How __hash__ is implemented in Python 3.2?

Question

I want to make custom object hash-able (via pickling). I could find __hash__ algorithm for Python 2.x (see code below), but it obviously differs

Answer 1

I looked up the new function in the source (in unicodeobject.c) and rebuilt it in Python. Here it is:

def my_hash(string):
    x = ord(string[0]) << 7
    for c in string:
        x = (1000003 * x) ^ ord(c)
    x ^= len(string)
    needCorrection =  x & (1 << 65)
    x %= 2 ** 64
    if needCorrection:
        x = -~(-x ^ 0xFFFFFFFFFFFFFFFF)
    if x == -1:
        x = -2
    return x

This is 64-bit only, though. Now with correction for Python's weird behavior when numbers become negative. (You better don't think about this too much.)

Answer 2

The answer why they are different is written there:

Hash values are now values of a new type, Py_hash_t, which is defined to be the same size as a pointer. Previously they were of type long, which on some 64-bit operating systems is still only 32 bits long.

The hashing also consider new values to be calculate, take a look at

 sys.hash_info 

For strings, you can take a look at http://svn.python.org/view/python/trunk/Objects/stringobject.c?view=markup line 1263 string_hash(PyStringObject *a)