Question

I’m having a problem correctly hashing my objects. Consider the following code:

class Foo:
    def __init__(self, bar):
        self.keys = list(bar.keys())
        self.values = list(bar.values())    
    def __str__(self):
        return ', '.join( '%s: %s' % z for z in zip(self.keys, self.values))    
    def __hash__(self):
        return hash(str(self))

if __name__ == '__main__':
    result = set()
    d = { 1: 2, 3: 4, 5: 6, 7: 8 }
    for i in range(10):
        result.add(Foo(d))
    for r in result:
        print r, hash(r)

I expect the result set to contain a single element, since all the added Foo objects have the same contents, and therefore the same hash.

However, this is the result:

misha@misha-K42Jr:~/Desktop/stackoverflow$ python hashproblem.py 
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338
1: 2, 3: 4, 5: 6, 7: 8 2131119371379196338

What is the problem here? The hashes do look the same, so shouldn’t they be treated as duplicates by the built-in set object? Why does the set contain duplicates?

I’ve noticed that if I use str(Foo(d)) instead of Foo(d) when adding elements to the set, things work as expected. Why does it matter?

Python version is:

misha@misha-K42Jr:~/Desktop/stackoverflow$ python --version
Python 2.6.6

Answer 1

Since the __hash__ method is only use for the internal hash-table, you need to redefine __eq__ as well.

Overriding only __eq__ is not correct either. If two object are equal, ie, a.__eq__(b) == True, then both hash(a) and hash(b) must be equal as well.

The default __hash__ method is:

def __hash__(self):
    return id(self)