Add support for topological filters

[huard]

This PR allows users to filter WFS requests using spatial filters: Intercepts, Contains, Within, Touches, Disjoint, Overlaps, Equals.

• Add spatial filters in fes2.py;
• Create gml:Point object;

Notes

• Only works with WFS 2.0 and POST.
• The set_filter method in GetFeature is weird. It expects a GetFeature xml request, instead of just the Filter node. I've added a shortcut to allow passing a filter XML object directly. Suggestions welcome. I suggest that when passing filter to getfeature, we only pass the filter node, not the getfeature node.
• I've tested it on our own geoserver instance, but unclear how to test this actually works without a live instance somewhere.
• The implementation bypasses the FilterRequest class, which seemed to target CSW.
• I used dataclasses, but that's only available from 3.7 onward so added it to dependencies if python < 3.7.

Related

#128

[huard]

Example usage:

from owslib.wfs import WebFeatureService
from owslib.fes2 import Intersects
from owslib.gml import Point
import geojson


def test_intersects():
    p = Point(id="qc", srsName="http://www.opengis.net/gml/srs/epsg.xml#4326", pos=[-71, 46])
    f = Intersects(propertyname="the_geom", geometry=p)
    intersects = f.toXML()
    wfs = WebFeatureService("https://pavics.ouranos.ca/geoserver/wfs", version="2.0.0")
    resp = wfs.getfeature(typename="public:canada_admin_boundaries", outputFormat="application/json",
                   propertyname="PRENAME", method="POST", filter=intersects)
    out = geojson.load(resp)
    assert out["totalFeatures"] == 1
    assert out["features"][0]["properties"]["PRENAME"] == "Quebec"

[tomkralidis]

Thanks @huard. There are no plans to drop 3.6 anytime soon given it is the default Python version on major OSs (i.e. bionic).

[huard]

Ok, will modify the PR.

[huard]

As mentioned, at the moment the <Filter> tag is built into each topological filter. I think ideally, users should be able to compose filters with AND, OR operations. This would require a smarter composition mechanism.

[coveralls]

Coverage increased (+0.4%) to 59.155% when pulling 943f40f on huard:topological_filters into 38c88e5 on geopython:master.

[huard]

@kwilcox @lpinner @jorisvandenbossche Is this what you were looking for in issue #128 ?

[huard]

@tomkralidis Ready for review.

[tomkralidis]

Thanks @huard !

[GabrielFerrariR]

is there any provision for stable implementation of geometry types other than Point?

[huard]

Not from my part. This PR was meant to solve a problem I had.