Linked Connections Server
The Linked Connections Server enables anyone maintaining public transit data to publish this data for the purpose of route planning. The Linked Connections Server is written in nodejs/express.
Install the server
This server requires Node.js 0.10 or higher and is tested on Linux. To install, execute:
git clone {this repo}
cd repo
npm install
Use the server
Configure the data sources
Copy config-example.json to config.json and fill out all the details. If you have a proxy, such as with varnish, cloudflare, apache or nginx in place, do fill out the proxy details. If you just want to run it on localhost as a test, your can remove the proxy object.
Load connections in MongoDB
First, make sure that you have MongoDB installed.
Now you have two options:
mongoimport --db lc --collection connections --file connections.jsonldstream
mongo lc --eval 'db.connections.find().forEach(function(conn){conn["arrivalTime"] = new ISODate(conn["arrivalTime"]);conn["departureTime"] = new ISODate(conn["departureTime"]);db.connections.save(conn)});'
Now, fill out your config.json with the right collections and mongodb connection string.
From a GTFS file
- Use gtfs2arrdep script to transform a GTFS file into arrivals/departures.
- Use arrdep2connections to convert arrivals/departures into connections and stream directly into MongoDB by using
--mongodb
width the command. - Fill out config.json accordingly
Start the server
nodejs server.js
Background
On today's Web, public transport datasets are disseminated in different ways, which include GTFS zip files and route planning as a service APIs (such as navitia.io, the Dutch railways and many others). In the first case, the dataset is to be downloaded entirely, and routes are intended to be computed locally. Both server as client need to do a considerable amount of work in order to update the data. In the second case, one server handles all possible questions from end-users, but it's hardly possible to combine different APIs to form intermodal routes taking into account all the end-user's requirements.
This proof of concept server offers a type of Linked Data Fragments, a term used to indicate the trade-offs between client and server responsibilities to be made when publishing data, which we will call Linked Connections. Each Linked Connections document consists of a linked list of pages identified by URLs. These pages offer:
- Connection objects starting at a certain timestamp
- controls that lead to other fragments of this dataset (typically the previous and the next page of Connections)
This proof of concept shows that it is worthwhile restudying the trade-offs between server infrastructure and user agent responsibilities: other options are open to be discovered and have advantages and disadvantages of their own.
License
The Linked Data Fragments server is written by Pieter Colpaert.
The code is copyrighted by iMinds - Ghent University and released under the MIT license