libosrmc

This repository is Geographica version of libosrmc, focused only on Python 3 binding (osrmcpy). This repository is a fork of https://github.com/daniel-j-h/libosrmc , which is a C wrapper around the C++ libosrm library, useful for writing FFI bindings and guaranteeing ABI stability.

Using with Docker

Building

First you must build base Docker image:

$ cd docker
$ docker build --pull -t geographica/osrmcpy:latest .

Running with Jupyter

You must build your Docker container with Docker-Compose:

$ docker-compose build osrmcpy-jupyter

This container exposes the internal port 8888 to the host port 8889.

Running without Jupyter

You must build your Docker container with Docker-Compose:

$ docker-compose build osrmcpy

This container exposes the internal port 5000 to the host port 5050.

Jupyter up and running

Up Docker container:

$ docker-compose up osrmcpy-jupyter

And and use these examples through JupyterLab in http://localhost:8888:

• notebooks/osrmcpy_compute_matrix.ipynb.
• notebooks/osrmcpy_compute_route.ipynb.
• notebooks/osrmcpy_compute_nearest.ipynb.

Before to use examples you must prepare test datasets (see Data processing).

Data preprocessing

There are two preprocessing pipelines to use OSRM:

• MLD: Multi-Level Dijkstra.
• CH: Contraction Hierarchies. You should use CH when performance is important.

Preprocessing pipelines to generate test data:

• CH pipeline example:

  $ docker-compose exec osrmcpy bash -c 'cd data && ./preprocessing_monaco.sh'

• MLD pipeline example:

  $ docker-compose exec osrmcpy bash -c 'cd data && ./preprocessing_monaco.sh MLD'

You have scripts in data folder to generate test datasets for three locations:

• Monaco (little)
• Berlin (medium)
• Ireland (moderately large)
• Spain (large)
• France (extra large)

Notes on procesing data

Here is the information related to the performance procesing the data for France

8 x Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz
Mem:       16GB
Swap:       2GB

france-latest.osm.pbf  3.43G

--Processing - osm-extract...
RAM: peak bytes used: 12157050880
time: 13m23.498s

--Processing - osrm-contract...
RAM: peak bytes used: 5705818112
time: 40m27.110s

france_osrm_ch/ 7.5G

Using the API

Launch the API pointing to an osrm file

$ docker-compose run --service-ports osrmcpy osrm-routed ./data/osrm/france_osrm_ch/france-latest.osrm

Example

http://localhost:5050/table/v1/car/3.081427,48.809148;3.243775,48.811817;3.228700,48.759358;3.230363,48.800354?sources=0&annotations=distance,duration

Building without Docker

First you need to build and install osrm-backend.

Second you need to build and install C/C++ interface to OSRM.

$ cd libosrmc
$ make
$ sudo make install
$ sudo ldconfig

This compiles the libosrmc.so shared object and installs it into /usr/local (you may have to export LD_LIBRARY_PATH="/usr/local/lib") or install to /usr/lib. The library's interface osrmc.h gets installed into /usr/local/include/osrmc/osrmc.h. You can modify defaults via config.mk.

Last you can install Python3 binding:

$ python -m pip install .

You can test it with this examples:

• osrmcpy/examples/osrm_python3_matrix.py.
• osrmcpy/examples/osrm_python3_route.py.
• osrmcpy/examples/osrm_python3_nearest.py.

License

Copyright © 2018 Cayetano Benavent - Geographica (Python3 binding; more functionalities to C/C++ Interface)

Copyright © 2016 Daniel J. Hofmann (Creator of libosrmc project)

Distributed under the MIT License (MIT).