If one of the mock.patch methods is used with autospec=True on a staticmethod in an object, the mock library determines that it is not callable by checking for the __call__ attribute. This results in a NonCallableMagicMock being returned which of course dies with the following error when the mocked method is called:

TypeError: 'NonCallableMagicMock' object is not callable

It seems that the create_autospec needs to special case for classmethod and staticmethod.

The following change seems to fix it, however I am only vaguely familiar with the internals of mock so I'm not sure what this breaks.

diff -r d356250e275d mock.py
--- a/mock.py	Tue Apr 09 14:53:33 2013 +0100
+++ b/mock.py	Wed Dec 17 07:35:15 2014 -0800
@@ -2191,7 +2191,8 @@
         # descriptors don't have a spec
         # because we don't know what type they return
         _kwargs = {}
-    elif not _callable(spec):
+    elif not _callable(spec) and not isinstance(spec, (staticmethod,
+                                                       classmethod)):
         Klass = NonCallableMagicMock
     elif is_type and instance and not _instance_callable(spec):
         Klass = NonCallableMagicMock

Here's a patch which does this. One problem could be that staticmethod(some_callable) or classmethod(some_callable) aren't callable per se, but given the fact that users expects Foo.staticmethod or Foo.klassmethod to be callable when patching them, it's something we should consider for fixing.

Looks like b6ea3dc89a78 is not a valid changeset. Could you attach a patch from the Mercurial repo?

Ups, sorry about that, I'll update the patch.

Regarding Claudiu's comment about staticmethod(x) or classmethod(x) not being callable, would it suffice to add a specific check of the form (isinstance(x, (classmethod, staticmethod)) and _callable(x.__func__))?

Separately, would it be better to include the check for staticmethod and classmethod objects (with an underlying callable) inside the _callable function? Not sure if this would break anything, but it seems like conceptually the issue is with the definition of a callable object, not the selection of mock type to use.

The attached patch implements these changes through _callable instead. This patch also ensures that the underlying object that staticmethod and classmethod wrap is a callable object as well.

Hi,

I hit this problem wile mocking one static method and found this fix.
Tested it and works for me.
However, I did not find it pushed anywhere: no Python 3.x and no mock for 2.7.

Is there any reason why it is not pushed anywhere, yet?

I planned to upgrade Chromium OS's mock module to 2.0.0. I also encountered the issue that classmethod cannot be patched as callable mock.

Reviewers, can we start the review process?

This affects all versions from 3.4 up to 3.8-dev. Would be nice if someone could do the review of the supplied patch.

Thanks for awesome work on Python!

I'm here because it just hit me also and I was for 1 h thinking that I don't know how to use patch/mock. ;)

I've just come across this too, so would be great if the patch can be progressed.

Here's a minimal example so my comment is not totally vacuous:

import unittest
from unittest import mock

class Foo:
    @classmethod
    def bar(cls, baz):
        pass

class TestFoo(unittest.TestCase):
    def test_bar(self):
        with unittest.mock.patch.object(Foo, "bar", autospec=True):
            Foo.bar()

->

~/› python -m unittest example.py 
E
======================================================================
ERROR: test_bar (example.TestFoo)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "example.py", line 14, in test_bar
    Foo.bar()
TypeError: 'NonCallableMagicMock' object is not callable

----------------------------------------------------------------------
Ran 1 test in 0.001s

Adding to the list of "I just ran into this". The patch submitted by fov seems straightforward enough: what can we do to help shepherd it along?

Thanks for the pings. I will work on this issue this weekend. Note that 3.4 and 3.5 are in security-fix-only mode now, so I removed them from the versions field.

Were you able to make any progress on this? Do you need any help?

@berker.peksag I have converted the patch at https://bugs.python.org/file40470/issue23078.patch and pushed it to a GitHub branch master...tirkarthi:bpo-23078 . I am willing to open a PR attributing to @fov in case you haven't had the time to look into this.

I am removing 3.6 since it's security fixes only mode.

Thanks

Please go ahead with the PR. I can't push this one through, but would be
great to have this finally land!

On Thu, 17 Jan 2019 at 03:42, Karthikeyan Singaravelan <
report@bugs.python.org> wrote:

Karthikeyan Singaravelan <tir.karthi@gmail.com> added the comment:

@berker.peksag I have converted the patch at
https://bugs.python.org/file40470/issue23078.patch and pushed it to a
GitHub branch
master...tirkarthi:bpo-23078 . I
am willing to open a PR attributing to @fov in case you haven't had the
time to look into this.

I am removing 3.6 since it's security fixes only mode.

Thanks

----------
nosy: +cjw296, mariocj89, xtreak
versions: -Python 3.6

---

Python tracker <report@bugs.python.org>
<https://bugs.python.org/issue23078\>

---

New changeset 9b21856 by Berker Peksag (Xtreak) in branch 'master':
bpo-23078: Add support for {class,static}method to mock.create_autospec() (GH-11613)
9b21856

New changeset 15a57a3 by Berker Peksag in branch '3.7':
bpo-23078: Add support for {class,static}method to mock.create_autospec() (GH-11613)
15a57a3

Thank you!