https://github.com/gzt/replacer
Drop-in replacement for the standalone R math library if all you need is runif().
https://github.com/gzt/replacer
c mersenne-twister r rng
Last synced: 4 months ago
JSON representation
Drop-in replacement for the standalone R math library if all you need is runif().
- Host: GitHub
- URL: https://github.com/gzt/replacer
- Owner: gzt
- License: apache-2.0
- Created: 2019-03-05T17:47:59.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-03-14T06:03:33.000Z (over 6 years ago)
- Last Synced: 2025-03-13T10:34:32.714Z (7 months ago)
- Topics: c, mersenne-twister, r, rng
- Language: C
- Size: 64.5 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# replaceR
Drop-in replacement for the standalone R math library if all you need is `runif()`.
This allows you to use `set_seed()` and `runif(a, b)` in `C` just as if you were using the `Rmath.h` standalone library.
While it uses the same random number generator as base `R`, I cannot guarantee that the seeds will produce the same results as `R`.
However, results are reproducible. There are further functions available in the Mersenne Twister code if you want.
Read their code or their `readme-mt.txt` for more details. Or check out their
[web-page](http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/emt.html). I have, essentially, only provided
a small wrapper around their code to make it mimic the behavior of the `R` library for those two functions.
If you want something more than that, don't use this! Either roll your own or, better, use the `R` standalone math
library.
I have also included a `randint(bound)` function which produces fast, unbiased uniform random integer
generation in the range `[0, bound)`. This is better than doing something like `n * runif(0,1)`.
I have included a couple other initialization functions as examples - one that uses a simple hash of the arguments
to the function and one that uses that plus the provided seed as well. Note the hash is a very simple one,
but this does give a way to quickly get reproducible results for testing without having to worry about explicitly
setting and storing a seed while also getting different results for different inputs.
The `mt19937ar.*` files are copyright their original owners subject to their
terms. See those files for details.
To compile and run an example, try:
gcc mt19937ar.c replaceR.c testreplaceR.c -o testreplaceR
./testreplaceR
The results here should agree with the results in `ReplaceResults.txt`.
Here is a histogram of 1000 random integers generated from the interval [0,100):
And results of a Kolmogorov-Smirnov test of the data in `R`:
```
Two-sample Kolmogorov-Smirnov test
data: res and rep(0:99, 10)
D = 0.011, p-value = 1
alternative hypothesis: two-sided
```
Your results should agree.