I don't think this is going to be easily fixable. Here's the ast for the expression A().f(''):

Expr(value=Call(
    func=Attribute(value=Call(func=Name(id='A', ctx=Load(), lineno=1, col_offset=0), args=[], keywords=[], lineno=1, col_offset=0),
                   attr='f', ctx=Load(), lineno=1, col_offset=0)
    , args=[Str(s='', lineno=1, col_offset=6)], keywords=[], lineno=1, col_offset=0),
    lineno=1, col_offset=0)]

Notice that all we know about f is that it's the string portion of an attribute lookup. It doesn't have any annotations (in the Python AST sense, i.e. line or col numbers) itself, aside from what's associated with the attribute lookup as a whole. Unfortunately, the column number associated with the attribute lookup is the first column of the value (in this case, A()), which starts in the 0th column.

So I'm guessing this would require an update to typed_ast. Oh well. Low priority is fine.

And what the fix would look like is not obvious (at least not to me), unfortunately.

(Raising priority in view of planned editor integration.)

I'm closing this issue: it doesn't appear to be a problem when mypy is run on 3.8+, as mypy uses the stdlib ast module on these newer versions of Python.

It seems unlikely to me that typed_ast will be updated to fix this issue, given that typed_ast is only really useful on Python <=3.7.