RNG 0.1.7

Random Number Generator, Functional C++ Random Library for Python3

RNG: Random Number Generator

Provides API to the C++ Random library as a c-extension for Python3

Default Random Engine: Mersenne Twister 64, with hardware entropy. Additional engines and seeding strategies are planned to be available in the unbounded future. More info about MT64: https://en.wikipedia.org/wiki/Mersenne_Twister

The RNG module is not suitable for cryptography, and perfect for other non-deterministic needs like A.I. or games of chance.

Recommended Installation: $ pip install RNG

RNG is not intended to be a drop-in replacement for the Python random module, RNG is a whole different beast.

Random Binary Functions

• random_bool(truth_factor: float) -> bool
  • Bernoulli distribution.
  • @param truth_factor :: the probability of True. Expected input range: [0.0, 1.0]
  • @return :: True or False

Random Integer Functions

• random_int(lo: int, hi: int) -> int
  • Flat uniform distribution.
  • @param lo :: the lower bound. Param lo must not be greater than param hi.
  • @param hi :: the upper bound.
  • @return :: random integer in the inclusive range [lo..hi]
• random_binomial(number_of_trials: int, probability: float) -> int
  • Based on the idea of flipping a coin and counting how many heads come up after some number of flips.
  • @param number_of_trials :: how many times to flip a coin.
  • @param probability :: how likely heads will be flipped. 0.50 is a fair coin.
  • @return :: count of how many heads came up.
• random_negative_binomial(trial_successes: int, probability: float) -> int
  • Based on the idea of flipping a coin as long as it takes to succeed.
  • @param trial_successes :: the required number of heads flipped to succeed.
  • @param probability :: how likely heads will be flipped. 0.50 is a fair coin.
  • @return :: the count of how many tails came up before the required number of heads.
• random_geometric(probability: float) -> int
  • Same as random_negative_binomial(1, probability).
• random_poisson(mean: float) -> int
  • @param mean :: sets the average output of the function.
  • @return :: random integer, poisson distribution centered on the mean.
• random_discrete(count: int, xmin: int, xmax: int, step: int) -> int
  • @param count :: number of weighted values
  • @param xmin :: smallest weight of the set
  • @param xmin :: largest weight of the set
  • @param step :: value stepping

Random Floating Point Functions

• random_float(lo: float, hi: float) -> float
  • @param lo :: lower bound Float
  • @param hi :: upper bound Float
  • @return :: random Float in range {lo, hi} biclusive.
    • biclusive: feature/bug rendering the exclusivity of this function a bit more mysterious than desired. This is a known compiler bug.
    • The spec defines the output range to be [lo, hi).
• random_normal(mean: float, std_dev: float) -> float
• random_log_normal(log_mean: float, log_deviation: float) -> float
• random_gamma(shape: float, scale: float) -> float
• random_exponential(lambda_rate: float) -> float
• random_weibull(shape: float, scale: float) -> float
• random_extreme_value(shape: float, scale: float) -> float
• random_chi_squared(degrees_of_freedom: float) -> float
• random_cauchy(location: float, scale: float) -> float
• random_fisher_f(degrees_of_freedom_1: float, degrees_of_freedom_2: float) -> float
• random_student_t(degrees_of_freedom: float) -> float
• piecewise_constant_distribution coming soon, produces real values distributed on constant subintervals.
• piecewise_linear_distribution coming soon, produces real values distributed on defined subintervals.

Utilities

• generate_canonical() -> float
  • Evenly distributes real values of given precision across [0, 1). Suffers from the same biclusive feature/bug noted for random_float.
  • Currently set to max precision for long double. Precision could be parameterized in the future.
• seed_seq maybe coming soon, general-purpose bias-eliminating scrambled seed sequence generator.
  • Currently RNG uses hardware seeding exclusively. Software seeding may be a feature in a future release.

Engines

• mersenne_twister_engine 64 bit.
  • Implements Mersenne twister algorithm. Default engine on most modern systems.
• linear_congruential_engine coming soon, implements linear congruential algorithm.
• subtract_with_carry_engine coming soon, implements a subtract-with-carry (lagged Fibonacci) algorithm.

Engine Adaptors

Engine adaptors generate pseudo-random numbers using another random number engine as entropy source. They are generally used to alter the spectral characteristics of the underlying engine.

• discard_block_engine coming soon, discards some output of a random number engine.
• independent_bits_engine coming soon, packs the output of a random number engine into blocks of a specified number of bits.
• shuffle_order_engine coming soon, delivers the output of a random number engine in a different order.

Seeding or Entropy Source

• Non-deterministic hardware entropy source: std::random_device(). RNG is non-deterministic by default.
• Repeatable deterministic software seeding: maybe coming soon.

Development Log

RNG 0.1.7 beta

• The random_floating_point function renamed to random_float.
• The function c_rand() has been removed as well as all the cruft it required.
• Major Documentation Upgrade.
• Fixed an issue where keyword arguments would fail to propagate. Both, positional args and kwargs now work as intended.
• Added this Dev Log.

RNG 0.0.6 alpha

• Minor ABI changes.

RNG 0.0.5 alpha

• Tests redesigned slightly for Float functions.

RNG 0.0.4 alpha

• Random Float Functions Implemented.

RNG 0.0.3 alpha

• Random Integer Functions Implemented.

RNG 0.0.2 alpha

• Random Bool Function Implemented.

RNG 0.0.1 pre-alpha

• Planning & Design.

Distribution and Performance Test Suite

RNG 0.1.7 BETA

Binary RNG Tests

random_bool(truth_factor=1/3)
Time: min: 62ns, mode: 62ns, mean: 68ns, max: 125ns
Distribution:
False: 66.6449%
True: 33.3551%


Integer RNG Tests

random_int(lo=1, hi=6)
Time: min: 62ns, mode: 93ns, mean: 97ns, max: 281ns
Distribution:
1: 16.7303%
2: 16.652%
3: 16.6614%
4: 16.6513%
5: 16.6245%
6: 16.6805%

random_binomial(number_of_trials=4, probability=0.5)
Time: min: 156ns, mode: 187ns, mean: 194ns, max: 718ns
Distribution:
0: 6.251%
1: 25.011%
2: 37.5306%
3: 24.9723%
4: 6.2351%

random_negative_binomial(number_of_trials=5, probability=0.75)
Time: min: 93ns, mode: 125ns, mean: 123ns, max: 187ns
Distribution:
0: 23.6627%
1: 29.7502%
2: 22.2298%
3: 12.997%
4: 6.4596%
5: 2.9343%
6: 1.1996%
7: 0.488%
8: 0.1793%
9: 0.0657%
10: 0.0231%
11: 0.0073%
12: 0.0023%
13: 0.0008%
14: 0.0003%

random_geometric(probability=0.75)
Time: min: 62ns, mode: 62ns, mean: 66ns, max: 93ns
Distribution:
0: 75.0096%
1: 18.755%
2: 4.6704%
3: 1.1768%
4: 0.2893%
5: 0.0731%
6: 0.0197%
7: 0.0044%
8: 0.001%
9: 0.0006%
10: 0.0001%

random_poisson(mean=4.5)
Time: min: 93ns, mode: 125ns, mean: 123ns, max: 500ns
Distribution:
0: 1.1293%
1: 5.0529%
2: 11.2235%
3: 16.9171%
4: 18.9297%
5: 17.0588%
6: 12.7932%
7: 8.2155%
8: 4.6406%
9: 2.3242%
10: 1.051%
11: 0.4248%
12: 0.1615%
13: 0.0541%
14: 0.017%
15: 0.005%
16: 0.0015%
17: 0.0002%
18: 0.0001%

random_discrete(count=6, xmin=0.7, xmax=21.0, step=1)
Time: min: 437ns, mode: 468ns, mean: 469ns, max: 781ns
Distribution:
0: 4.8145%
1: 9.5061%
2: 14.2351%
3: 19.0452%
4: 23.8392%
5: 28.5599%


Floating Point RNG Tests

random_float(lo=0.0, hi=10.0)
Time: min: 62ns, mode: 62ns, mean: 67ns, max: 187ns
Floored Distribution:
0: 10.0477%
1: 10.0009%
2: 10.0118%
3: 9.9884%
4: 9.9892%
5: 10.0382%
6: 10.0036%
7: 9.9524%
8: 9.968%
9: 9.9998%

random_exponential(lambda_rate=1.0)
Time: min: 93ns, mode: 93ns, mean: 105ns, max: 156ns
Floored Distribution:
0: 63.327%
1: 23.2008%
2: 8.4878%
3: 3.1426%
4: 1.1715%
5: 0.428%
6: 0.1545%
7: 0.0548%
8: 0.0214%
9: 0.0057%
10: 0.0038%
11: 0.0014%
12: 0.0005%
13: 0.0001%
14: 0.0001%

random_gamma(shape=1.0, scale=1.0)
Time: min: 93ns, mode: 93ns, mean: 98ns, max: 156ns
Floored Distribution:
0: 63.1574%
1: 23.3084%
2: 8.5438%
3: 3.1526%
4: 1.1626%
5: 0.4212%
6: 0.1634%
7: 0.0577%
8: 0.0193%
9: 0.0087%
10: 0.0031%
11: 0.0011%
12: 0.0006%
13: 0.0001%

random_weibull(location=1.0, scale=1.0)
Time: min: 125ns, mode: 156ns, mean: 143ns, max: 187ns
Floored Distribution:
0: 63.2328%
1: 23.2939%
2: 8.5137%
3: 3.1451%
4: 1.1392%
5: 0.4273%
6: 0.1569%
7: 0.0556%
8: 0.0231%
9: 0.008%
10: 0.0031%
11: 0.0009%
12: 0.0003%
13: 0.0001%

random_extreme_value(location=0.0, scale=1.0)
Time: min: 93ns, mode: 125ns, mean: 124ns, max: 312ns
Rounded Distribution:
-3: 0.0001%
-2: 1.1381%
-1: 18.1035%
0: 35.3055%
1: 25.4613%
2: 12.0974%
3: 4.937%
4: 1.8778%
5: 0.6859%
6: 0.2486%
7: 0.0893%
8: 0.035%
9: 0.0138%
10: 0.0042%
11: 0.0021%
12: 0.0003%
13: 0.0001%

random_normal(mean=5.0, std_dev=2.0)
Time: min: 93ns, mode: 125ns, mean: 127ns, max: 187ns
Rounded Distribution:
-5: 0.0004%
-4: 0.0009%
-3: 0.0085%
-2: 0.0509%
-1: 0.2398%
0: 0.9404%
1: 2.7851%
2: 6.5624%
3: 12.1515%
4: 17.4516%
5: 19.6875%
6: 17.4993%
7: 12.0453%
8: 6.5669%
9: 2.771%
10: 0.9409%
11: 0.2379%
12: 0.0505%
13: 0.008%
14: 0.0011%
15: 0.0001%

random_log_normal(log_mean=1.6, log_deviation=0.25)
Time: min: 156ns, mode: 187ns, mean: 197ns, max: 687ns
Rounded Distribution:
1: 0.0001%
2: 0.3093%
3: 7.8934%
4: 26.8253%
5: 31.1587%
6: 19.9274%
7: 9.0125%
8: 3.3209%
9: 1.0976%
10: 0.323%
11: 0.0955%
12: 0.0276%
13: 0.0067%
14: 0.0014%
15: 0.0005%
16: 0.0001%

random_chi_squared(degrees_of_freedom=1.0)
Time: min: 156ns, mode: 187ns, mean: 182ns, max: 250ns
Floored Distribution:
0: 68.2285%
1: 16.0431%
2: 7.3764%
3: 3.8011%
4: 2.0362%
5: 1.0959%
6: 0.5994%
7: 0.3479%
8: 0.1973%
9: 0.1185%
10: 0.0685%
11: 0.0362%
12: 0.0218%
13: 0.0117%
14: 0.008%
15: 0.0044%
16: 0.002%
17: 0.0013%
18: 0.0007%
19: 0.001%
20: 0.0001%

random_cauchy(location=0.0, scale=0.0005)
Time: min: 93ns, mode: 125ns, mean: 116ns, max: 156ns
Rounded Distribution:
-181: 0.0001%
-177: 0.0001%
-83: 0.0001%
-43: 0.0001%
-23: 0.0001%
-13: 0.0002%
-11: 0.0001%
-10: 0.0001%
-9: 0.0002%
-7: 0.0005%
-6: 0.0001%
-5: 0.0005%
-4: 0.0015%
-3: 0.0015%
-2: 0.0049%
-1: 0.0232%
0: 99.9354%
1: 0.0218%
2: 0.0037%
3: 0.0015%
4: 0.0008%
5: 0.0007%
6: 0.0005%
7: 0.0008%
9: 0.0001%
10: 0.0005%
12: 0.0001%
13: 0.0001%
15: 0.0001%
20: 0.0001%
21: 0.0001%
32: 0.0001%
41: 0.0001%
51: 0.0001%
659: 0.0001%

random_fisher_f(degrees_of_freedom_1=8.0, degrees_of_freedom_2=8.0)
Time: min: 250ns, mode: 312ns, mean: 301ns, max: 375ns
Floored Distribution:
0: 50.009%
1: 32.6384%
2: 10.2679%
3: 3.7382%
4: 1.5732%
5: 0.7518%
6: 0.3945%
7: 0.2201%
8: 0.1308%
9: 0.0798%
10: 0.0567%
11: 0.0355%
12: 0.0254%
13: 0.0162%
14: 0.0141%
15: 0.011%
16: 0.0062%
17: 0.0065%
18: 0.0046%
19: 0.0033%
20: 0.003%
21: 0.0028%
22: 0.0009%
23: 0.0012%
24: 0.0013%
25: 0.0012%
26: 0.0009%
27: 0.0008%
28: 0.0004%
29: 0.0004%
30: 0.0008%
31: 0.0009%
32: 0.0004%
33: 0.0003%
34: 0.0003%
35: 0.0002%
38: 0.0002%
39: 0.0001%
40: 0.0001%
42: 0.0001%
47: 0.0001%
48: 0.0001%
56: 0.0002%
86: 0.0001%

random_student_t(degrees_of_freedom=8.0)
Time: min: 218ns, mode: 250ns, mean: 241ns, max: 343ns
Rounded Distribution:
-14: 0.0001%
-12: 0.0001%
-11: 0.0002%
-10: 0.0009%
-9: 0.0011%
-8: 0.0018%
-7: 0.0047%
-6: 0.0197%
-5: 0.0724%
-4: 0.3%
-3: 1.4413%
-2: 6.7645%
-1: 22.9623%
0: 36.9402%
1: 22.9035%
2: 6.7485%
3: 1.4407%
4: 0.2963%
5: 0.0736%
6: 0.0177%
7: 0.0062%
8: 0.0024%
9: 0.001%
10: 0.0005%
11: 0.0002%
13: 0.0001%


All tests passed!