DeepDiff pathlib.PosixPath objects

[jmccreight]

First off, deepdiff is great. Thanks.

I found the following behavior somewhat surprising. (The failure is on the last line, my expectations are set on the preceeding lines)

import pathlib
from deepdiff import DeepDiff

# Great... 
p1='foobar'
p2='barfoo'
assert p1 != p2
assert not p1.__eq__(p2)
assert bool(DeepDiff(p1, p2)) is True # True == not-empty

# Huh?...
pp1=pathlib.PosixPath(p1)
pp2=pathlib.PosixPath(p2)
assert pp1 != pp2
assert not pp1.__eq__(pp2)
assert bool(DeepDiff(pp1, pp2)) is True # True == not-empty

I just verified this with the master branch.

I dug into the code a bit and found the following:

1. deepdiff calls __diff_obj() on the pathlib.PosixPath objects
2. these objects have no __dict__
3. these objects have empty slots (e.g. bool(pp1.__slots__) is False).

So that's how the failure to detect the diff goes down. I have no idea about the expected behavior.

So, I'm perty new to python, but my naive expectations would be that before diffing the __dict__ or a dict based on the __slots__ of an object, you'd first check that object for __eq__ and use that to compare the objects. Or, if that didnt make sense, maybe there would be an optional list of types to __eq__ passed to DeepDiff instead of traversing __dict__? I honestly dont know where/how these ideas or expectations might fit in to how deepdiff is/should be structured... or if there is already something I'm missing that solves this.

Above I'm giving the MRE, but in my use case I have these pathlib.PosixPath objects inside other objects (which dont have__eq__). Thanks and I'm happy to be helpful however I can since I will rely on deepdiff!

(I also found dunders werent designed for communication in markdown! geeseess) Cheers,

[jmccreight]

see #104.