Generate GeoJSON data designed to expose edge cases in your code. Implemented as a custom extension of hypothesis for Python testing and as a CLI.

pip install hypothesis-geojson

Hypothesis is an advanced testing library for Python. It lets you write tests which are parametrized by a source of examples, and then generates simple and comprehensible examples that make your tests fail. This lets you find more bugs in your code with less work.

Hypothesis is rad; read the manifesto

The key to making great maps is to test your code with ugly data

• Generate GeoJSON using a hypothesis composite strategy.
• The output must be valid GeoJSON (RFC7946).

The edge-features script let's you pipe GeoJSON features to stdout

⭆ ./edge-features 2  # optional, defaults to 10 features
{"geometry": {"type": "Point", "coordinates": [9.1806517e-317, 1.8734333207298816e-300]}, "type": "Feature", "id": 134, "properties": {}}
{"geometry": {"type": "MultiLineString", "coordinates": [[[180.0, -74.59075745399575], [-161.12031862665756, 90.0], [-89.97485098724994, -17.139247698198062]]]}, "type": "Feature", "id": "\u00a0", "properties": {"": null, "196": null, "116": -8.339789954230072e+18}}

From test_features.py

from hypothesis import given

from hypothesis_geojson import features


def find_name(feature):
    """This silly function checks a geojson-like feature
    for the `name` property. 
    
    This assumes that feature['properties']
    is always a dictionary or mapping. But the GeoJSON spec
    allows for `null` (becomes a python `None`) and causes this
    to fail on otherwise valid geojson objects.

    Would you think to test that? Hypothesis does.
    """
    # Get name property or default to empty string
    return feature['properties'].get('name', '')


# To use the hypothesis GeoJSON strategy,
# decorate the test with the @given(feautures())
@given(features())
def test_find_name(feature):
    find_name(feature)

Now put Hypothesis to work finding a GeoJSON feature that will break the function.

⭆ py.test test_features.py
...
Falsifying example: test_find_name(feature={
 'geometry': {'coordinates': (0.0, 0.0), 'type': 'Point'},
 'id': u'',
 'properties': None,
 'type': 'Feature'})

🎉 Hypothesis found the bug.

• At some point we may want to specify more about the aesthetics, geometry validity and spatial relationships between the features (e.g. "Give me 10 rectangular polygons with overlapping edges in this extent").