False negative for `no-else-return` after `except`

[nickdrozd]

Bug description

This snippet returns in the except block, and therefore the else is unnecessary:

def f() -> bool:
    try:
        print('asdf')
    except:  # pylint: disable = bare-except
        return False
    else:
        return True

It should raise no-else-return.

Configuration

No response

Command used

`pylint asdf.py`

Pylint output

No warnings raised

Expected behavior

Should raise no-else-return

Pylint version

pylint 2.16.0-dev
astroid 2.13.0-dev0
Python 3.8.10 (default, Mar 15 2022, 12:22:08) 

[nickdrozd]

A similar case with raise:

    def _load_reporter_by_name(self, reporter_name: str) -> reporters.BaseReporter:
        name = reporter_name.lower()
        if name in self._reporters:
            return self._reporters[name]()

        try:
            reporter_class = _load_reporter_by_class(reporter_name)
        except (ImportError, AttributeError, AssertionError) as e:
            raise exceptions.InvalidReporterError(name) from e
        else:
            return reporter_class()

The except raises an exception, so the else is not needed.

[marcinnformula]

What?!

I don't understand why no-else-return affects try-except-else.

I also don't understand why else is "unnecessary" here:

def test_exception(raise_exception=True):
    try:
        if raise_exception:
            raise ValueError
    except ValueError:
        return "exception raised"
    else:
        return "exception NOT raised"

Yes, this is an equivalent of:

def test_exception(raise_exception=True):
    try:
        if raise_exception:
            raise ValueError
    except ValueError:
        return "exception raised"
    return "exception NOT raised"

But please note that explicit else block is for:

1. readability (PEP20)
2. avoiding negative cons due to possible refactoring in except block when removing return (that would change the flow)

#7888 should be reverted

[Pierre-Sassoulas]

Could you open a new issue @marcinnformula, a comment on a closed issue is going to be forgotten. I don't think we should revert but we could introduce an option or a way to distinguish the two use cases.