rsg

RSG - Random Struct Generator

Rsg is a tool that helps creating randomly generated nested data structures, from simple built-in structures as lists, dictionaries and tuples, to custom composite class hierarchies.

In its current state, this library is merely a toy project with a few basic utilities, created with the intent of experimenting with python decorators and language inspection. It was originally a function that generated random yaml files for testing purposes, that quickly grew to become a generic subpackage that had no dependencies and could be exported as a separate python package.

Basics

Generators are subclasses of Rsg with special "generator" methods:

class RsgFooBar(Rsg):
    @generator("foo")
    def generate_foo(self):
        return "foo"

    @generator("bar")
    def generate_bar(self):
        return "bar"

This class genertes either "foo" or "bar" with equal chance:

rsg = RsgFooBar()
[print(next(rsg)) for _ in range(10)]
# bar foo foo bar foo foo bar bar foo foo

You can set a custom chance for each generator method, by setting the keyword agument named <METHOD_NAME>_chance in the object constructor. Probabilities are normalized automatically, so you don't need to ensure they sum up to one.

rsg = RsgFooBar(foo_chance=0.8, bar_chance=0.2)
[print(next(rsg)) for _ in range(10)]
# foo foo bar foo foo foo foo foo foo bar

Chances values all default to 1.0, you can set a custom default value when decorating a generator method. Default values are used only when a chance value is not specified in the object constructor.

@generator("foo", default_chance=0.2)

Parameters

What if a generator method needs parameters? In this example we want to make a generator of random powers of a number (a parameter), with the maximum exponent bounded by anohter parameter.

To do so, just add any amount of parameters to a generator method, you can use both positional and keyword arguments, without the need to implement a custom constructor for your class.

class RsgPowers(Rsg):
    @generator("power")
    def generate_power(self, base: int, max_exp: int = 5):
        exp = randint(0, max_exp)
        return base**exp

To set the declared parameters, just pass them to the class constructor, as if you had actually implemented it!

rsg = RsgPowers(base=2, max_exp=10)
[print(next(rsg)) for _ in range(10)]
# 512 1024 16 128 32 1024 64 8 8 1

In the previous example, the parameter max_exp has a default value of 5, this means that you don't have to explicitly set it in the object constructor. On the other hand, the base parameter doesn't have a default value, and so it must be set explicitly.

Composite Structures

A generator method can be of two types:

• Leaf - Objects with no children, like the ones used so far.
• Composite - Objects with randomly generated children, which in turn can have children recursively.

Let's make a generator of lists-of-lists-of-lists-of-l...

class RsgList(Rsg):
    @generator("list")
    def generate_list(self, children: Iterable[Any]) -> list:
        return list(children)

To make it recursive, you just need to add an argument named children to the method. This argument will contain an iterable of N randomly generated children objects. The children objects are generated with the same generator used for the parent object.

Recusion is governed by the following parameters:

• min_depth - The minimum number of recursions before a leaf object can be generated. Ignored if no composite generators are available.
• max_depth - The maximum number of recursions. Must be >= min_depth.
• min_breadth - The minimum number of children per recursion step. If no leaf generators are available, on the last recursion step min_breadth is forced to 0.
• max_breadth - The maximum number of children per recursion step.

rsg = RsgList(min_depth=1, max_depth=3, min_breadth=2, max_breadth=4)
print(next(rsg))
# [[[[], []], [[], []], [[], []], [[], []]], [[[], [], [], []], [[], []], [[], []], [[], []]]]

Mixin

You can combine any Rsg by simply inheriting from them:

class RsgMixed(RsgFooBar, RsgList, RsgPowers):
    pass

You can still set all parameters of the base classes, remember to set the required ones, or otherwise you will get an AttributeError!

rsg = RsgMixed(
    base=3,
    foo_chance=0.1,
    bar_chance=0.1,
    power_chance=0.4,
    list_chance=0.4,
    min_depth=2,
    max_depth=3,
    min_breadth=2,
    max_breadth=4,
)
print(next(rsg))
# [[[9, 243, 'foo', 'bar'], 27, ['foo', 3]], ['foo', [243, 81, 243]], ['foo', 243]]

Built-in Generators

Rsg provides generators for the following built-in types:

• int - rsg.core.RsgInt generates uniform integer values.
• float - rsg.core.RsgFloat generates uniform float values.
• str - rsg.core.RsgStr generates random strings containing a uniform random number of letters, digits and punctuations.
• list - rsg.core.RsgList generates nested lists.
• tuple - rsg.core.RsgTuple generates nested tuples.
• dict - rsg.core.RsgDict generates nested dicts with valid python identifiers as keys.

In addition, rsg.core.RsgBase combines all previous generators into one:

from rsg.core import RsgBase

rsg = RsgBase(min_depth=1, max_depth=4, min_breadth=1, max_breadth=3)
print(next(rsg))
# [
#     ((0.73354,), (0.12835, '[^U`zq')),
#     [643, ["QiX0@'", [0.95934, 0.27090, 0.66838]]],
#     [{'N_dHLJUZA': 0.00379, 'D0N0uK1_eH': [667, 0.29715], 'SjWKcs_D': {'IMUUZH': 630}}]
# ]