In programming, we often need fairly distribution of data. For example, this can be used when assigning work to one of several running processes and partitioning.
Recently, while realizing such a problem, I decided to approach the subject more “mathematically”.
Using erlang:phash2/2 you can get the same hash for the same Erlang term regardless of machine architecture and ERTS version.
It is possible to indicate the maximum range in the form 0..Range-1.
The function should always be used for hashing terms; thanks to its speed, it is also great for binaries.
We can hash any data and limit it to a range of integers.
If you expect to use atoms as input, it makes sense to use a preprocessing step with :erlang.term_to_binary/1 so that the results obtained are more spread out in the given range.
Example from GitHub discussions:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
iex(1)> :erlang.phash2(:foo)
27999
iex(2)> :erlang.phash2(:erlang.term_to_binary(:foo))
86811783
iex(3)> :erlang.phash2(:erlang.term_to_binary(:a))
64216263
iex(4)> :erlang.phash2(:erlang.term_to_binary(:aaaaaa))
86861957
iex(5)> :erlang.phash2(:erlang.term_to_binary(:b))
14428281
iex(6)> :erlang.phash2(:erlang.term_to_binary(:b), trunc(:math.pow(2,32)-1))
3235653753
iex(7)> :erlang.phash2(:foo, trunc(:math.pow(2,32)-1))
27999
iex(8)> :erlang.phash2(:erlang.term_to_binary(:foo), trunc(:math.pow(2,32)-1))
355247239