This is related to bpo-44649. This is the simplest non-dataclasses case I could come up with.

Given:

class Example:
    __slots__ = ('a', 'b')
    def __init__(self, a):
        self.a = a

obj = Example(42)
print(obj.b)

I get this in 3.10 and main:

Traceback (most recent call last):
  File "C:\home\eric\local\python\cpython\testdc.py", line 7, in <module>
    print(obj.b)
          ^^^^^
AttributeError: b. Did you mean: 'b'?

The error message is confusing.

3.8 gives:

Traceback (most recent call last):
  File "testdc.py", line 7, in <module>
    print(obj.b)
AttributeError: b

I don't have 3.9 around to test with.

Maybe don't print the "Did you mean" part if the suggestion is the same as the requested attribute?

The fact that the instance variable isn't initialized is the actual error in bpo-44649, and I'll fix it there.

Also, the dot after the first 'b' was confusing to me: I thought it had something to do with an attribute of b. And the quotes around the second 'b' were also confusing, but that's mostly because the original example initialized a class variable named 'b' with the str value 'b'.

Maybe a better error would use a colon:

AttributeError: foo: Did you mean: 'foo1'?

Looking at 3.10s behavior, I can't decide what logic it's using to suggest something:

>>> 'foo'.formatx
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'str' object has no attribute 'formatx'. Did you mean: 'format'?
>>> class C: pass
...
>>> C().a
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'C' object has no attribute 'a'

For attribute errors we just call dir() on the object and we do a suggestion based on the names so if a name appears in the dir() then we will consider it for the suggestion.

Maybe don't print the "Did you mean" part if the suggestion is the same as the requested attribute?

I think this is a good idea and will take care of many other similar cases. Thanks for the suggestion, Eric! I will prepare a PR

It's obviously not super important, but it would be nicer if the version with the suggestion were more like the version without the suggestion. Specifically, mentioning the object type:

>>> class E:
...     __slots__=('a')
...
>>> E().a
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: a. Did you mean: 'a'?
>>> E().b
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'E' object has no attribute 'b'
>>>

Thanks, Pablo!

it would be nicer if the version with the suggestion were more like the version without the suggestion

Agreed, but that is actually not related with the suggestions. We only append a string at the end of whatever exception it was there. This difference is based on the attribute error for a slot not initialized vs an attribute error for something that is not there. For example, in 3.9:

---

class E:
    __slots__=('blech')

E().blech

Traceback (most recent call last):
  File "/home/pablogsal/github/cpython/lel.py", line 6, in <module>
    E().blech
AttributeError: blech

while

class E:
    __slots__=('blech')

E().bluch

Traceback (most recent call last):
  File "/home/pablogsal/github/cpython/lel.py", line 4, in <module>
    E().bluch
AttributeError: 'E' object has no attribute 'bluch'

New changeset 6714dec by Pablo Galindo Salgado in branch 'main':
bpo-44655: Don't include suggestions for attributes that are the same as the missing one (GH-27197)
6714dec

Got it. Thanks for jumping on this quickly.

New changeset a0b1d40 by Miss Islington (bot) in branch '3.10':
bpo-44655: Don't include suggestions for attributes that are the same as the missing one (GH-27197) (GH-27198)
a0b1d40

New changeset f783428 by Pablo Galindo Salgado in branch 'main':
bpo-44655: Include the name of the type in unset __slots__ attribute errors (GH-27199)
f783428

New changeset efda905 by Miss Islington (bot) in branch '3.10':
bpo-44655: Include the name of the type in unset __slots__ attribute errors (GH-27199) (GH-27201)
efda905