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:
- Mapbox, either Valhalla or OSRM
- Openrouteservice
- Here Maps
- Google Maps
- Graphhopper
- OpenTripPlannerV2
- Local Valhalla
- Local OSRM
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 3.8, 3.9, 3.10, 3.11. PyPy3 is currently not supported, see #60.
© routingpy contributors 2022 under the Apache 2.0 License.
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.
Recommended: Install via poetry:
poetry add routingpy
Install using pip
with
pip install routingpy
Or the lastest from source
pip install git+git://github.com/gis-ops/routingpy
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 offeradditional options to create a location element.
|
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 routersoffer 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 routerdependent 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, thiswill return a symmetrical matrix. Usually this is done
with
[Lon, Lat] tuples, but some routers offeradditional 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 totake as sources for the matrix calculation. If not
specified all
locations are considered to be sources. |
|
destinations | The indices of the
locations parameter iterable totake as destinations for the matrix calculation. If not
specified all
locations are considered to bedestinations.
|
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.
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.
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))
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))
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
)
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 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