{Async,}ExitStack's close() and aclose() should have an optional exception parameter

[dcolascione]

Feature or enhancement

Proposal:

ExitStack and AsyncExitStack have close and aclose methods, respectively, that finalize all entries in the stack before scope exit would otherwise do it. Right now, close and aclose accept no parameters and all stack entries are finalized as successes --- i.e. a call to __exit__ with no exception information.

It would be useful to add an optional exception parameter to close and aclose that would allow us to explicitly close a stack as if it had failed, e.g. with a synthetic timeout exception. Today, you can do the job by manually invoking __exit__ or __aexit__ on the stack, but .close() is meant to be more ergonomic.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

[frenzymadness]

Could you please explain what you would use this new feature for? Are there any existing examples where the code needs to use __exit__ directly because of the absence of the requested feature?

[dcolascione]

Just as you can use .close() when you want to "reset" a stack without changing its identity, e.g. when iterating over something in a loop, you might want to .close(with_error) a stack when doing the same thing when retrying a failed attempt to do something. The use cases are symmetric.

[picnixz]

Please discuss this proposal on https://discuss.python.org/c/ideas/6 with a concrete example (not just with words, but a code example). I don't see the needs to actually pre-emptively close a stack as if an error occurred as this would also invoke the callbacks on that specific error. Couldn't you just raise an exception after entering the stack? with stack: raise ... I think this would do the same right?

For now I'm going to close it as not planned because it's not clear whether it's needed enough to warrant its place in the stdlib. If the proposal gathers engouh support, I can reopen it.

[dcolascione]

Do you not see an argument from symmetry here? Would you add a .close() if it didn't exist today? This is a tiny bug fix.

[picnixz]

Sorry, but this would still be a feature request, not a bug fix. In addition, the documented method says:

Immediately unwinds the callback stack, invoking callbacks in the reverse order of registration. For any context managers and exit callbacks registered, the arguments passed in will indicate that no exception occurred.

So changing this is both a semantic and behavioral change. Since the use cases are niche (at least, unless proven otherwise), it's also not enough to warrant such addition in the stdlib. The "symmetry" argument doesn't convince me enough. I would rather have a separate method to indicate the intent of an error.