Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

array_equal(a, b, equal_nan=True) throws errors for array with non-numeric values

See original GitHub issue

Using the newly added (#16128) np.array_equal keyword argument equal_nan fails when the values of the passed array aren’t accepted by np.isnan. This occurs for values like strings.

Reproducing code example:

import numpy as np
np.array_equal(
    np.array(list("abc")), np.array(list("abc")), equal_nan=True
)

Error message:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<__array_function__ internals>", line 6, in array_equal
  File "/Users/isaac/github/numpy/numpy/core/numeric.py", line 2381, in array_equal
    a1nan, a2nan = isnan(a1), isnan(a2)
TypeError: ufunc 'isnan' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''

Numpy/Python version information:

This occurs on both master and 1.19.0rc1

1.20.0.dev0+3f11db4 3.7.7 (default, Apr 25 2020, 15:09:12) 
[Clang 11.0.3 (clang-1103.0.32.59)]

1.19.0rc1 3.7.7 (default, May  6 2020, 04:59:01) 
[Clang 4.0.1 (tags/RELEASE_401/final)]

Issue Analytics

State:
Created 3 years ago
Reactions:1
Comments:12 (12 by maintainers)

Top GitHub Comments

2reactions

sebergcommented, May 26, 2020

Yes, lets ignore object here for now, its trickier.

String I think we can just propose, may need a bit of discussion (or maybe digging up the old discussion when we added bools and ints). We have no string ufuncs (this is mainly because you have to set the output string length correctly, which made, also me, always think that its tricky). The other point is that string promotion is still broken (weakly typed) which could lead to strange things (I would like to move that concat + string promotion deprecation forward!). But this is a unary ufunc, with boolean output that is always False, so neither of those issues would stand in the way.

1reaction

anirudh2290commented, May 26, 2020

okay thanks @seberg for the guidance ! So @ivirshup if you are interested in proposing something for string support for isnan that would be great here. Maybe some of the discussion topics that @seberg mentioned may be : https://github.com/numpy/numpy/pull/12988 and the xref’d issues.

Given the current discussion, I think we should close the current PR #16378 .

Top Results From Across the Web

ARRAY_EQUAL - L3HarrisGeospatial.com

The ARRAY_EQUAL function is a fast way to compare data for equality in situations where the index of the elements that differ are...

numpy.array_equal — NumPy v1.24 Manual

Returns True if two arrays are element-wise equal within a tolerance. array_equiv. Returns True if input arrays are shape consistent and all elements...

Comparing numpy arrays containing NaN - Stack Overflow

I just added np.testing.assert_equal(a,b) in my unittest, and if it raises the exception, the test fails (no error), and I ...

array_equals

With this function you can check to see if two arrays are equal (equivalent or the ... which will not check to see...

How can I set a matrix value to equal a non-numeric coordinate?

For example, if part of a matrix equals 3, I want that coordinate to be (a,b). Then I want to be able to...