Hum , I think we might be able to get something better with

def structured_traceback(...):
    from difflib import get_close_matches
    get_close_matches(value.name, dir(value.obj)

And re-implemented the match logic (even if slightly different). That way we can properly highlight tokens.

I believe that the suggestion to use difflib would only work for AttributeError cases but not for NameError (where the misspelled name can be found in the local or global scope).

Furthermore, difflib gives different results from those given by cPython.

>>> import math
>>> math.Pi
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: module 'math' has no attribute 'Pi'. Did you mean: 'pi'?
>>> from difflib import get_close_matches
>>> get_close_matches('Pi', dir(math))
[]

CPython has done several tweaks to get better results. IMO, it's better to use the suggestions from the compiler itself to avoid any discrepancy and to avoid reimplementing/maintaining the algorithm.

python/cpython#25412
python/cpython#25443
python/cpython#25460
python/cpython#25584
python/cpython#25776

Closing old stale PRs, sorry if this was not addressed, but this has been a few years, so there are few chances of someone actually pushing this through.

Feel free to resubmit if you feel this has a chance to be revived.