In isspace, isalpha, isalnum and isdigit, I see code like:

/* Shortcut for single character strings */
if (PyString_GET_SIZE(self) == 1 &&
    isspace(*p))
    return PyBool_FromLong(1);

Is it intentional to not use:

if (PyString_GET_SIZE(self) == 1))
    return PyBool_FromLong(isspace(*p) != 0);

which would be faster when the result is False (but a tad slower when it is True because of the extra comparison).

Also, is there a reason (other than historical) why the macros Py_RETURN_TRUE and Py_RETURN_FALSE are not used instead of their equivalent functions PyBool_FromLong(1) and PyBool_FromLong(0)?

See:
http://hg.python.org/cpython/file/e87364449954/Objects/stringobject.c#l3324

The shortcut seems fairly pointless to me.

If you would like to improve Python, you have to focus on the development version which is Python 3.4. In this version, the code is different:

    if (length == 1)
        return PyBool_FromLong(
            Py_UNICODE_ISSPACE(PyUnicode_READ(kind, data, 0)));

I'm not sure that having a special case for string of 1 character provide any speed up...

There's still stuff in bytes_methods.c which looks like the old string code.

Argl. I know I should have used 3.4... but that is what I thought I did. 
I used http://hg.python.org/cpython then "browse", and assumed it was the default branch... I know realize that since the last commit at that time was on the 2.7 branch, that is what I got. And I didn't even realize my mistake by the string vs unicode differences. Now, I will just go hide somewhere in shame...

Benjamin, if that comment means there's still something to be done, please reopen.