GTFS-Hub

This project aims at providing community tested, probably quality/content enhanced, partially merged or filtered GTFS-feeds of (currently German) transport agencies.

Motivation

Since April, 1st, 2020, DELFI e.V. provides a (75%) Germany wide GTFS feed. However, for some use cases only a regional subset is needed, or locally published feeds need to be merged to retain original trip_ids to match e.g. GTFS-RT feeds.

Additionally, the locally published datasets as well as the Germany-wide DELFI GTFS feed sometimes have quality issues or miss e.g. shape information.

While we hope, that all these problems are overcome soon, we currently still see a need to postprocess published data to overcome these shortcomings.

Inner workings

Updating, checking, enhancing GTFS

GTFS-Hub regularly checks on a list of well known GTFS-feeds for updates.

If they have been updated, they get

• downloaded,
• quality checked via mecatran's GTFSVTOR feedvalidator
• optionally enhanced with shapes using OSM data and the pfaedle tool
• optionally transformed with the OneBusAway GTFS transformer tool (fed with a feed specific rule file)
• and optionally merged into larger aggregated GTFS feeds or filtered to a regional subset

Updating and preparing OSM data

Before GFTS data is updated, the OSM data which is used to generate GFTFS shapes is updated. To avoid daily downloading large pbf datasets (GTFS-Hub uses DACH (Germany, Austria, Switzerland) and Alsace (France)) from scratch, we only download the original datasets once and update these afterwards via pyosmium-up-to-date and prepares some region clipped extracts (namely Baden-Wuerttemberg including a buffer of some kilometers around the border).

As this extract will serve as input to OpenTripPlanner as well, we do some additionally data processing on it to enhance some infos, e.g.

• For parkings close to stations with no park_ride tag yet, set park_ride=yes.
• Set some well-known parkings to park_ride=hov.

Publishing

After updating OSM and GTFS data, you'll find the datasets in the folder data/www, ready to publish e.g. via a web serve serving this directory.

External references

This project uses a couple of other Docker images:

• adfreiburg/pfaedle: a dockerized version of Patrick Brosi's tool pfaedle the enhance GTFS feeds by shapes map matched using OSM data
• mfdz/pyosmium: a dockerized version of (py)osmium to quickly update / cut / merge OSM data
• mfdz/osmosis: a dockerized version of osmosis to enhance OSM data with a declarative instruction set
• mfdz/transitfeed: a dockerized version of mecatran's GTFSVTOR feedvalidator.
• mfdz/otp-data-tools: a dockerized version of onebusaway's GTFS transform and merge tools

Thanks to everybody contributing to these tools, the OSM community and Geofabrik and the transit agencies providing the data download services.

How to start gtfs-hub

Prerequisites

You'll need to have Docker installed.

Also, make sure you're using the GNU flavor of make. The BSD make (e.g. on macOS) is not fully compatible with this project.

Running GTFS-Hub

All configuration files necessary for the aforementioned processing steps reside in the config directory. The data will be downloaded into and processing within data.

make all

Running GTFS-Hub within a Docker container

GTFS-Hub's Makefile is designed to work both directly on your machine, as well as within a Docker container. This section describes the 2nd setup.

We'll mount the config & data directories into the GTFS-hub container.

Note that, because the Makefile will run the processing tools via docker run, we'll have to enable a Docker-in-Docker setup by

• mounting /var/run/docker.sock to give the GTFS-Hub container the ability to create new containers
• passing in the path of the host (your machine) GTFS-Hub directory as a $HOST_MOUNT environment variable

docker run -it --rm \
	-v $PWD/config:/gtfs-hub/config:ro -v $PWD/data:/gtfs-hub/data:rw \
	-v /var/run/docker.sock:/var/run/docker.sock -e HOST_MOUNT=$PWD \
	mfdz/gtfs-hub