Hashing repr of object instead of tuple of attributes

Asked May 12 '17 at 19:52

Active May 12 '17 at 19:52

Viewed 258 times

A common way of overriding the __hash__() method for custom objects is to make a tuple t with the object attributes and apply return hash(t), As described in this this question.

I have a program doing this, and after running cProfile, i've observed that this approach is the bottleneck of my program. I've made the following test and got the result:

>>> t1 = tm.Timer(setup="x = str((10,12))", stmt="hash(x)")
>>> t2 = tm.Timer(setup="x = (10,12)", stmt="hash(x)")
>>> min(t1.repeat(100)), min(t2.repeat(100))
(0.04663395881652832, 0.05523419380187988)

From which i conclude that using an string on the hash function is faster than using a tuple. Given that i've made the __repr__() implementation unique for every object of my class, i think that shouldn't be any problems.Before changing my program, is there any possible problem with this approach(or my test metodology)?

edited May 23 '17 at 12:10

Community

asked May 12 '17 at 19:52

Victor Hugo

Shouldn't you include the string creation and the tuple creation in your benchmark? – Liran Funaro May 18 '17 at 22:28
In my case not. The tuple/string are created only once, and the hash function is executed with them many times. – Victor Hugo May 19 '17 at 17:41
If the string/tuple isn't changing, so the hash will have the same result. Why do you need to call it again and again? – Liran Funaro May 19 '17 at 17:44

Hashing repr of object instead of tuple of attributes

0 Answers0