_zeros, np.i0, np.msort, np.trapz, np.ix_, np.fill_diagonal, np.diag_indices_from, np.lib.scimath.sqrt, np.lib.scimath.log, np.lib.scimath.log10, np.lib.scimath.logn, np.lib.scimath.log2, np.lib.scimath.power, np.lib.scimath.arccos, np.lib.scimath.arcsin, np.lib.scimath.arctanh, np.poly, np.roots, np.polyint, np.polyder, np.polyfit, np.polyval, np.polyadd, np.polysub, np.polymul, np.polydiv, np.intersect1d, np.setxor1d, np.in1d, np.isin, np.union1d, np.setdiff1d, np.fv, np.pmt, np.nper, np.ipmt, np.ppmt, np.pv, np.rate, np.irr, np.npv, np.mirr, np.save, np.savez, np.savez_compressed, np.savetxt,

ArrayCollection TODO list

TODO list:

• complete most npy-api functions