Repository navigation
Add back typo detection to sub-attribute access #137967
Description
Bug report
Bug description:
I remember we used to have this when all the typo-detection machinery was in the C code but somehow that has not survived the transition to the new traceback machinery (or perhaps I am misremembering but I recall adding this originally when we added the first round). Either way this PR ensures that we have a nice error message when the typo is across attribute access:
from dataclasses import dataclass
from math import pi
from types import SimpleNamespace
from typing import Any, Dict
@dataclass
class Circle:
radius: float
@property
def area(self) -> float:
return pi * self.radius**2
@property
def perimeter(self) -> float:
return 2 * pi * self.radius
@dataclass
class Square:
side: float
@property
def area(self) -> float:
return self.side**2
@property
def perimeter(self) -> float:
return 4 * self.side
class Container:
def __init__(self, inner: Any) -> None:
self.inner = inner
if __name__ == "__main__":
square = Square(side=4)
container = Container(square)
print(container.area)Should print:
Traceback (most recent call last):
File "/home/pablogsal/github/python/main/lel.py", line 42, in <module>
print(container.area)
^^^^^^^^^^^^^^
AttributeError: 'Container' object has no attribute 'area'. Did you mean: 'inner.area'?
CPython versions tested on:
CPython main branch
Operating systems tested on:
No response
Activity
pythongh-137967: Restore suggestions on nested attribute access
pythongh-137967: Restore suggestions on nested attribute access
SylvainDe commentedon Aug 20, 2025
Looks great!
Out of curiosity, is there any way for not having the following suggestion :
AttributeError: 'Container' object has no attribute 'area'. Did you mean: 'container.inner.area'?? (Maybe it'd harder to implement, maybe it's just not wanted in the first place)
pablogsal commentedon Aug 20, 2025
Not sure I follow, are you saying you want
container.inner.areainstead ofinner.area?SylvainDe commentedon Aug 20, 2025
pablogsal commentedon Aug 21, 2025
Yeah unfortunately is not possible or at least not very easy: at the time we produce the message we don't have access to the variable name the object is assigned to. We could find it in the frame but that's slow and is a but complicated so it will have some performance impact. Furthermore the rest of the similar error messages don't have the name and that will make it a bit inconsistent, so I think I am going to go with this as it is.
Thanks for the suggestion though!
SylvainDe commentedon Aug 21, 2025
Sure! That totally makes sense! Thanks for everything you've done for error messages and for replying to my suggestion!
gh-137967: Restore suggestions on nested attribute access (#137968)
pythongh-137967: Fix "Whats New" doc example
pythongh-137967: Fix "Whats New" doc example
gh-137967: Fix "Whats New" doc example (#138028)