Making a python user-defined class sortable, hashable

Question

What methods need to be overridden/implemented when making user-defined classes sortable and/or hashable in python?

What are the gotchas to watch out for?

I type dir({}) into my interpreter to get a list of methods on built-in dicts. Of those, I assume I need to some implement some subset of

['__cmp__', '__eq__', '__ge__', '__gt__', '__hash__', '__le__', '__lt__', '__ne__'] 

Is there a difference in which methods must be implemented for Python3 as opposed to Python2?

Answer 1

I almost posted this as a comment to the other answers but it's really an answer in and of itself.

To make your items sortable, they only need to implement __lt__. That's the only method used by the built in sort.

The other comparisons or functools.total_ordering are only needed if you actually want to use the comparison operators with your class.

To make your items hashable, you implement __hash__ as others noted. You should also implement __eq__ in a compatible way -- items that are equivalent should hash the same.

Answer 2

There isn't any difference between Python 2 and 3.

For sortability:

You should define comparision methods. This makes your items sortable. Generally, you shouldn't prefer __cmp__().

I usually use functools.total_ordering decorator.

functools.total_ordering(cls) Given a class defining one or more rich comparison ordering methods, this class decorator supplies the rest. This simplifies the effort involved in specifying all of the possible rich comparison operations:

The class must define one of __lt__(), __le__(), __gt__(), or __ge__(). In addition, the class should supply an __eq__() method.

You should be careful that your comparison methods do not have any side effects. (change any of the values of the object)

For hashing:

You should implement __hash__() method. I think the best way is returning hash(repr(self)), so your hash would be unique.