routing-py

One lib to route them all - routingpy is a Python 3 client for several popular routing webservices.

Inspired by geopy and its great community of contributors, routingpy enables easy and consistent access to third-party spatial webservices to request route directions, isochrones or time-distance matrices.

routingpy currently includes support for the following services:

This list is hopefully growing with time and contributions by other developers. An up-to-date list is always available in our documentation.

routing-py is tested against CPython versions 3.7, 3.8, 3.9 and against PyPy3. As other major libraries like numpy and pandas decided to drop Python 2 support, we did not see any reason to burden the project with the compatibility weight.

Why routing-py?

You want to

get from A to B by transit, foot, bike, car or hgv
compute a region of reachability
calculate a time distance matrix for a N x M table

and don't know which provider to use? Great. Then routingpy is exactly what you're looking for.

For the better or worse, every provider works on different spatial global datasets and uses a plethora of algorithms on top. While Google or HERE build on top of proprietary datasets, providers such as Mapbox or Graphhopper consume OpenStreetMap data for their base network. Also, all providers offer a different amount of options one can use to restrict the wayfinding. Ultimately this means that results may differ - and our experience tells us: they do, and not too little. This calls for a careful evaluation which service provider to use for which specific use case.

With routingpy we have made an attempt to simplify this process for you.

Installation

Recommended: Install via poetry:

poetry add routingpy

Install using pip with

pip install routingpy

Or with conda (deprecated)

conda install -c nilsnolde routingpy

Or the lastest from source

pip install git+git://github.com/gis-ops/routing-py

API

Every provider has its own specifications and features. However the basic blueprints are the same across all. We tried hard to make the transition from one provider to the other as seamless as possible. We follow two dogmas for all implementations:

All basic parameters have to be the same for all routers for each endpoint
All routers still retain their special parameters for their endpoints, which make them unique in the end

This naturally means that usually those basic parameters are not named the same way as the endpoints they query. However, all provider specific parameters are named the exact same as their remote counterparts.

The following table gives you an overview which basic arguments are abstracted:

Endpoint	Argument	Function
`directions`	locations	Specify the locations to be visited in order. Usually this is done with `[Lon, Lat]` tuples, but some routers offer additional options to create a location element.
`directions`	profile	The mode of transport, i.e. car, bicycle, pedestrian. Each router specifies their own profiles.
`isochrones`	locations	Specify the locations to calculate isochrones for. Usually this is done with `[Lon, Lat]` tuples, but some routers offer additional options to create a location element.
	profile	The mode of transport, i.e. car, bicycle, pedestrian. Each router specifies their own profiles.
	intervals	The ranges to calculate isochrones for. Either in seconds or in meters, depending on `interval_type`.
	intervals _type	The dimension of `intervals`, which takes router dependent values, but generally describes time or distance
`matrix`	locations	Specify all locations you want to calculate a matrix for. If `sources` or `destinations` is not set, this will return a symmetrical matrix. Usually this is done with `[Lon, Lat]` tuples, but some routers offer additional options to create a location element.
	profile	The mode of transport, i.e. car, bicycle, pedestrian. Each router specifies their own profiles.
	sources	The indices of the `locations` parameter iterable to take as sources for the matrix calculation. If not specified all `locations` are considered to be sources.
	sources	The indices of the `locations` parameter iterable to take as destinations for the matrix calculation. If not specified all `locations` are considered to be destinations.

Contributing

We ❤️ contributions and realistically think that's the only way to support and maintain most routing engines in the long run. To get you started, we created a Contribution guideline.

Examples

Follow our examples to understand how simple routingpy is to use.

On top of the examples listed below, find interactive notebook(s) on mybinder.org.

Basic Usage

Get all attributes

routingpy is designed to take the burden off your shoulder to parse the JSON response of each provider, exposing the most important information of the response as attributes of the response object. The actual JSON is always accessible via the raw attribute:

from routingpy import MapboxValhalla
from pprint import pprint

# Some locations in Berlin
coords = [[13.413706, 52.490202], [13.421838, 52.514105],
          [13.453649, 52.507987], [13.401947, 52.543373]]
client = MapboxValhalla(api_key='mapbox_key')

route = client.directions(locations=coords, profile='pedestrian')
isochrones = client.isochrones(locations=coords[0], profile='pedestrian', intervals=[600, 1200])
matrix = client.matrix(locations=coords, profile='pedestrian')

pprint((route.geometry, route.duration, route.distance, route.raw))
pprint((isochrones.raw, isochrones[0].geometry, isochrones[0].center, isochrones[0].interval))
pprint((matrix.durations, matrix.distances, matrix.raw))

Multi Provider

Easily calculate routes, isochrones and matrices for multiple providers:

from routingpy import Graphhopper, ORS, MapboxOSRM
from shapely.geometry import Polygon

# Define the clients and their profile parameter
apis = (
   (ORS(api_key='ors_key'), 'cycling-regular'),
   (Graphhopper(api_key='gh_key'), 'bike'),
   (MapboxOSRM(api_key='mapbox_key'), 'cycling')
)
# Some locations in Berlin
coords = [[13.413706, 52.490202], [13.421838, 52.514105],
          [13.453649, 52.507987], [13.401947, 52.543373]]

for api in apis:
    client, profile = api
    route = client.directions(locations=coords, profile=profile)
    print("Direction - {}:\n\tDuration: {}\n\tDistance: {}".format(client.__class__.__name__,
                                                                   route.duration,
                                                                   route.distance))
    isochrones = client.isochrones(locations=coords[0], profile=profile, intervals=[600, 1200])
    for iso in isochrones:
        print("Isochrone {} secs - {}:\n\tArea: {} sqm".format(client.__class__.__name__,
                                                               iso.interval,
                                                               Polygon(iso.geometry).area))
    matrix = client.matrix(locations=coords, profile=profile)
    print("Matrix - {}:\n\tDurations: {}\n\tDistances: {}".format(client.__class__.__name__,
                                                                  matrix.durations,
                                                                  matrix.distances))

Dry run - Debug

Often it is crucial to examine the request before it is sent. Mostly useful for debugging:

from routingpy import ORS

client = ORS(api_key='ors_key')
route = client.directions(
    locations = [[13.413706, 52.490202], [13.421838, 52.514105]],
    profile='driving-hgv',
    dry_run=True
)

Advanced Usage

Local instance of FOSS router

All FOSS routing engines can be run locally, such as openrouteservice, Valhalla, OSRM and GraphHopper. To be able to use routingpy with a local installation, just change the base_url of the client. This assumes that you did not change the URL(s) of the exposed endpoint(s):

from routingpy import Valhalla

# no trailing slash, api_key is not necessary
client = Valhalla(base_url='http://localhost:8088/v1')

Proxies, Rate limiters and API errors

Proxies are easily set up using following requests scheme for proxying. Also, when batch requesting, routingpy can be set up to resume its requests when the remote API rate limits (i.e. responds with HTTP 429). Also, it can be set up to ignore API errors and instead print them as warnings to stdout. Be careful, when ignoring RouterApiErrors, those often count towards your rate limit.

All these parameters, and more, can optionally be globally set for all router modules or individually per instance:

from routingpy import Graphhopper, ORS
from routingpy.routers import options

request_kwargs = dict(proxies=dict(https='129.125.12.0'))

client = Graphhopper(
    api_key='gh_key',
    retry_over_query_limit=False,
    skip_api_error=True,
    requests_kwargs=request_kwargs
)

# Or alternvatively, set these options globally:
options.default_proxies = {'https': '129.125.12.0'}
options.default_retry_over_query_limit = False
options.default_skip_api_error = True

codyfcook/routing-py