What information should the fingerprint be based on?

[konstin]

We want to generate the right fingerprint values from ruff for integration with gitlab code quality (PR). The question is, on what information should the fingerprint be based?

If we e.g. take the following python code ...

def a(x=[]):
    x.append(1)
    print(x)


def b(y=[]):
    y.append(2)
    print(y)

... and run ruff on it, we get two B006 violations:

$ ruff --select B --show-source scratch.py
scratch.py:1:9: B006 [*] Do not use mutable data structures for argument defaults
  |
1 | def a(x=[]):
  |         ^^ B006
2 |     x.append(1)
3 |     print(x)
  |
  = help: Replace with `None`; initialize within function

scratch.py:6:9: B006 [*] Do not use mutable data structures for argument defaults
  |
6 | def b(y=[]):
  |         ^^ B006
7 |     y.append(2)
8 |     print(y)
  |
  = help: Replace with `None`; initialize within function

How should these be hashed for the fingerprint? If we include only the message and the source of the violation ([]), we get to identical fingerprints. If we include the line number on the other hand, the fingerprint will change if any line is inserted or removed before them.