1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
|
SyncRNG
=======
A synchronized Tausworthe RNG usable in R and Python.
Why?
----
This program was created because it was desired to have the same random
numbers in both R and Python programs. Although both languages implement a
Mersenne-Twister RNG, the implementations are so different that it is not
possible to get the same random numbers with the same seed.
SyncRNG is a Tausworthe RNG implemented in `syncrng.c`, and linked to both R
and Python. Since both use the same underlying C code, the random numbers will
be the same in both languages, provided the same seed is used.
How
---
In Python, the interface SyncRNG.py can be used as an importable module. In R,
it is adviced to simply source the 'SyncRNG.R' file. Before use, make sure to
build both shared libraries using:
make
Then, in a Python script located in the same directory as `syncrng.so` and
`SyncRNG.py`, you can do:
from pysyncrng import SyncRNG
s = SyncRNG(seed=123456)
for i in range(10):
print(s.randi())
Similarly, in an R script located in the same directory as `RSyncRNG.so` and
`SyncRNG.R`, you can do:
source('./SyncRNG.R')
s = SyncRNG(seed=123456)
for (i in 1:10) {
cat(s$randi(), '\n')
}
You'll notice that the random numbers are indeed the same.
Notes
-----
Since R is not capable of reliable handling integers larger than `2^32 - 1`,
the random numbers are internally capped at this value (using a bitwise and
with `0x7FFFFFF`), this influences the quality of the random numbers. The
random numbers are no longer uniformly distributed on `[0, 2^32 -1]`. For the
intended use of SyncRNG this is not a problem, but it is a compromise worth
considering when using SyncRNG. SyncRNG should definitely not be used for any
cryptographic purposes.
TODO
----
Future versions may include a random number generator that does not need
capping, and is uniform. It may also provide easier system-wide installation
through an R package and a Python module.
|
