class Node:
    def __init__(self, edges = set()):
        self.edges = edges


def main():
    foo = Node()
    bar = Node()
    quz = Node()

    foo.edges.add(bar)
    bar.edges.add(foo)

    assert(foo is not bar) # assertion succeeds
    assert(foo is not quz) # assertion succeeds
    assert(bar is not quz) # assertion succeeds
    assert(len(quz.edges) == 0) # assertion fails??


main()
spoiler

Mutable default values are shared across objects. The set in this case.

  • Kevin@programming.dev
    link
    fedilink
    arrow-up
    10
    ·
    edit-2
    11 days ago

    Oh I had a similar bug but with defaulted dicts. Default args are constructed once and reused. Not a problem for immutable args, but mutables like dicts (and sets I’d also assume) are all shared.

    EDIT: whoops, didn’t see you spoilered the answer, my bad! If it helps, i found my bug when dealing with cross-thread stuff, so that was a fun moment to bisect

    • unalivejoy@lemm.ee
      link
      fedilink
      English
      arrow-up
      4
      ·
      11 days ago

      You may like collections.defaultdict. Pass the constructor a factory function to be run when a key is missing.

      dd = defaultdict(list)
      dd['key'].append("value")
      print(dd['key'])  # ["value"]
      
  • m_f@discuss.online
    link
    fedilink
    English
    arrow-up
    3
    ·
    edit-2
    11 days ago

    Yeah, I discovered this when a coworker wrote code like def foo(timestamp = now()) and had fun debugging why there were a bunch of duplicate timestamps.

    PEP 671 would add new syntax to ease the pain, but it’s not accepted yet. It would allow for writing function definitions like one of these:

    def bisect_right(a, x, lo=0, hi=>len(a), *, key=None):
    def connect(timeout=>default_timeout):
    def add_item(item, target=>[]):
    def format_time(fmt, time_t=>time.time()):
    
      • FizzyOrange@programming.dev
        link
        fedilink
        arrow-up
        3
        ·
        10 days ago

        Yeah I tried Ruff about a year ago and it only had really trivial lints so it wasn’t a good replacement for Pylint. Is it on par yet?

        • milkisklim@lemm.ee
          link
          fedilink
          arrow-up
          2
          ·
          edit-2
          10 days ago

          What do you mean by trivial? I am not necessarily the most experienced coder, but it does a great job yelling at me to keep methods short and simple.

          I’d suggest taking five minutes whenever and look up the ruff ruleset to see if it would be helpful for you.

          Also maybe because I don’t know how to use pylint in vs code, but the only semi useful thing it catches for me is if my venv doesn’t have a library the code imports.

          Edit: For example, Ruff has caught this bug (mutable argument defaults) in my code before.

          • FizzyOrange@programming.dev
            link
            fedilink
            arrow-up
            1
            ·
            10 days ago

            it does a great job yelling at me to keep methods short and simple

            Yes style things like that are what I would consider trivial. I also think those are actively bad lints. Yes methods should be short in general, but making it a hard enforced limit means you end up getting sidetracked by refactoring when you only wanted to add one line to a method.

  • waz@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    edit-2
    10 days ago

    Wow, I learned this bug during a job interview I had this week. I’m not much of a Python guy, but I was givin a Python coding challenge and I tried default initializing a parameter to an empty list. The guy interviewing me looked horrified and explained the problem and sent me an article about it to read later. It’s a odd coincidence coming across it twice in one week.