When python builtin hash() is just wired cross-platform. I have an app use builtin hash() for 'test'. Both systems are 64bit, python 2.7.12
windows:
>>> hash('test')
1308370872
linux:
>>> hash('test')
2314058222102390712
Why is this?
Asked
Active
Viewed 1,648 times
4
thefourtheye
- 233,700
- 52
- 457
- 497
wiwengweng
- 352
- 1
- 4
- 16
-
when use hashlib, everything is fine. Is there any doc refered to this difference? – wiwengweng Oct 12 '16 at 02:23
-
1Do you expect them to return the same values? Why? Running `hash("test")` twice in different sessions on the same system, returns different values each time. – Oct 12 '16 at 02:27
-
1Maybe one python is 32bit and another is 64bit? You can run python 32bit on a 64bit operating system – Ella Sharakanski Oct 12 '16 at 02:34
-
@donkopotamus nope – Antti Haapala -- Слава Україні Oct 12 '16 at 03:58
-
OP explicitly mentioned that they are using Python 2.7.12. So the dupe is not relevant, as it is about Python 3.3. Reopening the question. – thefourtheye Oct 13 '16 at 06:45
-
I can't image all of you disccus so much, as I completely know nothing about the build_in function. True that hashlib solve my problem: I need the same strings return the same num squence. Thanks all. – wiwengweng Oct 18 '16 at 03:12
1 Answers
4
There are no guarantees about the value hash returns in Python. It appears that you're using a 32-bit Windows Python (that's a guess), and that you're using a 64-bit python on linux (again, a guess). IIRC (and I haven't checked), The default hash(item) returns the address of item as its hash value.
If you want to have values you can compare across operating systems, look at hashlib.
cco
- 5,873
- 1
- 16
- 21
-
I just reread the question; `hash(item)` returning the address of `item` could return wildly different values betweens different runs on one platform, much less on different ones, regardless of 32/64-bit issues. – cco Oct 13 '16 at 07:19
-
Worse, two strings with equal value don't necessarily have the same memory address, but they need to have the same hash or they could appear twice in dictionary keys. So this can't be it. – RemcoGerlich Oct 13 '16 at 07:23
-
Yep. Python does us a lot of favors by interning strings transparently, so `hash('foo')` could be either the same or different from `xx='foo'; hash(xx)`. `hashlib` has solutions for this problem. – cco Oct 13 '16 at 07:33
-
hashlib is completely immune to the "problem" as it simply doesn't care about the address, the hash depends on the string's value. Same for hash(). – RemcoGerlich Oct 13 '16 at 07:46
-
1True, but hash() isn't guaranteed to be platform-independent, and hashlib is. – cco Oct 13 '16 at 07:50