Oh.. this is a Python spec omg.

Python 3.11.2 (main, Feb 16 2023, 02:55:59) [Clang 14.0.0 (clang-1400.0.29.202)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import ast
>>> ast.parse("""
... @decorator
... def x(): pass
... """)
<ast.Module object at 0x100738be0>
>>> a = _
>>> f = a.body[0]
>>> f.lineno
3
>>> f.decorator_list[0].lineno
2

We will really need a feature flag for ruff.

I'd better to understand it better if this is also related to symbol table or SyntaxError.
If not, it can be handled in ast module with this patch

@MichaReiser Do you want to make it has different value only for more-attributes feature right now? That has no blocker.

@MichaReiser Do you want to make it has different value only for more-attributes feature right now? That has no blocker.

In my view, this stretches feature flags to the limit, or even over the limit. We would need to rename it at least. I also plan to make other range changes:

def a():
	pass

	# comment

The range of the function should include the #comment, which it does not today.

I think maintaining this change in a fork is straightforward enough and probably easier than dealing with even more conditional code that has very limited test coverage.

The range of the function should include the #comment, which it does not today.

it seems going really different.

I think maintaining this change in a fork is straightforward enough and probably easier than dealing with even more conditional code that has very limited test coverage.

Yes, that sounds like truth.

Reusing ast crate but making a fork of parser crate makes sense to me.

We can make a new crate like rustpython-parser-util to share fundamental parser utilities except for the details.

We can make a new crate like rustpython-parser-util to share fundamental parser utilities except for the details.

I don't think that this is necessary. I still plan to pull in upstream changes. I don't expect to many conflicts since the grammar shouldn't change frequently