GeezerVison

this sets a mutable default value which does not work as you would expect in python -- please don't change this, the original argument handling was correct

>>> def f(x=[]):
...     x.append('wat')
...     print(x)
... 
>>> f(['hi'])
['hi', 'wat']
>>> f(['hi'])
['hi', 'wat']
>>> f()
['wat']
>>> f()
['wat', 'wat']
>>> f()
['wat', 'wat', 'wat']
>>> f()
['wat', 'wat', 'wat', 'wat']
>>> f()
['wat', 'wat', 'wat', 'wat', 'wat']

Yep, I'm aware of the behavior of mutable defaults in python. In this case it's not a problem because the function will always be called only once and without arguments by setuptools machinery. However, I agree that using it may be fragile, for example, in tests.

this shouldn't be needed -- use the arguments that are passed from pre-commit directly (you'll want to set up an argparse.ArgumentParser -- see how the rest of the tools do this)

Initially I thought that too and wanted to rewrite it using git status instead of git diff to query mode. However, the hook is called "forbid new submodules" which apparently means that bumping somehow already existent submodules should be allowed. That's why we need --diff-filter=A and we need to query PRE_COMMIT_*_REF variables in order to ensure that no submodules have been added in any commits being validated, not only the last one.

use mock.patch.dict, though I think this code will be deleted anyway

pre-commit will never call this with zero arguments

Yeah, I've added it because of git diff or git status behavior difference with one or more path arguments (check them) vs. zero (check all, not none). But I agree that this is not needed.

Tuple[str, ...]

why are you capitalizing local variables

Fixed FROM and TO, preserved REV_PARSE_HEAD because it's a constant. If you insist, I can put it to the top of the module (since it only needed in this function I've put it there).