strip_zeros kwarg no longer available in FloatFormat

Reproducing code example:

I’ve been using this context manager in some code to control array printing:

import contextlib

import numpy as np
import numpy.core.arrayprint as arrayprint


@contextlib.contextmanager
def printoptions(strip_zeros=False, **kwargs):
    """Allows you to set NumPy array print formatting locally.
    Taken from:
    http://stackoverflow.com/questions/2891790/pretty-printing-of-numpy-array
    """

    origcall = arrayprint.FloatFormat.__call__

    def __call__(self, x, strip_zeros=strip_zeros):
        return origcall.__call__(self, x, strip_zeros)

    arrayprint.FloatFormat.__call__ = __call__
    original = np.get_printoptions()
    np.set_printoptions(**kwargs)
    yield
    np.set_printoptions(**original)
    arrayprint.FloatFormat.__call__ = origcall

FloatFormat was removed in NumPy 1.18.0 but prior to that https://github.com/numpy/numpy/pull/9941 removed the strip_zeros kwarg from FloatFormat.__call__. This seems to be a backwards incompatible change. It also seems like that kwarg likely did not have a unit test.

Regardless, I’m trying to figure out how to reproduce the behavior of strip_zeros=False prior to PR #9941.

Error message:

Numpy/Python version information:

In [10]: import sys, numpy; print(numpy.__version__, sys.version)                                                                           
1.18.1 3.6.10 | packaged by conda-forge | (default, Apr  6 2020, 14:52:36) 
[GCC 7.3.0]

[moorepants]

This has been inside NumPy for 3 releases now in that case. I am a bit curious whether this is more something you noticed, or something that actually creates issues for you?

I have a stable library that I developed a decade ago and I update it about once a year now to ensure it works with the recent libraries. I’m behind on the update a bit but it is blocking some travis builds and I need to come up with a work around. I’m mostly looking for that, not expecting you all to undo changes.

[rossbar]

The discussion in #16869 turned up the source of the issue, see #11924 . I will close this for now as picking up the linked NEP seems like the best way for preventing issues like this in the future.