UndeclaredException is the only place where I encountered empty hints. An alternative solution is to add the check there to prevent the hint from being an empty string in the first place.

It seems like a code smell that we are creating empty string hints in the first place. Do we know where they are coming from?

It seems like a code smell that we are creating empty string hints in the first place. Do we know where they are coming from?

I think its as in this - #4351 (comment) - just the UndeclaredException. But I can spend some more time if i encounter it elsewhere.

note the regression comes from #4154 -- the previous code checked for truthy value of self.hint.

note the regression comes from #4154 -- the previous code checked for truthy value of self.hint.

Yeah, I think there are two main solutions:

1. Have the check in the same way as pre fix[ux]: fix empty hints in error messages #4351 - (which is proposed in this PR, because i though hint can be only None or string, but I can rewrite it to if not hint to keep the truthyness).
2. Or keep the current test if hint is None and modify get_levenshtein_error_suggestions to return None instead of an empty string (which is used as the hint in one type of UndeclaredDefinition)

I think the second solution is cleaner, but the first solution ensures that we catch any missed empty string hints.
What do you prefer?

i think 2. is better -- we should not be supplying empty string as a hint!

i think 2. is better -- we should not be supplying empty string as a hint!

Hope its better now :)