deepdiff with include-paths is changing results, if a not considered path changes

[jlaba]

Comparing some paths of two dicts with deepdiff the results differ, depending on the presence or absence of other "not included" paths (see also Stackoverflow question).

from deepdiff import DeepDiff
DeepDiff(
    {
        'name': 'Testname',
        'code': 'bla',
        'noneCode': 'blu',
    }, {
        'uid': '12345',
        'name': 'Testname',
    },
    include_paths = "root['name']"
) # expected to be {}, but detects a change

DeepDiff(
    {
        'name': 'Testname',
        'code': 'bla',
    }, {
        'uid': '12345',
        'name': 'Testname',
    },
    include_paths = "root['name']"
) # expected to be {} and is {}

OS, DeepDiff version and Python version (please complete the following information):

• OS: OpenSuSe Leap 15.6 as well as Windows 11
• Python Version 3.10 (on linux), 3.11 (on windows)
• DeepDiff Version 8.0.1

[seperman]

@jlaba Thanks for reporting the bug.
We should switch our "include_path" logic so it first does the full diff, then it removes any reports that are outside the include_paths.

[seperman]

Also PRs are very welcome in case you have time to fix it!
Thanks

[jlaba]

So I'm not an expert in performance, but most likely in larger dicts, it might be good to only walk through the paths, which should be compared to speed up the process. I had a quick check at the code and most likely this line

deepdiff/deepdiff/diff.py

Line 481 in 6d8a4c7

if self.include_paths and level_path != 'root':

is the issue. My thesis is, that the problem occurs, because I have not included all root elements. I will try to build a test for it, to check whether the thesis is true :-)

[jlaba]

So my thesis proofed to be wrong. The problem is, if the number of items in a dict is changing, even if those added or removed items are not in the included path.