Skip to main content

Extension to hypothesis to generate inputs for general universal (GU) numpy functions.

Project description

https://api.travis-ci.com/uber/hypothesis-gufunc.png?token=RSemjpisB7uiZv78DVwd&branch=master

This project is experimental and the APIs are not considered stable.

Only Python>=3.6 is officially supported, but older versions of Python likely work as well.

This package includes support for strategies which generate arguments to functions that follow the numpy general universal function API. So, it can automatically generate the matrices with shapes that follow the shape constraints. For example, to generate test inputs for np.dot, one can use,

import numpy as np
from hypothesis import given
from hypothesis.strategies import floats
from hypothesis_gufunc.gufunc import gufunc_args

easy_floats = floats(min_value=-10, max_value=10)

@given(gufunc_args('(m,n),(n,p)->(m,p)', dtype=np.float_, elements=easy_floats))
def test_np_dot(args):
    x, y = args
    assert np.allclose(np.dot(x, y), np.dot(y.T, x.T).T)

test_np_dot()  # Run the test

We also allow for adding extra dimensions that follow the numpy broadcasting conventions. This allows one to test that the broadcasting follows the vectorization conventions:

@given(gufunc_args('(m,n),(n,p)->(m,p)', dtype=np.float_, elements=easy_floats, max_dims_extra=3))
def test_np_matmul(args):
    x, y = args
    f_vec = np.vectorize(np.matmul, signature='(m,n),(n,p)->(m,p)', otypes=[np.float_])
    assert np.allclose(np.matmul(x, y), f_vec(x, y))

test_np_matmul()  # Run the test

Providing max_dims_extra=3 gives up to 3 broadcast compatible dimensions on each of the arguments.

Quick Start/Installation

Simply install with pip:

pip install hypothesis-gufunc

If one would like the same pinned requirements we use during testing, then one can install from the repo with:

git clone git@github.com:uber/hypothesis-gufunc.git
cd hypothesis-gufunc
pip install -r requirements/base.txt
pip install -e .

Running the Tests

The tests for this package can be run by first doing a cd to its root directory and then

./test.sh

The script creates a conda environment using the requirements found in requirements/test.txt.

Hypothesis for Xarray

This package also contains an extension to hypothesis to generate xarray data structures.

To install the package with the xarray dependencies install it with pip as

pip install hypothesis-gufunc[xarray]

Once installed, one can generate a data array as follows:

from hypothesis.strategies import integers, lists
from hypothesis_gufunc.extra.xr import fixed_dataarrays

S = fixed_dataarrays(("a", "b"), coords_st={"a": lists(integers(0, 3))})
S.example()

Here, coords_st allows one to specify a custom strategy for the coordinates on a per-dimension basis. Likewise, if one has known coordinates one can call fixed_coords_dataarrays; or dataarrays if one wants both the dimensions and coordinates determined by the strategy.

The package also has the ability to generate a dataset:

from hypothesis_gufunc.extra.xr import fixed_datasets

S = fixed_datasets({5: ("a", "b"), "bar": ("b"), "baz": ()}, coords_st={"a": lists(integers(0, 3))})
S.example()

One can use fixed_coords_datasets when the coordinates are determined; or simply datasets to have both the dimensions and coordinates generated.

License

This project is licensed under the Apache 2 License - see the LICENSE file for details.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for hypothesis-gufunc, version 0.0.5
Filename, size File type Python version Upload date Hashes
Filename, size hypothesis_gufunc-0.0.5.tar.gz (16.6 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page