Welcome to reasonable’s documentation!

reasonable makes it easy to compare floats using the generative testing package hypothesis.

All the documentation is on this one page, so it’s easy to search from your browser.

Quickstart

To use reasonable, replace the floats import from hypothesis with

from reasonable import floats, x

floats is a replacement for the hypothesis floats function.

x is a helper object which makes it trivial to define constraints on the generated floats.

For example, if you were minimising a function and wanted to ensure that all input values above 1.5 will converge on the correct answer, you could write a test like

from hypothesis import given
from reasonable import floats, x
from pytest import approx


@given(floats(x > 1.5))
def test_my_minimisation(n):
    result = minimise(my_func, x0=n)
    assert result == approx(0.23456789)

Alternatively, you could embrace property testing as opposed to the more traditional unit-testing shown above, by validating that the result of the minimisation is less than the result of the objective function for every input

from hypothesis import given
from reasonable import floats, x


@given(floats(x > 1.5))
def test_actually_minimises(n):
    result minimise(my_func, x0=n)
    assert result < my_func(n)

Both of these are useful tests to perform, however neither is easy using just hypothesis because of the float values generated - with the reasonable package, we generate more reasonable float values, allowing simpler testing :)

Floats

The x object

The x object is an instantiated object, not a class or function definition.

It can be imported from the top-level reasonable module:

from reasonable import x

It is a helper object - by using it in a comparison with an int or float, it will set the constraints on the generated values.

This will generate uniformly-distributed random floats between 1.0 and sys.maxsize:

float(1.0 < x)

Gaussian