Python gRPC Geometry Client

Why wrap shapely? We wanted three different abilities. Protobuf support, a required spatial reference and spatial reference aware geometry operations.

Protobuf

We wanted to import, export and operate on protobuf geometries and envelopes. Our microservices communicate using protobufs, so it made sense for us to have a helper library that allowed us to work with the geometries on our client and quickly construct protobufs for messaging.

Spatial Reference

We never want a geometry separate from it's spatial reference details. We could have solved this by working exclusively with the geojson idea of always Wgs84, but we've found time and again that projected geometries exist and must be worked with. So instead of letting geometries float around without their coordinate system information we decided to tie it to the geometry, much like ESRI does with ArcObjects.

Geometry Service and SpatialReference

We wanted easy projections and operators that returned results according to their spatial reference tolerance (geodetic buffer, area, and topo relationships). Most of the old shapely operators behave the same as they did before (with exception of project and generalize), but now they use a remote geometry service that requires spatial reference information. We've kept the native shapely topo operators and changed their names to use the s_ prefix.

Changed Behavior

There are a few differences between how shapely behaves and how this wrapper functions.

• project is now mean to project a geometry from one spatial reference to another
• simplify fixes a broken geometry. To remove vertices from a geometry use s_simplify or generalize
• area is no longer a property. it's a method and it defaults to geodetic (internally it uses geodetic_area). the default result unit is meters squared. to force non-geodetic use geodetic=False
• carto_bounds property is used for the bounds order cartopy prefers
• carto_geom returns a shapely geometry (as does shapely_dump). Cartopy doesn't want wrapped shapely geometries, it wants the real thing.
• buffer defaults to geodetic (like area). The distance you specify is in meters. You can override with geodetic=False
• sr field on geometry returns a SpatialReferenceData protobuf object
• initializing geometry requires a spatial reference, sr, wkid or proj4 definition

Requirements

This requires the gRPC geometry library and docker. You need to build the service as the docker hosted echoparklabs/geometry-service-java is out of date.

git clone https://github.com/geo-grpc/geometry-service-java.git
cd geometry-service-java
docker build -t echoparklabs/geometry-service-java .
docker run -p 8980:8980 -d echoparklabs/geometry-service-java

Install Python gRPC Geometry Client

install using pip (depends on shapely, epl.protobuf and grpc)

pip install epl.geometry

install for development

pip install -r requirements.txt
pip install -r requirements-test.txt

Install and run demo

pip install -r requirements.txt
pip install -r requirements-demo.txt
jupyter notebook

if you have virtualenv installed

python -m ipykernel install --user --name=geometry-client-python

if cartopy crashes (issues 738 and 879):

pip uninstall cartopy
pip uninstall shapely
pip install shapely --no-binary shapely
pip install cartopy --no-binary cartopy
jupyter notebook